EVALUATION OF DYNAMIC SPEED FEEDBACK SIGNS ON FREEWAY INTERCHANGE RAMPS By Md Shakir Mahmud A DISSERTATION Submitted to Michigan State University in partial fulfillment of requirements for the degree of Civil Engineering-Doctor of Philosophy 2022 ABSTRACT EVALUATION OF DYNAMIC SPEED FEEDBACK SIGNS ON FREEWAY INTERCHANGE RAMPS By Md Shakir Mahmud Dynamic speed feedback signs (DSFS) are promising countermeasure to reduce curve speeds and subsequent lane departures at freeway interchange ramps, although their use in such contexts has been limited. Consequently, the impact of DSFS on driver performance at interchange ramps has remained unproven. To that end, research was performed to determine the effect of DSFS installed at freeway interchange ramps on measures of driver behavior, particularly vehicular speeds approaching and entering the ramp curve. To accomplish this objective, a series of field evaluations were conducted at six freeway interchange ramps possessing significant horizontal curvature. Several factors were tested during these evaluations, including: DSFS messaging strategy, positioning of the DSFS with respect to the start of the curve and side of the ramp, physical characteristics of the DSFS, radar detection range, interchange type, and temporal effects. Speed data were collected using one of these three techniques selected based on the site characteristics and types of data desired: 1) a series of high-definition video cameras, 2) handheld LIDAR, and 3) a speed-trailer. Several measures of effectiveness including speed at different locations, speed of drivers at different speed percentiles, and initial braking location were analyzed using appropriate regression techniques. Considering all phases of the field evaluation, it was concluded that DSFS are an effective countermeasure for reducing the speeds of vehicles approaching and entering horizontal curves on freeway exit ramps. The most critical aspect influencing the effectiveness of the DSFS as a speed reduction countermeasure was the longitudinal positioning of the sign relative to the ramp curve. Overall, the presence of a DSFS ii positioned near the start of the curve resulted in curve entry speeds that were, on average, 1.5 mph to 5.3 mph lower than without a DSFS present at the site. When the DSFS was present near the start of the curve, the lowest curve entry speeds were observed for cases where the feedback message activated when vehicles were within 250 to 400 ft of the start of the curve. Regarding DSFS lateral position, both the right-side-mounted and forward-mounted (i.e., ramp gore area) installations resulted in similar curve entry speeds. Furthermore, there were no discernable differences in curve entry speeds between 15-inch and 18-inch display panels, nor were speeds impacted by the inclusion of an advisory speed panel. In terms of feedback message, the most effective strategy was to display the measured speed alternating with a SLOW DOWN message. The effects on driver behavior associated with the DSFS were consistent between system interchanges and service interchanges, and across all vehicle types. The DSFS was equally effective irrespective of the mainline speed limit or ramp advisory speed. However, more pronounced speed reduction effects were observed for faster drivers (i.e., those within the highest quantiles) compared to the slower or average drivers. Furthermore, there was no evidence of temporal changes in driver behavior during the initial 14 months of operation of the permanent DSFS installation evaluated here. Finally, message activation did not show any significant effect on the speeds of mainline (i.e., non-exiting) vehicles. These findings were subsequently utilized towards the development of guidelines and recommendations for future use of DSFS at freeway interchange ramps. iii ACKNOWLEDGEMENTS I would like to acknowledge my advisor, Dr. Timothy Gates for his continuous support throughout my Ph.D., his guidance to achieve goals, and his advice during the down times. Sincere thanks go to my committee members, Dr. Peter Savolainen, Dr. Ali Zockaie, Dr. Mehrnaz Ghamami, and Dr. Dong Zhao for their guidance and support throughout the course of my Ph.D. I would also like to thank the Michigan Department of Transportation (MDOT) for sponsoring this study and MDOT crews for the continuous help during the data collection efforts. In addition, I would also like to thank my family, partner, friends, and the Civil and Environmental Engineering department staff for their continued help and support during my time at Michigan State University. iv This thesis is dedicated to my Mom, Dad, and Wife. Thank you, mom, for loving me like no one else in this world. Thank you, dad, for pushing me for the Ph.D. and always supporting me. Thank you, my dear wife Faria Amin, for being the best partner in every aspect and taking care of me every single day. v TABLE OF CONTENTS LIST OF TABLES…………………………………….………………….……………….…......ix LIST OF FIGURES……………………………...…………………...……......................….…...xi 1.INTRODUCTION AND BACKGROUND ................................................................................ 1 1.1 Lane Departure Crashes ........................................................................................ 1 1.2 Dynamic Speed Feedback Signs and Other Lane Departure Countermeasures .... 5 1.3 Research Problem and Objectives ......................................................................... 6 1.4 Research Tasks ...................................................................................................... 7 1.5 Dissertation Structure ............................................................................................ 8 2. LITERATURE REVIEW AND STATE AGENCY SURVEY................................................ 11 2.1 Policies or Guidelines for DSFS ......................................................................... 11 2.1.1 MUTCD guidance for DSFS ......................................................................... 11 2.1.2 General guidance for the deployment of DSFS ............................................ 13 2.1.3 Location-specific guidelines for deployment of DSFS ................................. 15 2.2 Operational Benefits of DSFS ............................................................................. 16 2.2.1 DSFS in work zones ...................................................................................... 16 2.2.2 DSFS at horizontal curves ............................................................................. 22 2.2.3 DSFS at speed transition zones ..................................................................... 25 2.2.4 DSFS on high-speed arterials ........................................................................ 27 2.2.5 DSFS in school zones .................................................................................... 29 2.2.6 DSFS in residential neighborhoods ............................................................... 30 2.3 Safety Benefits of DSFS ..................................................................................... 32 2.4 Public Perception of DSFS .................................................................................. 33 2.5 Message Type and Position of DSFS Installation ............................................... 34 2.6 State Agency Survey ........................................................................................... 35 2.7 Summary of Literature Review and State-of-the-Practice Survey ...................... 37 3. FIELD EVALUATION METHODOLOGY ............................................................................ 38 3.1 DSFS Selection ................................................................................................... 38 3.2 DSFS Programming ............................................................................................ 41 3.2.1 TraffiCalm sign ............................................................................................. 41 3.2.2 All Traffic Solution sign ............................................................................... 43 3.3 Site Selection ....................................................................................................... 45 3.4 Data Collection Methods ..................................................................................... 47 3.4.1 Data collection using cameras ....................................................................... 48 3.4.2 Data collection using LIDAR ........................................................................ 51 3.4.3 Data collection using speed trailer ................................................................ 54 3.5 Measures of Effectiveness and Analytical Methods ........................................... 55 4. EVALUATION OF MESSAGE TYPE AND LONGITUDINAL SIGN POSITION ............. 58 4.1 Study Design and Site Characteristics ................................................................ 58 vi 4.2 Data Summary ..................................................................................................... 61 4.3 Results and Discussion ........................................................................................ 63 4.3.1 Curve entry speeds ........................................................................................ 63 4.3.1.1 Effect of DSFS position ......................................................................... 65 4.3.1.2 Effect of DSFS messaging strategy ....................................................... 66 4.3.1.3 Effect of vehicle type ............................................................................. 66 4.3.2 Extreme curve entry speeds........................................................................... 67 4.3.3 Initial braking location .................................................................................. 69 4.3.3.1 Effect of DSFS position ......................................................................... 69 4.3.3.2 Effect of vehicle type ............................................................................. 69 4.4 Conclusion and Direction for Further Evaluation ............................................... 71 5. EVALUATION OF LONGITUDINAL SIGN POSITION, INTERCHANGE TYPE, AND TIME OF DAY...... . ..................................................................................................................... 72 5.1 Study Design and Site Characteristics ................................................................ 72 5.2 Data Summary ..................................................................................................... 75 5.2.1 Speed data from cameras............................................................................... 75 5.2.2 Speed data from speed trailer ........................................................................ 76 5.3 Results and Discussions ...................................................................................... 78 5.3.1 Effect of sign position ................................................................................... 81 5.3.2 Effect of interchange type ............................................................................. 82 5.3.3 Effect of time of day...................................................................................... 83 5.4 Conclusion and Direction for Further Evaluation ............................................... 85 6. EVALUATION OF SIGN SIZE, RADAR DETECTION RANGE, AND LATERAL SIGN POSITION……….. ...................................................................................................................... 87 6.1 Study Design and Site Characteristics ................................................................ 87 6.1.1 Sign lateral installation position .................................................................... 88 6.1.2 Feedback message activation location .......................................................... 91 6.2 Data Summary ..................................................................................................... 92 6.3 Results and Discussion ........................................................................................ 93 6.3.1 Effect of sign lateral position ........................................................................ 96 6.3.2 Effect of sign border and display size ........................................................... 97 6.3.3 Effect of message activation location............................................................ 97 6.4 Conclusion and Direction for Further Evaluation ............................................... 98 7. EVALUATION OF SIGN POSITION AND IMPACT TO MAINLINE TRAFFIC............. 101 7.1 Study Design and Site Characteristics .............................................................. 101 7.1.1 Sign test conditions ..................................................................................... 103 7.1.2 Mainline vehicle speeds .............................................................................. 104 7.2 Data Summary ................................................................................................... 104 7.3 Results and Discussions .................................................................................... 105 7.3.1 Effect of longitudinal sign position ............................................................. 107 7.3.2 Effect of lateral sign position ...................................................................... 108 7.3.3 Effect on speeds of mainline vehicles ......................................................... 108 7.4 Conclusion and Direction for Further Evaluation ............................................. 109 vii 8. EFFECTS ON DRIVERS AT DIFFERENT SPEED PERCENTILES .................................. 111 8.1 Study Design ..................................................................................................... 111 8.2 Data Summary ................................................................................................... 112 8.3 Results and Discussion ...................................................................................... 113 8.4 Conclusions and Direction for Further Evaluation ........................................... 116 9. TEMPORAL CHANGES IN DSFS EFFECTIVENESS ....................................................... 118 9.1 Study Design and Site Characteristics .............................................................. 118 9.2 Data Summary ................................................................................................... 118 9.3 Results and Discussions .................................................................................... 119 9.3.1 Curve entry speeds ...................................................................................... 120 9.3.2 Extreme curve entry speeds......................................................................... 121 9.4 Conclusion and Direction for Further Evaluation ............................................. 122 10. CONCLUSIONS AND RECOMMENDATIONS ............................................................... 123 10.1 Conclusions ....................................................................................................... 123 10.2 Recommendations ............................................................................................. 125 10.3 Limitations and Direction for the Future Research ........................................... 128 APPENDIX……… . ................................................................................................................... 129 BIBLIOGRAPHY.. . ................................................................................................................... 137 viii LIST OF TABLES Table 1 Freeway interchange ramp-related crashes in Michigan, 2012 - 2016 .............................. 3 Table 2 Interchange ramps selected for DSFS field evaluation .................................................... 46 Table 3 Descriptive statistics (Site 1) ........................................................................................... 62 Table 4 Multiple linear regression results for curve entry speed (Site 1) ..................................... 64 Table 5 Binary logistic regression results for curve entry speed < 45 mph (Site 1) ..................... 68 Table 6 Binary logistic regression results for braking 200-600 ft prior to the curve (Site 1) ....... 70 Table 7 Site characteristics and DSFS test conditions .................................................................. 73 Table 8 Descriptive statistics: video camera data (Sites 2-4) ....................................................... 76 Table 9 Descriptive statistics: speed trailer data (Site 3) .............................................................. 78 Table 10 Random intercept linear regression results for curve entry speed (Sites 2-4) ............... 80 Table 11 Random intercept logistic regression results for curve entry speed within 15 mph of the advisory speed (Sites 2-4) ........................................................................................... 80 Table 12 Random intercept logistic regression result for brake response location (Sites 2-4)..... 80 Table 13 Random intercept linear regression results for curve entry speed by peak/off-peak period and light condition (Site 3) ........................................................................................ 84 Table 14 Descriptive statistics for ramp vehicles (Site 3) ............................................................ 92 Table 15 Frequency distribution of vehicle detection ranges based on sign positions (Site 3) .... 93 Table 16 Multiple linear regression results for speeds approaching and entering the ramp curve (passenger cars, n=1,983)............................................................................................ 94 Table 17 Multiple linear regression results for speeds approaching and entering the ramp curve (heavy vehicles, n=64) ................................................................................................ 95 Table 18 Descriptive statistics for ramp vehicles (Site 5) .......................................................... 105 Table 19 Descriptive statistics for ramp vehicles (Site 6) .......................................................... 105 Table 20 Multiple linear regression results for curve approach and entry speeds (site 5) .......... 106 ix Table 21 Multiple linear regression results for curve approach and entry speeds (site 6) .......... 106 Table 22 Mainline vehicle speed summary ................................................................................ 109 Table 23 Descriptive statistics for vehicle speeds for different driver types .............................. 113 Table 24 Multiple linear regression results for speeds approaching and entering the freeway exit ramp curve ................................................................................................................... 114 Table 25 Descriptive statistics for evaluation of temporal effects of DSFS ............................... 119 Table 26 Linear regression results for curve entry speed ........................................................... 120 Table 27 Binary logistic regression results for curve entry speed <45 mph ............................... 121 Table 28 Key findings from the statewide survey ...................................................................... 132 x LIST OF FIGURES Figure 1 Innovative speed control treatments used by MDOT ....................................................... 5 Figure 2 State DOT use of DSFS on horizontal curves ................................................................ 36 Figure 3 Dynamic radar speed feedback sign displaying example feedback messages ............... 39 Figure 4 Dynamic radar speed feedback signs tested during the field evaluations ...................... 40 Figure 5 TraffiCalm sign programming using SafetyCalm software ........................................... 43 Figure 6 ATS sign programming using TraffiCloud sign manager software ............................... 45 Figure 7 Example camera setup and field of view provided by each camera............................... 49 Figure 8 A typical two-person LIDAR data collection setup ....................................................... 52 Figure 9 Raw and interpolated vehicle speed data from LIDAR.................................................. 53 Figure 10 Speed trailer at NB US-127 to Round Lake Rd ........................................................... 55 Figure 11 EB Business I-69 Ramp to WB I-69 (Site 1) ............................................................... 59 Figure 12 Example DSFS messaging conditions and installation locations (Site 1) .................... 60 Figure 13 Mean curve entry speeds by DSFS position and messaging strategy (PSGR only) ..... 65 Figure 14 Mean curve entry speeds by DSFS presence and vehicle type .................................... 67 Figure 15 WB I-96 Exit Ramp to SB I-69 (Site 2) ....................................................................... 74 Figure 16 NB US-127 Exit Ramp to Round Lake Rd (Site 3)...................................................... 74 Figure 17 EB I-96 Exit Ramp to 36th St (Site 4) ........................................................................... 75 Figure 18 DSFS and speed trailer at US-127 exit to Round Lake Rd .......................................... 77 Figure 19 Mean curve entry speeds by site and DSFS presence .................................................. 83 Figure 20 Mean curve entry speeds by DSFS presence and time of day ...................................... 85 Figure 21 NB US-127 Exit Ramp to Round Lake Rd .................................................................. 88 xi Figure 22 Forward-mounted vs. side-mounted DSFS installation................................................ 89 Figure 23 Linear regression parameter estimates for the reduction in curve entry speed (with 95% confidence intervals) ........................................................................................... 95 Figure 24 NB US-127 Exit Ramp to Dunckel Rd (Site 5) .......................................................... 102 Figure 25 EB I-96 Exit Ramp to NB US-127 (Site 6) ................................................................ 102 Figure 26 DSFS installation locations (Site 6) ........................................................................... 103 Figure 27 Mean curve entry speeds based on DSFS presence and longitudinal position........... 108 Figure 28 Average vehicle speed profiles for drivers at different speed percentiles .................. 113 Figure 29 Linear regression parameter estimates for the reduction in speed at freeway exit ramps (with 95% confidence intervals) ....................................................................... 116 Figure 30 Average curve entry speeds before DSFS installation and during the 14-months after DSFS installation ........................................................................................................ 121 xii 1. INTRODUCTION AND BACKGROUND 1.1 Lane Departure Crashes Between 2012 and 2016, 1.47 million crashes occurred on public roadways in Michigan, resulting in 4,790 fatalities, 26,367 severe injuries, and 340,441 other injuries (1). Among the most severe types of crashes are those involving lane departure, which occurs when a vehicle crosses over either the roadway centerline or edge line, often resulting in a head-on, sideswipe, or run-off-road crash. These lane departure crashes have been a long-standing safety issue given the severe injury outcomes associated with such crashes. These types of crashes are generally due to a variety of factors, including driver distraction, drowsiness, limited visibility, and poor pavement surface conditions. The initial (1998) AASHTO Strategic Highway Safety Plan (SHSP) identified numerous safety emphasis areas of nationwide importance related to lane departure crashes (2). The SHSP culminated in the publication of NCHRP Report 500, which provided guidance towards lane departure mitigation strategies to reduce head-on and cross-median collisions, keep vehicles on the roadway, and minimize the consequences of leaving the roadway (2, 3). Shortly after the publication of the AAHSTO SHSP, the Michigan Governor’s Traffic Safety Advisory Commission (GTSAC) began the process of developing the initial SHSP for the State of Michigan, which included the identification of 12 traffic safety emphasis areas (4). Lane departure crashes were identified as an emphasis area, and an action plan was subsequently developed. Lane departure crashes have also remained a significant roadway safety problem in Michigan. Although lane departure crashes accounted for 18.0 percent of all crashes in the state from 2012-2016, such crashes accounted for 46.5 percent of all fatalities and 37.4 percent of all serious injuries (1). Not long after the development of the initial SHSP in Michigan, MDOT began several high-profile statewide initiatives aimed at reducing lane departure crashes. These 1 1 initiatives included the installation of rumble strips on non-freeways and cable median barriers on freeways, and each program showed substantial reductions in target lane departure crashes (5, 6). Despite numerous statewide highway safety initiatives to address lane departures, such crashes have continued to occur at relatively steady annual rates. As a result, the prevention of lane departures has remained as a primary emphasis area in each edition of the Michigan SHSP (4, 7, 8). A recent safety evaluation of rural highways in Michigan showed that the risk of a lane departure crash is elevated where horizontal curves are present (9). One particularly vulnerable area for curve-related lane departure crashes is freeway ramps, which, due to right-of-way constraints and other factors, often include horizontal curves requiring a significant reduction in speed to be safely negotiated. A query of Michigan crash data coded as “freeway crash - entrance/exit ramp related” found 39,276 such crashes to have occurred between 2012 and 2016, of which 12,581 (32.0 percent) involved lane departure (1). Table 1 displays these data categorized by crash type versus vehicle type, crash severity, road condition, and lighting condition. Table 1 presents several interesting findings. First, lane departure crashes were overrepresented among severe ramp crashes, accounting for 59.2 percent of total fatal and A-injury crashes on freeway ramps. Furthermore, the specific type of lane departure crash clearly influences severity outcome. Overturning crashes on ramps result in an alarming rate of severe injuries, as 6.5 percent (approximately 1 in 15) result in a fatal or A-injury, compared to 1.2 percent for all other ramp crashes. Consequently, crashes involving an overturn were greatly overrepresented among all severe ramp crashes, accounting for 18.8 percent of all fatal and A-injury ramp crashes, but only 4.1 percent of total ramp crashes. 2 Table 1 Freeway interchange ramp-related crashes in Michigan, 2012 - 2016 VEHICLE TYPE Fixed Object Overturn Other Total Lane Departure Crashes % of Lane Crashes % of Lane Crashes % of Lane Crashes % of Lane Departure Departure Departure Departure Truck/bus over 10,000 lbs 213 45.3% 127 27.0% 130 27.7% 470 100.0% Truck under 10,000 lbs 103 71.0% 32 22.1% 10 6.9% 145 100.0% Passenger Vehicle 8,771 78.8% 1,280 11.5% 1,085 9.7% 11,136 100.0% Other Vehicle 489 56.5% 155 17.9% 222 25.6% 866 100.0% CRASH SEVERITY Fixed Object Overturn Other Total Lane Departure Crashes % of Lane Crashes % of Lane Crashes % of Lane Crashes % of Lane Departure Departure Departure Departure Fatal (K) 27 48.2% 20 35.7% 9 16.1% 56 100.0% Incapacitating (A) 143 53.2% 83 30.9% 43 16.0% 269 100.0% Non-Incapacitating (B) 482 54.2% 286 32.2% 121 13.7% 889 100.0% Possible Injury (C) 1,073 63.9% 426 25.4% 180 10.7% 1,679 100.0% No Injury (O) 7,851 81.0% 779 8.0% 1,058 10.9% 9,688 100.0% ROAD CONDITION Fixed Object Overturn Other Total Lane Departure Crashes % of Lane Crashes % of Lane Crashes % of Lane Crashes % of Lane Departure Departure Departure Departure Dry 3,065 70.7% 849 16.6% 648 12.7% 5,102 100.0% Wet 2,369 76.6% 387 12.5% 337 10.9% 3,093 100.0% Ice/Snow/Slush 3,539 82.3% 348 8.1% 415 9.6% 4,308 100.0% Other/Unknow 63 75.0% 10 11.9% 11 13.1% 84 100.0% LIGHTING CONDITION Fixed Object Overturn Other Total Lane Departure Crashes % of Lane Crashes % of Lane Crashes % of Lane Crashes % of Lane Departure Departure Departure Departure Daylight 5,144 76.5% 834 12.4% 746 11.1% 6,724 100% Dawn/Dusk 480 74.8% 79 12.3% 83 12.9% 642 100% Dark 3,907 75.7% 680 13.2% 573 11.1% 5,160 100% Other/Unknown 45 81.8% 1 1.8% 9 16.4% 55 100% Table 1 also clearly indicates the influence of vehicle type on the risk of an overturn crash. Lane departure crashes were far more likely to result in an overturn if a heavy truck or bus was involved (overturn in 27.0 percent of lane departure crashes) compared to a passenger vehicle (overturn in 11.5 percent of lane departure crashes). However, passenger vehicles were more likely to collide with a fixed object (78.8 percent of lane departure crashes) compared to heavy trucks/buses (45.3 percent of lane departure crashes). Table 1 also provides insight into the environmental conditions that influence the occurrence of a lane departure crash. Lane departure ramp crashes are far more likely to occur during wet conditions and especially ice/snow/slush conditions compared to all other ramp crashes. While only 15.4 percent of non-lane departure ramp crashes occurred during wet road conditions, 3 24.6 percent of lane departure ramp crashes occurred under wet road conditions. Ice/snow/slush road conditions had an even greater influence on lane departure ramp crash occurrence, as 34.2 percent of all lane departure ramp crashes occurred during these conditions, compared to only 10.6 percent of all other ramp crashes. This implies that lane departure crashes were 3.2 times more likely to occur under ice/snow/slush conditions than other types of ramp crashes. Finally, 41 percent of lane departure ramp crashes occurred during darkness, which was more than double the proportion of dark crashes for non-lane departure ramp crashes. Issues with lane departure crashes on interchange ramps in Michigan may be further exacerbated by the 2017 increase in speed limits from 70 to 75 mph for passenger vehicles on over 600 miles of rural freeways and from 60 to 65 mph for trucks and buses on all freeways (10, 11). Recent research found that freeway segments where the aforementioned speed limit increases were applied experienced increases in free-flow speed ranging between 2.1 and 4.6 mph and overall speed increases ranging between 2.0 and 3.1 mph (10). The trend of increasing speed limits on freeways increases the differential between speeds on the mainline freeway and adjacent exit ramps. A recent study evaluated the safety performance of freeway exit ramps, with particular emphasis on the speed differentials that are introduced between mainline traffic and the lower advisory speeds on exit ramps (12). Results indicate this speed difference to have a strong relationship with crash risk, particularly, for lane departure crashes. Approximately 29 percent and 66 percent more crashes were observed at the exit ramps possessing speed differentials with the mainline that were between 40-45 mph and 50-55 mph, respectively, compared to the exit ramps with speed differentials between 25-35 mph. 4 1.2 Dynamic Speed Feedback Signs and Other Lane Departure Countermeasures The Michigan Department of Transportation (MDOT) has deployed various traffic control strategies to mitigate lane departure crashes on interchange ramps, including signs (W1-11, W1- 13, W1-15, W13-6/7, chevrons, etc.), warning beacons (Figure 1a), and chevron pavement markings (Figure 1b). Despite these efforts, the problem continues to persist statewide and new treatments are needed. A recent SHSP engineering action plan for Michigan recommended the implementation of innovative countermeasures to prevent lane departure crashes, including speed control technologies such as dynamic speed feedback signs (DSFS) (13). Utilizing a speed measuring device, typically a radar unit embedded in the sign face, DSFS display real-time feedback to the driver in a variety of formats, including: • the measured speed of the approaching vehicle, • a speed warning message (e.g., “SLOW DOWN” or “TOO FAST”), or • activation of warning lights or beacons on signs with static warning messages. a. Flashing Ramp Warning Sign b. Chevron Markings at Ramp c. DSFS on US-2 Figure 1 Innovative speed control treatments used by MDOT The feedback messages are typically activated only when drivers are exceeding a preset speed threshold. The signs may be programmed to provide different messages based on a preset speed threshold; for example, providing the measured speed below a certain threshold and “SLOW 5 DOWN” above that threshold. Furthermore, combinations of messages may also be used; for example, the measured speed alternating with “SLOW DOWN”. 1.3 Research Problem and Objectives DSFS have been widely implemented across Michigan, although almost exclusively as temporary/portable installations in work zones, school zones, and general municipal speed control applications. Permanent installations of DSFS, such as that previously installed on US-2 west of St. Ignace (Figure 1c), have seen limited use in Michigan and mostly in school zone applications. Although DSFS have been found to be an effective crash reduction countermeasure at horizontal curves (14), they have experienced only very limited installation at such locations in Michigan. Furthermore, prior to the onset of this research project, DSFS had seen very little implementation on freeway interchange ramps both within Michigan and nationwide, and the effectiveness of the signs in such settings has consequently remained unproven. DSFS possess the potential to reduce lane departure crashes on interchange ramps, particularly when used on ramps with substantial horizontal curvature, such as loop ramps at typical cloverleaf interchanges. To this end, research was undertaken to determine effective applications for DSFS when used as a speed control measure on freeway ramps with sharp horizontal curvature. The primary objective of this research was to determine the effect of various DSFS configurations at freeway ramp curves on measures of driver behavior, particularly speed approaching and entering the ramp curve. The findings and conclusions from this study allowed for the development of guidance towards further deployment of DSFS on interchange ramps in Michigan in support of efforts to reduce lane departure crashes and associated injuries and fatalities statewide. 6 1.4 Research Tasks The research described herein evaluated the effectiveness of DSFS when used on interchange ramps in order to provide guidance related to where DSFS should be deployed and what types of DSFS to deploy in this context. The specific tasks of this research study were as follows: • Using a literature review and state agency survey, determine the nationwide state-of-the- art and state-of-the-practice for DSFS, with particular emphasis on deployment at interchange ramps; • Procure and test one or more prototype portable DSFS along with all necessary components for deployment and evaluation at interchange ramps; • Perform a series of field studies at freeway interchange ramps in Michigan to evaluate the impacts of DSFS on the behavior of motorists approaching and entering the ramp curves, including an assessment of various characteristics related to the DSFS, including: o sign messaging strategy, o longitudinal positioning of the sign relative to the ramp curve, o lateral positioning of the sign with respect to the side of the ramp, o sign dimensions and other physical characteristics, o radar activation range, o time of day, o interchange type, o impact on drivers at different speed percentiles, and 7 o changes in driver behavior over time. • Provide recommendations regarding the use of DSFS on interchange ramps in Michigan, including DSFS specifications, installation conditions, and future deployment locations. 1.5 Dissertation Structure This dissertation document has been structured to reflect each of the research tasks. As the field evaluation represented the most extensive task in this study, discussion of this task and the subsequent findings have been divided into multiple chapters based on the evaluated characteristics. The chapter structure for the remainder of this report is provided as follows: • LITERATURE REVIEW AND STATE AGENCY SURVEY (Chapter 2)- This chapter presents a brief overview of the research literature on policies and guidelines for DSFS, and effectiveness of DSFS at different critical locations. • FIELD EVALUATION METHODOLOGY (Chapter 3)- Although DSFS have been evaluated in multiple phases in the field, these evaluations generally employed similar data collection strategies and analytical methods. Consequently, a single chapter was devoted to describing the methodologies for all field evaluations. This chapter includes a discussion of signs tested, sign programming, site selection process, speed data collection using cameras, speed trailer, or LIDAR along with a description of the statistical methods employed. • EVALUATION OF MESSAGE TYPE AND LONGITUDINAL SIGN POSITION (Chapter 4)- This initial field evaluation was performed at a single freeway interchange ramp and was utilized to assess characteristics of the DSFS messaging strategy and sign installation location on various measures of driver behavior, including speed on the 8 approach and entry to the ramp curve and initial braking location. Three feedback messaging strategies were evaluated at three sign positions in advance of the ramp curve. • EVALUATION OF LONGITUDINAL SIGN POSITION, INTERCHANGE TYPE, AND TIME OF DAY (Chapter 5)- This phase expanded the field study to three additional ramps with variable site characteristics to further evaluation the effects of DSFS installation position in addition to interchange type and time of day on driver behavior. • EVALUATION OF SIGN SIZE, RADAR DETECTION RANGE, AND LATERAL SIGN POSITION (Chapter 6)- Further research was conducted at a freeway exit ramp to evaluate several additional aspects of the DSFS, including feedback message display size, sign border type, sign lateral placement position, and radar detection range. Three commercially available speed feedback signs, which varied in size, border type, and radar detection range, were utilized during this field evaluation. • EVALUATION OF SIGN POSITION AND IMPACT TO MAINLINE TRAFFIC (Chapter 7)- An extended field evaluation was conducted at two additional sites to further confirm the effects of longitudinal and lateral sign installation position with respect to the ramp curve. Additionally, the effect of the DSFS activation on mainline vehicles was evaluated. • EFFECTS ON DRIVERS AT DIFFERENT SPEED PERCENTILES (Chapter 8)- A subset of data from Chapter 6 were further analyzed to evaluate whether the DSFS had a variable impact on drivers approaching at different speed percentiles (e.g., faster, average, slower drivers). 9 • TEMPORAL CHANGES IN DSFS EFFECTIVENESS (Chapter 9)- A single freeway interchange ramp was observed periodically for 14 months following the permanent installation of a DSFS to determine if any temporal changes in driver behavior occurred • CONCLUSIONS AND RECOMMENDATIONS (Chapter 10)- This chapter summarizes the findings from all field evaluation phases and provides guidance and recommendations for future use of DSFS at freeway interchange ramps. 10 2. LITERATURE REVIEW AND STATE AGENCY SURVEY Dynamic speed feedback signs (also known as speed feedback signs or dynamic speed display signs) provide active information to the drivers on a digital display by detecting speed of the approaching vehicles. The speed feedback display can be alternated with any feedback message including “SLOW DOWN,” “YOU ARE SPEEDING SLOW DOWN,” “HIGH SPEED SLOW DOWN,” “REDUCE SPEED IN WORK ZONE,” and “EXCESSIVE SPEED SLOW DOWN.” It has been used to reduce speed and crash occurrence in different empirical settings that require a high level of attention from the drivers. The DSFS utilization area includes work zones, sharp horizontal curves, speed transition zones, high-speed arterials, school zones, and residential neighborhoods. To support the use of DSFS in various critical locations, different state agencies have recommended policies to install DSFS, in addition to the general guidelines provided by the Manual on Uniform Traffic Control Devices (MUTCD). 2.1 Policies or Guidelines for DSFS To better understand policies and guidelines for the use of DSFS at different locations requiring speed management, a review of the existing federal and state guidelines was conducted. MUTCD and different state agencies provide guidelines for installing and maintaining DSFS within the rights-of-way of the state-maintained roadways. The policies cover both permanent and temporary installation along with guidance specific to a location including school zones, work zones, and transition zones. 2.1.1 MUTCD guidance for DSFS MUTCD guidance specific to DSFS are provided in sections 2B.13 (Speed Limit Sign) and chapter 2L (Changeable Message Signs). Speed Limit Sign Guidance in section 2B.13 suggested 11 that “If a changeable message sign displaying approach speeds is installed, the legend YOUR SPEED XX MPH or such similar legend should be displayed. The color of the changeable message legend should be a yellow legend on a black background or the reverse of these colors.” The specific guidance for the DSFS (changeable message sign (CMS) in MUTCD) includes (15): • The message sign should be “blank‐out signs that display only single‐phase, predetermined electronic‐display legends that are limited by their composition and arrangement of pixels or other illuminated forms in a fixed arrangement…” • The message displays should be consistent along the roadway corridor and adjacent corridors when DSFS are used at multiple locations. • DSFS used on roadways with speed limits of 55 mph or higher should be visible from 1/2 mile under both day and night conditions. • DSFS message should be legible from a minimum distance of 600 feet for nighttime conditions and 800 feet for normal daylight conditions. • DSFS should be used as a supplement to and not as a substitute for conventional signs and markings. • DSFS message shall consist of no more than two phases and a phase shall consist of no more than three lines of text. • Permanent DSFS should be located sufficiently upstream of known bottlenecks, high crash locations, major diversion decision points such as interchanges to provide adequate response distance. • Permanent DSFS should not be installed within an interchange except for toll plazas or managed lanes, at a location with already high information load on drivers, or at the locations where frequent lane-changing maneuvers are performed. 12 2.1.2 General guidance for the deployment of DSFS Several state agencies have provided guidance on the identification of the need for a DSFS and general guidance to install one. The Vermont Agency of Transportation (VTrans), in their guidelines for the use of radar speed feedback signs on the state highway system recommended a study of the 85th percentile speed to first identify if the posted speed limit is appropriate (16). A comprehensive study by Veneziano et al. (17) recommended several criteria on when or how DSFS can be deployed and operated to address speeding and safety issues effectively. The study suggested, DSFS may be considered when 1. the observed 85th percentile or mean speeds at a site exceed the posted speed limit by 5 mph or more, 2. average daily traffic exceeds 500 vehicles, 3. sites exhibit a correctable speed-related crash history within a recent time period, 4. sites have a pedestrian-related crash history, and 5. the posted speed limit at a site is 25 mph or greater. The Montana Department of Transportation (MDT) and VTrans outlined a list of technical requirements for both permanent and temporary DSFS installation. Both of these agencies recommended to first confirm the MUTCD guidance and standards for DSFS installation. The additional technical requirements for the permanent installation include (16, 18): 1. Installation shall be in conjunction with a speed limit sign (standard or school speed zone), 2. Installation is restricted to one DSFS in each direction for the area being addressed, 3. The changeable display shall be programmed to go blank/no display or an optional word display of “SLOW DOWN” when the vehicle speed exceeds 15 MPH over the posted 13 speed. In either option, the speed of the vehicle will not be displayed when the speed exceeds 15 MPH over the posted speed. 4. When activated, the DSFS display shall give drivers immediate feedback on their individual driving speed when the posted speed is exceeded without animation, rapid flashing, or other dynamic elements. 5. When installed in association with school speed zones, the DSFS shall operate only when the school speed zone is in effect. Use of DSFS in conjunction with school speed zones “when children are present” is not allowed. If this same school zone location experiences documented speed issues, this DSFS can be utilized during ‘non-school-hours’ also. 6. DSFS sign support assembly and installation should meet the requirements for crash- worthiness as defined in the National Cooperative Highway Research Program (NCHRP) Report 350 (19) or Manual for Assessing Safety Hardware (MASH) (20). 7. The installation shall not interfere with the visibility and general effectiveness of any other signs in the area. 8. Identification and contact information for the local government in which it is installed shall be displayed on the case of the DSFS. The technical requirements for the temporary installation of DSFS in the form of a speed trailer include: 1. Speed trailers should be in place for a maximum of 2 weeks; (if needed longer than two weeks, the District Administrator can grant the extra time by email or other documentation) (16, 18). 2. Speed trailers should not replace any other safety measures already in place, i.e., school crossing guards, existing warning signs, enforcement (18). 14 3. Speed trailers shall not interfere with the operation of the highway for motor vehicles, bicyclists, or pedestrians (16, 18). 4. Care should be taken for the placement of the speed trailers in relationship to the clear zone or other signs. This would include placing behind a guardrail and if that is not practicable then placement on the shoulder of the highway with proper delineation of retroreflective temporary traffic control devices such as drums or cones (18). 5. Speed trailers shall include the legend “Speed Limit xx MPH” that is consistent with the regulatory speed limit of the road on which it is used (16, 18). 2.1.3 Location-specific guidelines for deployment of DSFS The FHWA’s ENTERPRISE Pooled Fund Study developed initial planning guidance for several Intelligent Transportation System (ITS) devices to assist agencies to make informed decisions while implementing those (21). A list of warrants for the application of DSFS on the three most common areas was compiled and has focused on the transition zones (19, 20), posted speed adherence (18–20), and intelligent work zones (16, 18–20, 22). The purpose of DSFS in these locations is to promote speed limit adherence in general while focusing on high-speed vehicles or temporary speed reduction due to construction. DSFS should be considered if certain warrants are met at each location. At the speed transition zones, DSFS should be considered if 1. The 85th percentile speed (as determined by a speed study) at a location within the lower speed limit area exceeds the posted speed limit by at least 10 mph, and 2. The zone experiences a posted speed limit reduction of at least 10 mph, and 15 3. There are no other DSFS along the route encountering the speed transition, within 5 miles in either direction (excluding DSFS within school zones). At the posted speed adherence locations, DSFS should be considered if, 1. The 85th percentile speed (as determined by a speed study) exceeds the posted speed limit by at least 5 mph, or by at least 5 mph in a school zone, and 2. The area is within 500 yards of a major pedestrian generator (e.g., school, park, library, senior center, office building) or the area is primarily a residential area or a heavily traveled pedestrian area, and 3. The posted speed limit is 35 mph or less, and 4. There are no other DSFS along the route within 5-mile in either direction of the proposed sign (excluding DSFS within school zones). At the intelligent work zones, DSFS should be considered if, 1. The work zone is currently in operation and observations suggest that the 85th percentile speed at a location within the work zone exceeds the posted speed limit by at least 10 mph or hazardous roadway conditions, such as a temporary unusually tight curve, or a rough road surface, requiring extra driving precaution, and 2. The posted speed limit is 35 mph or greater. 2.2 Operational Benefits of DSFS 2.2.1 DSFS in work zones DSFS was first used as a speed control measure in work zones. It has been used in different forms including portable changeable message signs, speed trailers, and dynamic speed feedback signs. Prior studies have consistently shown the effectiveness of DSFS in reducing speed in work 16 zones. In 1994, Garber and Patel (23, 24) evaluated the change in travel speed by using four different feedback messages at seven work zones on two Interstate highways in Virginia work zones. DSFS were placed on 65 mph highways, intended to drop the speed to 55 mph. While all four types of feedback messages significantly reduced the average speed of vehicles traveling 59 mph or faster, “YOU ARE SPEEDING SLOW DOWN” was the most effective, followed by “HIGH SPEED SLOW DOWN”, “REDUCE SPEED IN WORK ZONE” and “EXCESSIVE SPEED SLOW DOWN”. Results showed a speed reduction of 8 to 10 mph following the installation of the speed feedback signs. The study also found a reduced percentage of vehicles speeding by any amount over 55 mph including by 5 mph and 10 mph. In 1995, McCoy et al. (25) evaluated DSFS in work zones on an interstate highway in South Dakota. The study tested a 20-inch by 28-inch speed display panel in a trailer combined with additional signs including a WORK ZONE warning sign, an advisory speed limit sign, and a YOUR SPEED guide sign. Two DSFS were positioned at the edge of the shoulder on either side and 310 feet upstream of the first taper. Before and after speed data collected from the sites indicated a reduction in average speed by 4 mph for the vehicles with two axles and by 5 mph for the vehicles having more than 2 axles. A significant reduction by 20 to 40 percent in the number of vehicles traveling 10 mph over the speed limit was also observed. In 2001, Pesti and McCoy (26) evaluated the long-term effectiveness of DSFS in work zones on rural interstate highways that required dropping speeds from 75 mph to 55 mph. Three temporary speed trailers showing the driver speed on a panel with 24-inch LED numerals and having a “SPEED LIMIT 55” sign on top of it were installed before three critical segments. The study found a significant reduction of 3 to 4 mph in average speed, 2 to 7 mph in the 85th percentile 17 speed, and 20 to 40 percent increase in the vehicle compliance of speed limit and speed threshold. The results were consistent over the 5 weeks of operation. In 2002, a study by the Maine Department of Transportation evaluated a radar-activated trailer-mounted portable speed feedback sign (27). The sign was installed at the Interstate I-95 work zone that required dropping the speed to 45 mph. The sign was programmed to display “YOU ARE SPEEDING!!!” alternated with “SPEED LIMIT 45 MPH”. A before and after comparison found a reduction in average speed by 7 mph and along with an 11 percent reduction in vehicles exceeding the speed limit. In 2003, a study by Wang et al. (28) evaluated the potential of fluorescent orange sheeting, innovative message signs, and DSFS for reducing speeds in highway work zones. The DSFS displayed “YOU ARE SPEEDING, SLOW DOWN NOW” for vehicles traveling 5 mph over the posted work zone speed limit of 45 mph and displayed “ACTIVE WORK ZONE, REDUCE SPEED” for vehicles traveling below 50 mph. The reduction in average operating speed for fluorescent orange sheeting was 1 to 3 mph, for innovative message signs, it was 0.2 to 1.8 mph. But when DSFS was installed, the average speed was significantly reduced by 7 to 8 mph, in addition to the reduction in speed variance. The influence of the DSFS remained similar throughout the implementation period of three weeks. In 2006, a study conducted by Sorrell et al. (29) in South Carolina work zones evaluated four messaging sequences including “YOU ARE SPEEDING” followed by “SLOW DOWN”, “YOUR SPEED IS______” followed by “SLOW DOWN”, “YOUR SPEED IS______” followed by “THANKS FOR NOT SPEEDING” or “SLOW DOWN”, and “YOU ARE SPEEDING” followed by “MINIMUM FINE $200”. The study found a reduction in average speed by 3 to 10 18 mph. A comparison of the messaging strategies showed that providing positive feedback did not significantly increase the speed reduction, and neither did negative feedback. In 2007, Mattox et al. (30) evaluated the effectiveness of a speed-activated sign at the work zones to reduce speeds on two-lane primary and secondary highways, a multilane divided highway, and an interstate freeway. The 4-feet by 4-feet plastic reflective “YOU ARE SPEEDING IF FLASHING” sign was equipped with radar and flashing lights that get triggered when the speed exceeds by 5 mph. The study showed a significant reduction in average speed by 2 to 6 mph and a reduction in speeding 3 mph over the posted speed limit by 15 to 41.5 percent. In 2009, Medina et al. (31). evaluated the effects of speed feedback trailer, presence of police cars, and speed photo enforcement on drivers’ speed at 1.5 mi downstream of the work zones. While the results demonstrated speed photo enforcement to be the most effective in continuing speed reduction, speed feedback trailer also reduced the average speed by 0.6 to 1.6 mph at one site, and by 0.1 to 1.2 mph at another site. In 2011, Hajbabaie et al. (32) extended the study by Medina et al. (31) and evaluated three work zone treatments that include an SFT, a police car, and a combination of these two. Only the speed feedback trailer reduced the average speed by 0.7 to 2.1 mph. The results also found a reduction of 8.7 percent in the proportion of vehicles speeding by more than 10 mph. In 2012, Roberts and Smaglik (33) used a changeable message sign with radar (CMSR) at a work zone location on a four-lane divided highway in Arizona. The CMSR alternated one speed feedback message, “YOUR SPEED XX MPH,” with another, “POSSIBLE FINE $XXX”. The speed feedback only message reduced the 85th percentile speed by 2 mph compared to “no sign” condition. However, the speed feedback that alternated with the traffic-fine feedback performed better and reduced the 85th percentile speeds by 3 mph compared to the “no sign” condition. 19 In 2013 a study by Thapa et al. (34) used time series traces to analyze the change in driver response from the upstream of work zones to different work zone features on four-lane roadways with both shoulder and lane closure using naturalistic driving study data. The analysis found drivers were 5.07 times more likely to respond when a work zone sign included a DSFS in it. The study also found drivers driving over the speed limit to be more likely to show response than the drives maintaining the speed limit and the likelihood of showing a response increased by 1.06 times with a 1 mph increase in speed over the speed limit. In 2014, a study by Huang and Bai (35) evaluated different messaging strategies on a PCMS at a work zone on two-lane rural highways in reducing speeds while approaching one. The results of the analysis found typical text messages (i.e., WORK ZONE AHEAD SLOWDOWN, FLAGGER AHD PREP TO STOP) to reduce average speed by 13 percent, text message alternating with a graphic depicting the same message at every 3 seconds by 10 percent, and only graphic message to reduce average speed by 17 percent. In the extended study (36), the graphic was redesigned following a survey that showed some confusion in understanding the graphic by around 12 percent of respondents. The redesigned graphic that alternated with text message reduced the average speed by 13 percent. The authors recommended a well-designed graphic to aid the text message to be effective in reducing speeds. The study also reported that 52 to 71 percent of drivers prefer graphics in the PCMS messages. In 2015, a study by Gambatese and Jafarnejad (37) evaluated radar speed feedback display for mobile maintenance operations including relamping, sweeping, vactoring, and spraying in high-speed roadways (speed limit of 55-65 mph) in Oregon. Truck-mounted radar speed signs capable of displaying approaching vehicle speed were deployed. The results showed the sign to be effective in reducing vehicles’ speeds through the work zones. The 85th percentile speed was 20 reduced by 2 to 5 mph when the sign was not in operation, but by 3 to 13 mph when the sign was in operation. Additionally, the study also reported a reduction of 27 to 48 percent in the number of vehicles speeding over the posted limits. In 2017, Zhang and Gambatese (38) evaluated the effect of different traffic-control devices on speed reduction in work zones. The implemented measures were: 1) Static speed-limit (“Speed 50”) signs; 2) “Speed 50” signs and PCMS on trailers; 3) “Speed 50” signs and a radar speed display; and 4) “Speed 50” signs, PCMSs on trailers, and a radar speed display. The results indicated that, at the end of the taper zone, measures 2 and 3 had higher mean speeds than measure 1, whereas measure 4 had a lower mean speed compared to only speed limit signs. However, the locations of the aforementioned measures were not the same on every data collection date, which probably impacted the results to some extent. Overall, the authors concluded that the combination of PCMSs on trailers and radar speed displays is the most effective in reducing vehicle speed immediately at the end of the taper, however, the effect may not be the same over longer distances in a work zone. In 2019, Thapa et al. used time series traces to analyze the change in driver response from the upstream of work zones to different work zone features on four-lane roadways with both shoulder and lane closure using naturalistic driving study data (34). The analysis found drivers were 5.07 times more likely to respond when a work zone sign included a DSFS in it. The study also found drivers driving over the speed limit to be more likely to show response than the drives maintaining the speed limit and the likelihood of showing a response increased by 1.06 times with a 1 mph increase in speed over the speed limit. The same year Banerjee, Jeihani, and Khadem (38) tested three speed reduction measures in work zones using a driving simulator. The control condition represented a standard work zone with the advance warning signs and posted speed 21 limits, and to this setting, the following were added: 1) speed photo enforced signs; 2) DSFS; and 3) reduced speed limit signs. Moreover, four phases of driver speed were tested as they approached and passed the signs: 1) initial speed area - an arbitrary distance of 375 feet before the start of the next phase; 2) visible area - the distance at which the sign became visible; 3) readable area - the distance at which the sign became legible; and 4) post sign area - a distance of 250 feet after each measure. A total of 66 participants, ranging from 18 to 65 years of age, were included in the study. While the study reported photo-enforced signs to be the most effective in reducing speed, DSFS was also found effective. The results indicated that participants reduced their speeds gradually once the DSFS became visible and readable and continued after passing the sign as well. The overall reduction in speed compared to the second-most effective sign (DSFS) was 4 mph. In 2021, a study by Anderson et al. (39) evaluated DSFS on high-speed work zones on Kansas roadways. A speed feedback sign was installed on 70 mph roadways having a work zone speed limit of 55 mph. The sign was installed with a static work zone speed limit sign, and it was capable of displaying the speed of the approaching vehicle. The study found DSFS to be effective in reducing speeds at one site by reducing the average speed by 2 mph. 2.2.2 DSFS at horizontal curves DSFS have been successfully implemented on horizontal curves at several locations and they have been found to improve the level of safety on horizontal curves by reducing speeds and crash rates. In 2000, a study by Tribbett et al. (40) evaluated five dynamic curve warning signs equipped with speed detection radar installed on Interstate 5 in the Sacramento River Canyon. The sites had curve advisory speeds between 50 and 60 mph and upstream speeds of 55 mph and 65 mph. The findings of the study were mixed as some sites experienced a reduction in average speed while others experienced no change or increase. The results also varied for different vehicle types. 22 However, there have been some concerns about the stopwatch method utilized for the speed data collection. In 2005, a study by Ullman and Rose (41) evaluated two horizontal curve sections on a 55 mph speed limit roadway with 20 mph advisory speeds. The study found average car speed to drop by 2.1 to 3.5 mph in the short-term. The impact on the truck speed was not similar, where one site experienced a slight increase in average speed and another one experienced a slight decrease. However, at both study sites, the percentage of vehicles exceeding posted speed limit significantly decreased both in short-term and long-term. Speeding over the curve advisory speeds dropped by 26 percent for cars and 28 percent for trucks at one site, where it dropped by 13 percent for cars and 24 percent for trucks at another site in the short-term. In 2006, a before-after study by Bertini et al. (42) evaluated two overhead speed feedback signs on a horizontal curve in Interstate 5 in Oregon. The overhead sign displayed the fastest speed within the detection zone. The message displayed “SHARP CURVE AHEAD” when vehicle speeds are less than 50 mph, “YOUR SPEED XX MPH” when detected speeds are 50 to 70 mph, and “YOUR SPEED IS OVER 70 MPH” when detected speeds exceeded 70 mph. Even though the feedback was not for an individual vehicle, the strategy significantly reduced average speed as the study found a reduction in average car and truck speed by 3 mph in one direction and by 2 mph in another direction. In 2015, a nationwide DSFS study on horizontal curves by Hallmark et al. (43) on 22 two- lane rural horizontal curve sections evaluated two different speed feedback signs. The signs include a speed display sign showing the speed of the approaching vehicle and posted speed limit when speed is 20 mph over the posted speed limit, and a curve warning sign showing a curve sign and an alternating slowdown message to the vehicles exceeding the 50th percentile speed. The 23 results showed a significant reduction in average speed by 1.82 mph, 2.57 mph, and 1.97 mph after 1 month, 12 months, and 24 months of installation, respectively. The results also showed a significant reduction in the percent of vehicles traveling 5 mph over the advisory speed or posted speed limit by 11.8 percent, 18.6 percent, and 19.8 percent after 1 month, 12 months, and 24 months of installation, respectively. The reduction in the percentage of vehicles exceeding the advisory or posted speed limit by 10 mph, 15 mph, or 20 mph was even higher. The same year, using a driving simulator, Montella et al. (44) evaluated a variety of measures intended for warning drivers about approaching the horizontal curve at a potentially unsafe speed. A total of ten scenarios were created in which the following measures were tested: 1) curve warning sign; 2) curve warning sign and flashing beacons; 3) curve warning sign and DSFS; 4) curve warning sign and transverse rumble strips; 5) curve warning sign and colored transverse strips; 6) curve warning sign, transverse rumble strips, and dragon teeth markings; 7) curve warning sign, transverse rumble strips, and colored median island; 8) curve warning sign, transverse rumble strips, and chevrons; 9) curve warning sign, transverse rumble strips, chevrons, and fluorescent sheets; and 10) curve warning sign, transverse rumble strips, chevrons, and sequentially flashing beacons. When it comes to DSFS, the sign appeared 75 meters before the curve and displayed the driving speed 150 meters before the curve (the sign was activated only if the driver exceeded the speed limit). Overall, the DSFS reduced speed from 88.7 km/h to 82.5 km/h at the start of the curve after installation. In 2022, Mahmud et al. (45) evaluated the effect of DSFS on rural highways with 65 mph mainline speed limit approaching horizontal curves with curve advisory speeds varying between 25 and 60 mph. The study evaluated the effect of DSFS installed at two different locations: 1) at PC, and 2) at curve advisory sign. The study reported a reduction in curve entry speed by 0.1 to 24 2.8 mph when DSFS was installed at PC, and a reduction of 0.7 to 3.5 mph when installed at the curve advisory speed sign. The study also reported a 25 to 40 percent reduction in the number of vehicles traveling 10 mph over the curve advisory speeds. The study finally suggested installing the sign in advance of the PC and at the locations where the difference between upstream and curve advisory speed is 25 mph or more. 2.2.3 DSFS at speed transition zones DSFS has been successfully installed at the speed transition zones with various upstream speeds transitioning to different reduced speed limits. It has been successful to effectively convey the information pertaining to the pending reduced speed limit to approaching drivers. In 2005, a study by Ullman and Rose (41) evaluated two transition zones dropping the speed limit from 55 mph to 45 mph. After the installation of the DSFS, the average speed dropped by 3.4 mph and 2.6 mph at two sites in the short-term and by 1.4 mph in the long-term at both sites. The study also found the drivers traveling above the posted speed limit to significantly reduce their speed compared to the drivers complying with the speed limits. Sandberg et al. (46) evaluated the long-term effects of DSFS on 7 speed transition zones along two-lane rural highways in Minnesota. The selected sites had an upstream speed between 45-55 mph and transitioned into areas with reduced speed limits between 30-45 mph. DSFS was installed at the reduced speed limit sign and data were collected after 1 week, 2 months, 7 months, and 12 months. The results showed a significant decrease in the average, 50th, 85th, and 95th percentile speeds, and the 10-mph pace, particularly at the sites with higher speed differential between the upstream and reduced speed limits. Across all the sties evaluated, the 85 th percentile speed dropped by 7.0 mph, 7.5 mph, 6.3 mph, and 6.8 mph after 1 week, 2 months, 7 months, and 25 12 months of installation. The authors concluded DSFS to be effective in the long term at speed transition zones. In 2009, a study by Cruzado and Donnell (47) evaluated a total of 12 speed transition zones on two-lane rural highways in Pennsylvania to evaluate the effectiveness of DSFS in reducing the speed while entering the rural communities. The study sites had upstream speed limits between 45-55 mph, which transitioned to reduced speed limit areas with speeds between 25-40 mph. Results showed free-flowing average passenger car speed reduction in the transition zones to increase by 6.3 mph following the installation of the DSFS. The effectiveness of the DSFS in reducing the average free-flowing speed continued during the time DSFS was activated but faded as soon as they were removed. In 2015, Hallmark et al. (48) evaluated different types of DSFS installed at the transition zones in three small rural communities in Iowa. A simple feedback sign displaying only drivers’ speeds at a transition site from 55 mph to 25 mph found a decrease in average speed by 8 mph and driving 5 mph over the speed limit by 45 percent after one month of installation. A similar setup including a static “YOUR SPEED” sign and a separate display showing the driver’s speed found a decrease of 5 mph in average speed one month after the installation. Another DSFS capable of showing alphanumeric messages was installed at a 55 mph to 25 mph transition zone and programmed to display vehicle speed when the approach speed was between 26 and 39 mph and display “Slow Down 25” when the approach speed was between 40 mph and 75 mph. The average speed decreased by 5 mph and vehicles exceeding the speed limit by 5 mph decreased by 76 percent. Another type of DSFS assembly included a speed limit sign with embedded LED lights around the outside of the sign and found inconsistent results, where one sign location experienced 26 only a 0.4 mph decrease in average speed and another one had a 6 mph decrease in average speed following the installation. In 2022, Mahmud et al. (49) evaluated the effect of DSFS at 5 speed transition zones along high-speed rural highways. The DSFS was positioned in advance of the reduced speed limit sign and programmed to display approaching vehicle speeds alternating with reduced speed limit information. The study reported the greatest speed reduction to occur at the DSFS location, where speeds were 3.2 to 7.8 mph lower when the sign was in operation. At the reduced speed limit sign, the reduction in speed ranges between 3.6 and 6.2 mph. Along with that, drivers were found to be 78.8 to 92.4 percent less likely, on average, to exceed the speed limit when the DSFS was present. 2.2.4 DSFS on high-speed arterials DSFS has been installed and evaluated on high-speed arterials as a speed regulating feature, particularly in advance of the intersections, or horizontal curves. In 2005, a study by Ullman and Rose (41) evaluated two high-speed roadways with target speed limits of 55 mph and 45 mph in advance of signalized intersections. The study found the average speed to drop by 3.4 to 3.6 mph in the short-term. However, in the long term, the speed reduction effects were mixed, where at one site average speed returned to the prior conditions, and at another, average speeds were 4.0 mph lower than the prior conditions. In 2008, a study by Walter and Broughton (50) evaluated 10 sites on two-way single carriageway roads with a 30 mph speed limit. The study found a significant reduction in average free-flow speed by 1.4 mph and a 12 percent decrease in vehicles exceeding the speed limit. However, speeds return to their previous levels after the removal of DSFS from the sites. In 2009, the city of Bellevue’s Transportation Department (22) evaluated 11 arterial streets with curves having posted speed limits of 30 or 35 mph. Results showed a significant reduction in 27 the 85th percentile speeds during multiple years of observations ranging from 1-6 years. The reduction in the 85th percentile speeds was between 2.0 to 6.3 mph over the years. In 2014, a study by Ardeshiri and Jeihani (51) evaluated both short-term and long-term effects of DSFS showing only speed numbers on arterial roads with speed limits of 25 mph, 35 mph, and 45 mph. The study found an increase in speed limit compliance by 5 percent and speed reduction in 40 percent of cases. However, the study also found drivers increased their speed after passing the DSFS, and DSFS lost its effectiveness in the long-term. The study suggested the implementation of the DSFS only at critical locations (i.e., locations with high crash rates, school zones, work zones) and should be supplemented with occasional speed enforcement. In 2020, Krimpour et al. (52) evaluated a major signalized arterials in Arizona with a speed limit of 45 mph. A total of 4 DSFS were installed along the corridor to quantify the impact of DSFS on the link and intersection level. The study found no significant difference in the signal performance, but significantly lower average speed (average reduction of approximately 1 mph) at three out of the four links following the installation of the DSFS. In 2021, an extended study by Krimpour et al. (53) evaluated speed feedback signs in combination with law enforcement at nine high-speed arterials with speed limits ranging from 40 to 50 mph in Arizona. The results showed that only speed feedback sign decreases average speeds by 0.8 mph to 5.8 mph at two sites. However, at one site with a higher speed limit (50 mph) average speed increased at the location with a speed feedback sign. Additionally, when the speed feedback sign was supported with periodic law enforcement, reductions in average speeds continued beyond the location of the speed feedback sign. A decrease of 0.3 mph to 2.5 mph at the speed feedback sign and an additional decrease of 2.5 mph to 3.5 mph beyond that point was observed with the presence of enforcement. 28 2.2.5 DSFS in school zones DSFS has been widely used in the school zones to implement lower regulatory speed limits during school arrival and dismissal times around the school zones. The sign has been effective in conveying the reduced speed limit information and subsequently reducing school zone approaching speeds. In 2005, a study by Ullman and Rose (41) evaluated DSFS installed in advance of 3 school zones with speed limits of 35 mph to 45 mph. The study found a short-term average speed reduction of 9.2 mph and a long-term speed reduction of 8.8 mph following the installation of DSFS. Additionally, the number of drivers exceeding the speed limit also dropped from 95 percent to 34 percent and 44 percent in the short-term and long-term after the installation of the DSFS, respectively. In 2006, Ash (54) evaluated four different school zones with a 35 mph approach speed and 20 mph school zone speed limit in Utah following the installation of speed feedback signs. A 24- inch by 30-inch “YOUR SPEED” display sign displayed driver speed and small LED lights in the number provided flashing sign when the speed was 5 mph over the school-zone speed limit. The short-term effect after two months found drivers to drop their speed to or below the speed limits. The average speed reduction in this study was between 1 to 2 mph. In the long-term after six months, though average speed values slightly increased compared to the two-months after results, it was still around the school zone speed limits. In 2012, a study by O’Brien and Simpson (55) evaluated a DSFS installed in a North Carolina school zone as part of the Safe Routes to School (SRTS) program. The DSFS assembly included a “SCHOOL” sign, speed limit sign, school hours sign, and “YOUR SPEED” sign with 29 a display capable of showing approaching vehicle speeds. The study found a significant reduction of 3 to 4.5 mph over a 12-month post-installation period. In 2016, Williamson et al. (56) studied the long-term effectiveness of speed display signs in a university environment. The study found that 85.6 percent of drivers decrease their speed immediately after the deployment of the speed display sign and the percentage was 80 percent after one year, suggesting the long-term effect of the sign. 2.2.6 DSFS in residential neighborhoods DSFS has been installed as part of the traffic calming measure in residential neighborhoods and the effect was positive. The sign was usually installed in the residential areas when other traffic calming measures were not effective in reducing speeds to the intended levels. In 1998, Bloch (57) evaluated DSFS in the residential areas with or without enforcement and compared the results with photo-radar. The study was conducted on three sites in Riverside, California, along two-lane, residential roads with speed limits of 25 mph. Results from the study indicated an average speed reduction of 6.1 mph at the location of the speed trailer. The speed reductions downstream of the trailer were 2.9 mph and 5.9 mph without and with enforcement, respectively. One week after the removal of the speed trailer, speed reductions of 0.6 mph (at the former trailer location) and 1.7 mph (downstream) were observed for deployments that did not coincide with enforcement. Where enforcement was used in conjunction with the speed trailer, one week after sign removal, speed reductions of 0.6 mph occurred both at the trailer location and downstream of the trailer. In 2005, a study by Chang et al. (58) evaluated four 24-inch by 30-inch radar signs installed at the 25 mph residential neighborhood. The sign was assembled under the speed limit sign and designed to display drivers' speeds and to start blinking once the speed reaches 5 mph over the posted speed limit. The study found a significant reduction of 1.2 to 2.2 mph at three out of the 30 four sites. One site experienced an increase of 0.5 mph following the installation of the DSFS, however, this site had the lowest prior average speed suggesting the implementation of DSFS at the locations with greater speeding-related issues was more effective. In 2009, the city of Bellevue’s Transportation Department (22) evaluated 20 residential streets with a speed limit of 25 mph. The study showed a significant reduction in the 85th percentile speeds between 0.3 to 6.8 mph during multiple years of observations ranging from 1-7 years. In 2012, a study by Gehlert et al. (59) evaluated three types of DSFS messaging strategies including a standard DSFS showing driver’s actual speed, a standard DSFS showing driver’s actual speeds highlighted in red or green depending on whether the driver complied with or exceeded the speed limit, and a verbal colored DSFS that shows “THANK YOU” message in green letters when the speed is within the speed limit or “SLOW” in red letters when the driver exceeded the speed limit. The study was conducted on a residential road with a speed limit of 30 kmph (18.6 mph). Results showed providing verbal feedback was the most effective in reducing the average speed, followed by showing derivers’ speed in red or green color and showing just the driver’s speed. Results indicate that providing a hint of the drivers’ action along with the personalized speed feedback was more efficient. In 2016, a study by Churchill and Mishra (60) evaluated a trailer-mounted speed feedback sign on residential roads with 50 kmph (31 mph) and 60 kmph (37 mph). The study found the average speed to significantly reduce by 1.6 to 5.6 mph during the operation. The study also found a significant reduction in the percentage of vehicles traveling over speed limits. A permanent installation of a speed feedback sign on 30 kmph (18.6 mph) roadways found a significant reduction in the percentage of vehicles triggering the sign as the sign was programmed to be triggered for only the vehicles with a speed of 35 kmph (21.8 mph) or above. The study indicated 31 that putting a lower threshold for sign triggering works better in reducing average speed. The results of long-term evaluation of the sign were inconclusive. In 2020, a case study in the City of Campbell, CA by Jue and Jarzab (61) analyzed the effectiveness of a radar speed sign over 5 years. The study deployed 30-inch by 42-inch speed feedback signs at 10 different locations with speed limits of 25 mph (one site had variable speed limits of 25 and 35 mph). The average reduction in speed after three months was 0.5 mph, after 6 months was 0.4 mph, after one year was 0.8 mph, after 3 years was 0.5 mph, and after 5 years was 0.2 mph. All the local streets had reduced speed and one collector street had increased speed following the installation of the speed feedback sign. 2.3 Safety Benefits of DSFS There have been relatively few studies evaluating the safety benefits of DSFS. One reason may be the limiting number of locations used to study the operational effect of DSFS in every setting, in addition to the installation of DSFS mainly for the short-term speed reduction purpose in many instances. However, the limited number of studies that evaluated the safety effects of DSFS reported a significant reduction in crashes. In 2000, a study Tribbett et al. (40) in California evaluated the safety effects after installing five dynamic curve warning signs. The study reported a reduction in total truck-related crashes and mixed results for passenger vehicle crashes. However, the study could not draw a conclusion due to limited after-period crash data. In 2015, a study by Hallmark et al. (14, 43) conducted a comprehensive crash evaluation after the installation of DSFS at horizontal curves on two-lane rural highways. The study evaluated twenty-two study sites with two different DSFS systems and 37 control sites (similar sites where no DSFS was installed) to develop crash modification factors using Bayes modeling approach. 32 The research found a 5 to 7 percent reduction in crashes during the first three years after the installation of DSFS. In 2020, an empirical Bayes (EB) analysis by Wu et al. (62) on 192 DSFS installed on arterial and collector roads within the city of Edmonton, Alberta, Canada showed a significant reduction in crashes of all severity. The before-after EB analysis found a significant reduction in collisions that ranged from 32.5 percent to 44.9 percent with the greatest reduction in speed-related severe crashes. The overall reduction in total crashes was 36.1 percent including a reduction in rear-end by 38.0 percent, improper lane-changing by 32.5 percent, and speed-related by 38.2 percent. A detailed economic analysis using three different methods including direct costs, human capital, and willingness to pay further showed the benefit-cost ratio of installing DSFS ranged from 8.16 and 20.19 for 2-year service life and 19.84 to 49.06 for 5-year service life. 2.4 Public Perception of DSFS Driver’s acceptance and perception of the sign can also impact how their driving behavior will be changed. Several studies conducting a public perception survey of DSFS reported positive feedback on the sign. A survey done in California as part of a DSFS study reported an average of 80.0 percent of drivers to find the information provided by the sign to be useful (40). A questionnaire survey conducted on the students of Morgan State University revealed most of the respondents reduce their speed to the speed limit when they encounter a DSFS (63). The perception towards the DSFS was also very positive as 82.0 percent of the responses suggested DSFS increases safety or improves traffic flow or both. Another questionnaire survey on the use of DSFS in school zones conducted in Utah found very positive responses from the drivers (54). The survey found that 72.9 percent of drivers believe speed feedback signs increased awareness 33 of the potential danger ahead and 84.3 percent considered the signs effective for speed reduction in school zones. 2.5 Message Type and Position of DSFS Installation The effectiveness of dynamic speed feedback signs depends on their positioning and message. The location of warning signs is an important factor, not only to provide enough prior warning to the driver but also because drivers tend to increase their speed once they are past the sign. On straight sections of roadway, the greatest speed reduction has been observed to be 1200 feet to 1400 feet upstream of the sign (50, 64) and speeds began to increase 300 feet to 500 feet downstream of the sign (64). This speed increase shortly after the sign leads researchers to recommend their installation be at critical points on the roadway where safety is of the utmost importance (51). Another study evaluated the spatial effectiveness of the DSFS using vehicle trajectory data on a rural two-lane highway in Wisconsin (64). DSFS with flashing speed readings have shown greater speed reductions in advance of the sign compared to beyond the sign, with the greatest speed reductions observed 1200 - 1400 feet upstream of the DSFS and diminished effects 300 - 500 feet beyond the sign (64). The study suggests that once drivers pass the DSFS, the sign loses its effectivity significantly. A driving simulator study by Zhao et al. (65) also evaluated the effect of warning sign position on driving behavior in a sharp horizontal curve. The optimal location of DSFS on horizontal curves was found to be about 330 feet to 650 feet prior to the curve (65). The type of feedback message has also been shown to affect the DSFS effectiveness. Full matrix displays that are able to provide messages such as ‘SLOW DOWN’ when the driver is exceeding the speed threshold have been found to outperform signs that simply display the driver’s speed (30, 41–43, 59). 34 2.6 State Agency Survey A state DOT was developed to ascertain the state of the practice in terms of the deployment of DSFS, including at freeway interchange ramps. The questionnaire (found in Appendix) included general questions on DSFS utilization by roadway type and context, sign specifications, sign calibration, message display rules, the ability for the sign to communicate with an operations center or sensors, site selection, sign installation position with respect to the curve, measures to limit interference from mainline traffic, maintenance requirements, and lessons learned. The survey was sent to 49 state DOTs in the United States (excluding Michigan). A total of 22 responses were received. As reflected in Figure 2, 9 states indicated that DSFS were not utilized on highway curves, 13 indicated the utilization of DSFS as a speed control measure on highway curves, although only four of these states (Illinois, Mississippi, Pennsylvania, and Wisconsin) indicated DSFS use on freeway exit ramp curves. Note: although the survey was not sent to the Michigan DOT, DSFS have recently been utilized on freeway ramp curves within Michigan, and is indicated in the map as such. Detailed survey responses are provided in Appendix. 35 Figure 2 State DOT use of DSFS on horizontal curves A particularly noteworthy comment mentioned by several agencies that use DSFS on ramp curves was that interference from adjacent mainline traffic is common, which frequently causes unnecessary activation of the feedback sign, but may also cause missed activations for ramp vehicles. Interference from mainline traffic occurs because the forward-facing radar cannot isolate ramp traffic from the mainline, and is most common at locations where the ramp runs immediately adjacent to the mainline. Panels designed to occlude mainline vehicles from the radar have proven ineffective. Narrow band radar (e.g., 10 degrees or less) may help, but is not common in “off the shelf” signs. Some agencies use a maximum speed threshold (65 mph) to minimize interference 36 from mainline vehicles. However, caution must be exercised such that ramp vehicles approaching at excessive speeds are not excluded, which becomes less of an issue when the DSFS is posted near the start of the curve. 2.7 Summary of Literature Review and State-of-the-Practice Survey Collectively, the research literature suggests that DSFS are effective for reducing speeds and subsequent speed-related crashes across various roadway contexts, including rural highway curves. However, no prior studies considered the effects of DSFS when installed at interchange ramp curves. Furthermore, guidance towards the use of DSFS provided within the MUTCD and state agency specifications does not specifically focus on freeway ramps applications of DSFS. Of the 23 state agencies responding to the survey on DSFS use, only five indicated experience with using DSFS at freeway ramp curves. Several agencies noted issues associated with mainline traffic interfering with the operation of feedback signs when used at freeway exit ramps. Given the recent freeway speed limit increases coupled with the persistent lane departure crash issues on interchange ramps in Michigan, additional research was warranted to evaluate the effects of DSFS on driver behavior and provide MDOT with guidance and direction towards further deployment of DSFS at freeway interchanges. 37 3. FIELD EVALUATION METHODOLOGY A series of field evaluations were performed at multiple freeway interchange ramps to assess the impacts of speed feedback signs on various measures of driver behavior, particularly speed while approaching and entering the ramp curve. It was ultimately intended for the results of these field evaluations to inform future deployment of DSFS at freeway interchange ramps across Michigan. Thus, the field evaluations were performed across multiple phases, each of which assessed important aspects related to the design, operation, and/or installation of the DSFS. The following subsections provide a general overview of the field evaluations, including DSFS selection, DSFS programming, site selection, field data collection techniques, and analytical methods. Specific details related to the individual field evaluations are provided in subsequent chapters. 3.1 DSFS Selection An initial step of the research was to identify the specific signs to be used in the field test. The research team contacted vendors and other state DOTs with experience using DSFS at horizontal curves to identify several viable sign options, which were presented to the MDOT research advisory panel for review and vetting. Each of these signs was compliant with MDOT’s draft special provision for DSFS, which called for a yellow “YOUR SPEED” sign with an embedded radar and a speed feedback panel that possessed the ability to display either the speed digits or a “SLOW DOWN” message to approaching vehicles. The MDOT research advisory panel collectively decided to select a sign that MDOT had previous experience with and was a likely candidate for future installations. The selected sign was 40-inches by 31-inches with microprismatic reflective yellow sheeting with black “YOUR SPEED” text and a full matrix amber 38 LED feedback display capable of displaying characters of up to 15 inches in height. This test sign, which was manufactured by TraffiCalm, is shown displaying example messages in Figure 3. In addition to the 15-inch TraffiCalm sign that was utilized in the preliminary field evaluations, two additional signs were included in the field evaluations performed during the later phases of this project. This included a larger version of the aforementioned TraffiCalm sign that consisted of a 48-inch by 36-inch sign with an 18-inch full matrix amber feedback display. The third sign evaluated in this study was from All Traffic Solution (ATS). The ATS sign with an 18- inch full matrix amber feedback display with a smaller black-on-white “YOUR SPEED” panel on top of the display panel. This sign did not include an additional border and was consequently a much smaller overall size (30-inch by 20-inch) compared to the TrafficCalm signs. All three signs tested in this study are shown in Figure 4. Figure 3 Dynamic radar speed feedback sign displaying example feedback messages 39 a) TraffiCalm with b) TraffiCalm with c) ATS with 15-inch Display 18-inch Display 18-inch Display Figure 4 Dynamic radar speed feedback signs tested during the field evaluations Apart from the sign size and border, the radar systems utilized for vehicular detection varied between the two sign manufacturers, as follows: • The two TraffiCalm signs utilized a forward-facing radar with a 30-degree cone embedded into the sign face. • The ATS sign utilized a radar unit that was mounted independently from the sign and employed a more focused 10-degree radar cone, which was intended to better isolate the ramp vehicles and minimize the activation of the sign by the mainline vehicles. The radar units on all signs were calibrated in the field to achieve vehicle detection ranges that were as similar as possible. However, it was not possible to achieve identical performance due to differences in the radar designs. Specifically, after calibration, the TraffiCalm signs (both 15-inch and 18-inch) had a typical vehicle detection range (i.e., the location where vehicles would typically be detected by the radar) of approximately 400 feet for passenger cars, which extended up to approximately 600 feet for large trucks. However, the ATS sign, with its more concentrated radar band had a typical vehicle detection range of approximately 600 ft for passenger cars, which 40 extended up to approximately 1,000 ft for trucks. Each of the signs was able to display a variety of speed feedback messages and could be programmed to display different messages based on the speed of the approaching vehicle. During the evaluations, the signs were powered using a 140 amp-hour portable battery system that powered the sign for two weeks on a single charge. 3.2 DSFS Programming The DSFS utilized in this study afforded substantial flexibility to modify the message display and radar performance settings. Sign programming was performed using manufacturer- specific software and/or smartphone applications in the field prior to each evaluation. Details on the sign programming are provided in the following subsections. 3.2.1 TraffiCalm sign The TraffiCalm sign was programmed using either a smartphone app or a computer software program named SafetyCalm. The SafetyCalm software allows connection using Bluetooth, serial port, or dialup connections for the computer version and via Bluetooth for the Android app. The software provides direct access to the TraffiCalm driver feedback display’s configurable features, day plans, schedules, and data captured by the sign. A layout of the sign programming home screen is shown in Figure 5. Detailed information on the software can be found in the user manual (66). The sign is programmed by clicking on “Edit Display” or from the “Edit” menu. The TraffiCalm signs allowed for programming different message displays based on four speed thresholds. For this study, the following message display settings were applied: o “Min speed”, which is the minimum speed for activation of the display panel, was set at 15 mph in order to prevent rain and small objects (leaves, debris, etc.) from activating the 41 sign. For this study, the measured speed was displayed for vehicles exceeding the minimum speed. o “Speed limit” was set to match the ramp advisory speed at the site. For this study, no additional change was made to the message display, and the measured speed was displayed for vehicles exceeding the speed limit. o “Excess speed” was set to match the ramp advisory speed+10 mph at the site. For this study, the message was modified to include a “SLOW DOWN” message alternating with the measured speed at 1 hz cycles. To achieve this programming, the feedback message was set to flash mode, “Strobe” was ON, and “Strobe Modes” was set to SLOW DOWN. Please note that while a beacon/strobe was present as a part of the sign, it was not enabled during any portion of the testing. o “Max speed” sets the maximum speed, beyond which the sign displays a blank screen or static “SLOW DOWN” or “TOO FAST” feedback message. For this study, the feedback panel was programmed to go blank for vehicles exceeding 85 mph. This strategy is used to prevent motorists from accelerating in an attempt to achieve high speed feedback values displayed on the sign. o The “Squelch” setting can be modified to achieve different vehicle detection ranges. The squelch defines the sensitivity level of the radar embedded in the sign and the values range from 1 to 999, where 1 is the highest sensitivity and 999 is the lowest sensitivity, which essentially provides no vehicular detection. The manufacturer suggests not to use a squelch value of less than 50 as this could result in excessive false signals. A squelch of 60 is recommended in the user manual to achieve optimal results, which can be extended up to 100 if needed. For this study, a squelch value of 60 was utilized. 42 o The color was set to amber for all speed levels. Furthermore, a brightness level of 3 was found to provide proper message visibility during both night and daytime. Figure 5 TraffiCalm sign programming using SafetyCalm software 3.2.2 All Traffic Solution sign The DSFS from All Traffic Solution is programmed using TraffiCloud Sign Manager software provided by the sign vendor (67), which connects to the sign via laptop using USB, Bluetooth, serial port, or remotely from TraffiCloud’s website. A typical layout of the sign homepage is shown in Figure 6. Once the sign is connected, it can be programmed as desired. For the purpose of this study, changes were made only to the sections that include: 43 o “Sign Mode”, which is a dropdown menu in the top left and includes eight sign programming modes: Display Off, Speed Limit Sign, Speed Display, Single Message, All Messages, Dependent Messages, Daily Schedules, and Weekly Schedules. For the purpose of this study, the sign was in “dependent messages” mode. This mode allows feedback messages to be programmed based on the approaching speed measured by the radar device. o “Speed”, which are the data entry values below the “Sign Mode” menu. The speed limit was set to match the ramp advisory speed at the site, and the speed display was allowed from 15 mph to 85 mph, which matched the TraffiCalm sign programming. Radar sensitivity was kept at low (level 2), as this resulted in a detection range of approximately 600 ft for passenger cars and even higher for heavy vehicles. o “Messages on sign”, which are the data entry values to the right of the “Speed” area. The messaging strategy can be programmed with a set number of conditions based on different speed bins from the “Messages on sign” section. This sign afforded greater flexibility in feedback messaging than the TraffiCalm sign, allowing for selection from a robust library of pre-set messages, in addition to custom text messages. This sign also affords the ability to define any number of messaging screens for each speed threshold that can be programmed to display in an alternating manner after a vehicle is detected. For this study, the sign was programmed to display a blank message for no vehicles (one screen), speed number only for vehicle speeds less than ramp advisory speed+10 mph (one screen), and speed number alternating with SLOW DOWN message for vehicle speed 10 mph or above the ramp advisory speeds (two screens). For the two-screen messages, the duration of each screen was set to 0.5 seconds to match the TraffiCalm programming. 44 Figure 6 ATS sign programming using TraffiCloud sign manager software 3.3 Site Selection Several potential candidate interchange ramps for the field evaluation were identified by the research team based on the following characteristics: • Frequent lane departure ramp crashes (reported and/or on-site evidence of such), • Posted ramp advisory speeds at or below 35 mph, • Presence of existing MUTCD complaint curve warning devices, • High ramp AADT, including a considerable proportion of trucks, • DSFS sign installation capability, and • Suitability for data collection. 45 A total of six interchange ramps were ultimately selected, including a balance between system and service interchanges. The general site characteristics and test conditions are displayed in Table 2. Five of the six sites were exit ramps, while the remaining ramp (Site 1) was a business route connection ramp onto a freeway. Although the primary intent of this study was to investigate the effects of DSFS at freeway exit ramps, this ramp was selected for the initial evaluation as it was completely isolated, thereby eliminating sign activations from non-ramp traffic. Table 2 Interchange ramps selected for DSFS field evaluation Site Interchange Speed Limit, Ramp Interchange No. Ramp Location Advisory Speed (mph) Type Test Conditions 1 EB I-69 to 55, 30 System Message type, sign WB I-69 location with respect to curve 2 WB I-96 to 70, 30 System Changes in sign SB I-69 effectiveness over time 3 NB US-127 to 75, 30 Service Time of day, light Round Lake Rd condition, sign lateral position, sign activation range, sign size, sign border type 4 EB I-96 to 70, 20 Service Sign location with 36th St respect to curve 5 NB US-127 70, 25 Service Sign location with to Dunckel respect to curve, sign Rd lateral position 6 EB I-96 to 70, 25 System Sign location with NB US-127 respect to curve, mainline speeds Note: MUTCD-compliant signage, including W13-6, E5-1a, and W1-8R, were present at all sites during all phases. Note that the mainline ran adjacent to the exit ramp at sites 2-6. While activation of the DSFS by the mainline vehicles occurred at these locations, in all cases, the feedback message reflected the ramp vehicle’s speed immediately upon entering the radar detection range. This was 46 partially due to the sign being positioned closer to ramp traffic than mainline traffic, but also the design of the radar detection algorithms, which prioritize vehicles near the center of the radar. 3.4 Data Collection Methods Data were collected from each study location broadly in two phases: 1) under the existing site conditions without the DSFS present and 2) after the installation of the DSFS or after modifying the DSFS setup or operation. After collection of data under the existing site condition, the DSFS was first installed by MDOT crews on dual aluminum sign posts, with a 7-ft bottom mounting height from the pavement surface, and at an offset distance of approximately 12-ft from the near edge of the ramp travel way. The existing signage at each site was not modified in any way. Shortly after DSFS installation, the sign was programmed by the research team, and the accuracy of the displayed speed message was validated with a handheld LIDAR gun for a sample of vehicles. The sign remained operational for seven days prior to initiating data collection in order to allow for dissipation of any driver novelty effects associated with the new traffic control device. For each subsequent change to the DSFS condition, a period of two days was allowed to pass prior to data collection. The study utilized three different techniques for the collection of vehicular speeds: 1) a series of high-definition video cameras, 2) handheld LIDAR, and 3) speed-trailer. The selection of the data collection technique was based on the site characteristics and types of data desired. The same data collection procedures were utilized across all data collection periods for a given evaluation. All data were collected under dry daylight conditions on weekdays between the hours of 10:00 AM and 4:00 PM, with the exception of the speed trailer, which collected data continuously. The following sections provide additional details of the data collection methods. 47 3.4.1 Data collection using cameras This method, which was utilized at sites 1, 2, 3, and 4, involved temporary installation of a series of elevated high-definition video cameras mounted on aluminum poles that were temporarily attached to a roadside signpost behind the respective signs were used to collect the data. The cameras were elevated to a height of 15 feet and aimed towards a pre-determined location on the roadway. Each camera recorded approximately 3 hours of video per data collection period, after which they were removed from the site. Setup and removal of each camera/pole took approximately 5 minutes and the cameras were installed in the same locations to provide approximately identical views during each data collection period. A total of three cameras were utilized at each site, which were positioned to provide full visible coverage of vehicles approaching and entering the curve. The three setup locations, with examples displayed in Figure 7, included: • Around 1,000 feet upstream of the PC and aimed downstream to provide coverage of the area upstream of the DSFS detection zone; • Around 450 feet upstream of the PC and aimed downstream to provide coverage of the area where the DSFS typically became illuminated for approaching vehicles, in addition to the curve approach and entry; and • Around 150 feet after the PC and aimed upstream to provide coverage of the entry to the curve and approximately 100 feet beyond the PC. An example camera setup (middle location) is displayed in Figure 7 along with a screenshot of the general field of view provided by video from each of the three cameras. 48 a. Example camera setup (middle location) b. View from upstream camera c. View from middle camera d. View from curve camera Figure 7 Example camera setup and field of view provided by each camera After completion of the camera-based field data collection, the videos were manually reviewed by a team of trained technicians to assess various characteristics of driver behavior. The view afforded by the elevated camera setups allows direct observation of numerous vehicular and operational attributes, including vehicle type, headway, speed, deceleration rate, brake light indication, lateral positioning within the lane, and occurrence of edgeline or shoulder encroachments. Further, it was also possible to assess various additional aspects related to the DSFS and site condition from the videos, including location of the vehicle at DSFS message onset, DSFS message displayed to each approaching vehicle, traffic volumes, weather, and interference from vehicles parked on the shoulder. 49 The videos were reviewed and each vehicle was sequentially tracked through the videos across the three camera setups. Quicktime software was utilized, which allowed for frame-by- frame review to determine the relevant vehicular location and time information. The videos were recorded at a rate of 60 frames per second, allowing time to be recorded to the nearest 0.0167 seconds based on the frame number displayed in the video player. Paint marks placed on the shoulder at 50-ft intervals (visible in Figure 7) were used as field reference markers for determining the relative location of a vehicle with respect to the curve PC at any point in time. The following information was obtained from the videos for each vehicle traveling through the site: • Time to traverse the following speed measurement zones: o 1000-ft upstream of the PC (prior to the DSFS detection area), o 400-ft upstream of the PC (after entering the DSFS detection area), o At curve PC, and o 100-ft after curve PC. • Deceleration rate over the initial 100-ft beyond the PC, • Time headway from the prior vehicle, • Location (with respect to the PC) at the initial point of brake-light illumination, • Whether a message was displayed on the DSFS, and • Vehicle type: o Passenger vehicle (car, SUV, pickup, van, minivan, motorcycle) without a trailer, o Passenger vehicle with a trailer, o Tractor-trailer truck, and o Single unit truck /bus/RV. 50 Lateral lane positioning was also collected from the videos for vehicles entering the ramp curve during the initial evaluation at Site 1. The lateral lane positioning of vehicles upon curve entry was found to be consistent across all test conditions. Specifically, vehicles generally tracked to the right or the center while traversing the curve, and no vehicles committed an encroachment onto the outer (left) edge line or shoulder. Thus, no further analysis of lateral lane positioning was performed during subsequent evaluations. 3.4.2 Data collection using LIDAR This method, which was utilized at sites 3, 5, and 6, involved the use of handheld LIDAR guns to continuously track vehicle speeds along the entire exit ramp lane, beginning from the start of the upstream taper for the speed-change lane, and continuing to the curve entry point on the ramp. To track vehicles continuously over this distance, which typically exceeded 1,000 ft, a sequence of two handheld LIDAR guns operated by technicians from within separate vehicles parked just beyond the shoulder was used. The LIDAR guns utilized in this study were ProLaser III manufactured by Kustom Signals Inc. These devices are able to measure vehicular speed and distance three times per second with an accuracy of ±1 mph at a range of 6,000 ft. For purposes of this study, each LIDAR gun was typically only utilized over a range of 1,000 feet due to sight limitations caused by geometry or encroachment of other vehicles. Typically, the upstream and downstream LIDAR data collection vehicles were positioned on the roadside at strategic locations that were away from any critical speed measurement points (e.g., start of taper, feedback sign, start of the curve, etc.) to minimize the influence of the data collection vehicle on drivers. A sample data collection setup is shown in Figure 8, where the upstream and downstream vehicles were positioned 1,350 ft and 500 ft, respectively, upstream of the PC. The GPS coordinates were recorded for each data collection vehicle, in addition to the 51 distance to a common reference point, such as a permanent traffic sign. Data were collected from the same location for each test condition. Figure 8 A typical two-person LIDAR data collection setup The upstream data collector would begin to track each subject vehicle near the start of the taper for the auxiliary lane, and continue tracking at least 100 ft beyond the downstream LIDAR technician. At this point, the tracking responsibilities were then transferred to the downstream technician, who would track each subject vehicle over the remaining distance to slightly beyond the PC. The data collectors communicated via cellular communications to ensure a seamless “hand‐off” of the LIDAR speed tracking as each subject vehicle proceeded along the ramp. In doing so, the upstream technician would convey the type and color of each subject vehicle to the downstream LIDAR collector. In order to isolate driver response to the speed feedback sign, only freely flowing vehicles (e.g., minimum 5 second headway) were included. Each LIDAR gun was connected to a laptop using a data transfer cable, which allowed for all measurements to be recorded in real-time using proprietary software. The computer LIDAR recordings included timestamp, distance, and speed for each measurement. After completion of the LIDAR tracking for each subject vehicle, both data collectors entered remarks on the type and color of the vehicle, in addition to any other comments. This information was later used to 52 combine the two data sets into a continuous speed profile for each subject vehicle approaching and entering the ramp curve. Collecting data using this LIDAR tracking method provides a significant advantage over cameras or pneumatic tubes, as it provides continuous speed measurements over the entire segment of interest, as opposed to spot speeds at fixed points. After completion of the LIDAR tracking data collection from the field, both files from the upstream and downstream LIDAR collector were joined using vehicle sequence, type, and color. As the relative distances between the LIDAR collectors and the PC were known, all distances were converted to be relative to the PC. An example representation of the output of this process is shown in Figure 9a. Because LIDAR speeds can't be measured at the same locations on the roadway for every vehicle, it was necessary to convert this data to a series of spot speeds using an interpolation technique. The combined raw data were linearly interpolated at 1-ft increments using the adjacent speeds. Interpolated speeds were then selected at every 50 ft interval starting from the PC, as shown in Figure 9b. Compiling the LIDAR data in this manner provides a robust array of spot speeds at numerous points along the ramp. The data were compiled separately for passenger cars and heavy vehicles (e.g. trucks and buses). Vehicles that were found to be missing data for substantial distances during the speed tracking process were excluded. a) Raw LIDAR data (n=203 vehicles) b) LIDAR data interpolated at 50-ft increments Figure 9 Raw and interpolated vehicle speed data from LIDAR 53 3.4.3 Data collection using speed trailer Although collection of driver behavior data using video cameras and two-person LIDAR hand-off method afford vehicle tracking capabilities, their use is limited to daylight conditions and data extraction is very labor-intensive, which further limits sample sizes. Thus, in order to obtain a comprehensive sample of curve entry speeds across various times of day and light conditions, a speed data collection trailer was utilized at one site (Site 3: US127/Round Lake Rd) just beyond the right shoulder near the point of curvature (Figure 10). The speed trailer utilized an elevated side-firing Wavetronix SmartSensor HD radar unit, which was aimed and calibrated to continuously measure speed and length data for vehicles at the ramp curve entry point. Note that LIDAR speed data were also collected using the aforementioned methods during the DSFS evaluations performed at this site. The speed trailer remained in a fixed position and continuously recorded data for the entire field test period. To limit data file size, the radar was programmed to bin data into 30-second intervals. To isolate free-flowing vehicles, the data were screened to include only those intervals that included a single vehicle. Further screening was conducted to remove potential mainline observations and other anomalies. Vehicle length information was utilized to distinguish between passenger vehicles and heavy vehicles. After an extensive reliability assessment of the vehicle length data, all vehicles with a length of 32 ft or less were considered as passenger vehicles, while the remainder were considered heavy vehicles. 54 Figure 10 Speed trailer at NB US-127 to Round Lake Rd 3.5 Measures of Effectiveness and Analytical Methods Several measures of effectiveness related to speed, deceleration rate, and initial braking location were analyzed to determine the effects of the various DSFS conditions and other factors. To determine any obvious trends in the data, sources for potential bias, and data distributions, a preliminary comparison of the descriptive statistics (i.e., mean, standard deviation, percentiles, etc.) and graphical representations (i.e., frequency distribution, box plot, scatterplot) for the vehicular data were performed across the data collection periods. From there, three primary behavioral measures of effectiveness (MOE) associated with the DSFS were analyzed using appropriate statistical procedures, which are described in greater detail in the paragraphs that follow. The MOE for these analyses included: • Speed at PC (i.e., curve entry speed), 55 • Initial braking occurring between 200 and 600 ft upstream of the PC; this range was selected as it corresponded with the detection range of DSFS in the upstream position and afforded a comfortable braking distance prior to curve entry, and • Curve entry speed within 15 mph of the curve advisory speed; this MOE was used to assess the effect of the DSFS at limiting excessive curve entry speeds. The same general set of predictor variables were entered into all models and included: upstream approach speed (to control for general driver speed selection), DSFS test condition, and vehicle type (when separate models were not developed for different vehicle types). All analyses were performed using statistical software RStudio. Speeds were analyzed using multiple linear regression, while logistic regression was utilized to analyze the binary response variables, which included the probability of initial braking in advance of the curve and the probability of vehicles entering the curve at speeds exceeding 15 mph above the curve advisory speed. The general form of the multiple linear regression is shown in Equation 1: 𝑌𝑖 = 𝛽0 + 𝛽1 𝑋𝑖1 + 𝛽2 𝑋𝑖2 + ⋯ + 𝛽𝑘 𝑋𝑖𝑘 + 𝜀𝑖 (1) where Yi is the measured speed at the PC for vehicle i, 𝑋𝑖1 to 𝑋𝑖𝑘 are independent variables affecting the dependent variables (including DSFS test condition), β0 is an intercept, β1 to βk are estimated regression coefficients for each independent variable, and 𝜀𝑖 is a normally distributed error term with variance 𝜎 2 . When analyses were conducted using the data from multiple sites or from a single site on multiple dates, the linear regression included a random effect (intercept) term in the model, with the form shown in Equation 2: 𝑌𝑖 = 𝛽0 + 𝛽1 𝑋𝑖1 + 𝛽2 𝑋𝑖2 + ⋯ + 𝛽𝑘 𝑋𝑖𝑘 + 𝜀𝑖 + 𝛿𝑖 (2) 56 where, 𝛿𝑖 term is a random intercept term. This accounts for unobserved factors affecting driver behavior between the data collection periods. A site-specific random intercept term was utilized when video speed data were collected from multiple sites to account for the unobserved differences between sites. As the speed trailer was only used at a single site for multiple days, a date-specific random intercept term was utilized for these models to account for unobserved day- to-day differences. Binary logistic regression was utilized to analyze the binary response variables, which included the probability of braking 200-600 ft in advance of the curve and the probability of a vehicle entering the curve within 15 mph of the advisory speed. Binary logistic regression is a technique used to predict the probability of a dichotomous outcome based on values of a set of predictor variables (continuous or categorical) and is similar to linear regression except that the response variable is categorical rather than numeric. The binary logistic regression model has the form as shown in Equation 3: 𝑃 𝑌𝑖 = 𝑙𝑜𝑔𝑖𝑡(𝑃𝑖 ) = 𝑙𝑛 (1− 𝑖𝑃 ) = 𝛽0 + 𝛽1 𝑋𝑖1 + 𝛽2 𝑋𝑖2 + ⋯ + 𝛽𝑘 𝑋𝑖𝑘 (3) 𝑖 where the response variable, Yi, is the logistic transformation of the probability of the vehicle i braking 200-600 ft in advance of the curve or entering the curve within 15 mph of the advisory speed. This probability is denoted as Pi. As in the linear regression model, 𝑋𝑖1 to 𝑋𝑖𝑘 are independent variables affecting driver behavior (including DSFS test condition), β0 is an intercept, β1 to βk are estimated regression coefficients for each independent variable. 57 4. EVALUATION OF MESSAGE TYPE AND LONGITUDINAL SIGN POSITION 4.1 Study Design and Site Characteristics This initial field evaluation included an assessment of the general effectiveness of DSFS on driver behavior, in addition to an assessment of various messaging strategies and longitudinal sign positioning relative to the curve. To control for external site biases, a single DSFS (15-inch TraffiCalm Sign) was utilized at a single freeway interchange ramp (Site 1: EB I-69 to WB I-69) during this initial field evaluation. This site possessed an upstream speed limit of 55 mph and a curve advisory speed of 30 mph. Although the primary intent of this research was to investigate high-speed to low-speed transitions (e.g., freeway exit), this particular entrance ramp was selected for the initial evaluation as it was completely isolated, thereby eliminating sign activations from non-ramp traffic. Existing 60-inch by 48-inch “EXIT 30 MPH” (W13-6a) signs with overhead beacons were present on both sides, approximately 450-ft upstream of the curve, and chevrons were present within the curve. A plan view of the study site, including the three DSFS locations is provided in Figure 11. Three feedback messaging strategies were evaluated at three sign positions in advance of the curve, as follows: • DSFS feedback messages: o Message Strategy 1: Display the measured speed. o Message Strategy 2: ▪ For speeds at or below 40 mph (i.e., advisory speed + 10 mph), display the measured speed; 58 ▪ For speeds above 40 mph, display the measured speed alternating with a “SLOW DOWN” message on 1 hz cycles. This messaging strategy followed the current MDOT draft special provision for DSFS. o Message Strategy 3: Same feedback message as Strategy 2, but with a 30-inch square advisory speed panel (black text on fluorescent yellow reflective sheeting) mounted immediately below the speed feedback sign. • DSFS longitudinal position relative to the curve (see Figure 12): o Location 1: At the point of curvature (PC) o Location 2: 255-ft upstream of the PC, to correspond with the minimum braking distance necessary to accommodate deceleration at 11.2 ft/s2 from the 85th percentile speed (measured as 53 mph) to the 30 mph advisory speed upon entry to the curve, plus a 1 second perception-reaction time. o Location 3: 400-ft upstream of the PC and hung under the existing right most flashing “EXIT 30 MPH” warning sign. Figure 11 EB Business I-69 Ramp to WB I-69 (Site 1) 59 a. DSFS at Point of Curvature b. DSFS 255 ft Upstream of Curve (with 30 mph Advisory Speed Panel) c. DSFS 450 ft Upstream of Curve (Under Existing Curve Warning Sign) Figure 12 Example DSFS messaging conditions and installation locations (Site 1) In order to test the effects of the three messaging strategies at each of the three sign locations compared to the existing condition without the DSFS, a total of nine data collection periods were required. Note that because the DSFS was positioned under the existing ‘‘EXIT 30 MPH’’ sign at the 450 ft upstream location, it was not possible to test the DSFS without the 60 advisory speed at this position. Thus, a total of eight DSFS messaging and sign positioning combinations were tested and compared with the existing site condition without the DSFS. Existing conditions without the DSFS data were collected both before the installation of the DSFS and after completing data collection with all other test conditions. The order of the sign test conditions and corresponding data collection periods were as follows: 1. Existing site condition prior to DSFS installation, 2. DSFS at PC with speed number feedback only, 3. DSFS at PC with speed number alternating with SLOW DOWN message (Figure 12a), 4. DSFS at PC with speed number alternating with SLOW DOWN message and advisory speed panel, 5. DSFS at 255 feet upstream of PC with speed number alternating with SLOW DOWN message and advisory speed panel (Figure 12b), 6. DSFS 255 feet upstream of PC with speed number alternating with SLOW DOWN message, 7. DSFS 255 feet upstream of PC with speed number feedback only, 8. DSFS 400 feet upstream of PC with speed number feedback and advisory speed panel, 9. DSFS 400 feet upstream of PC with speed number alternating with SLOW DOWN message and advisory speed panel (Figure 12c), and 10. Existing site condition after completion of all the data collection with the DSFS. 4.2 Data Summary Driver behavior data were collected using a series of three elevated high-definition video cameras mounted on aluminum poles that were temporarily attached to a roadside signpost. The videos were reviewed using QuickTime software and each vehicle was sequentially tracked 61 through the videos across the three camera setups. The vehicular observation data were then tabulated, organized, and coded into a single data file for detailed analysis. The initial data set included complete records for 3,442 vehicles collected across the ten data collection periods. To eliminate the effects of platooning, vehicles with headways below 3.0 seconds were excluded from further analysis. Furthermore, the six cases where no message was displayed on the DSFS were also removed from the dataset. Table 3 presents the descriptive statistics for each variable included in the final dataset, which included 2,212 vehicles. Table 3 Descriptive statistics (Site 1) Variable Unit Min Max Mean Std. Dev Speed 1000 ft Upstream of the PC mph 30.080 77.187 47.957 5.388 Speed at PC mph 26.914 63.920 42.323 5.159 Speed <45 mph at PC (1 = yes) 0 1 0.742 0.438 Braking first initiated 200-600 ft prior to PC (1 = yes) 0 1 0.238 0.426 Existing site condition (No DSFS) 0 1 0.168 0.374 DSFS at PC with: Speed number only 0 1 0.121 0.326 Speed number alternating with slow down 0 1 0.113 0.317 message Speed number alternating with slow down 0 1 0.103 0.304 message plus advisory speed panel DSFS at 255 ft upstream with: Speed number only 0 1 0.094 0.291 Speed number alternating with slow down 0 1 0.082 0.275 message Speed number alternating with slow down 0 1 0.086 0.280 message plus advisory speed panel DSFS at 450 ft upstream with: Speed number and advisory speed panel 0 1 0.120 0.325 Speed number alternating with slow down 0 1 0.113 0.316 message plus advisory speed panel Passenger Vehicle Without Trailer 0 1 0.909 0.287 Passenger Vehicle With Trailer 0 1 0.029 0.168 Tractor Trailer Truck 0 1 0.025 0.156 Single Unit Truck/Bus/RV 0 1 0.037 0.189 N = 2,212 vehicles 62 4.3 Results and Discussion Preliminary models suggested large differences in the speeds of passenger vehicles and heavy vehicles. As a result, separate models were generated for passenger vehicles, single-unit trucks/buses, and tractor-trailer trucks. For the passenger vehicle models, DSFS test conditions (e.g., sign message and position) were included as separate binary variables and evaluated against the existing site condition without the DSFS. However, for the heavy vehicle models, because of small sample sizes, the DSFS was evaluated as a single binary variable (No DSFS and with DSFS) combined across all test conditions. Three measures of effectiveness related to speed and initial braking location were analyzed to determine the effects of the various DSFS conditions and other factors. The dependent variables for these analyses included: • Speed at the PC (curve entry), • Probability of initial braking occurring between 200 and 600 ft upstream of the PC, and • Probability of a vehicle entering the curve below 45 mph. The linear regression results for curve entry speeds (e.g., at the PC) are displayed for each vehicle type in Table 4. The logistic regression results for the probability of initial braking in advance of the curve and the probability of vehicles entering the curve at speeds below 45 mph are displayed in Tables 5 and 6. Further discussion is provided in the following sections. 4.3.1 Curve entry speeds A review of the parameter estimates in Table 4 indicates that curve entry speeds were generally lower across all DSFS test conditions and vehicle types compared to the site without the DSFS. Also, as expected, the speed of vehicles measured 1,000 ft prior to the curve was strongly 63 correlated with the speed at the point of curvature for each of the speed models. This finding is aligned with prior research and suggests that faster drivers tended to maintain such behaviors regardless of the DSFS characteristics. Further discussion of the primary variables of interest, including sign position, messaging strategy, and vehicle type is provided in the following sections. Table 4 Multiple linear regression results for curve entry speed (Site 1) Passenger Vehicles (N=2,075) Parameter Estimate Std. Error t-value p-value Intercept 13.126 0.787 16.680 <0.001 Speed 1000 ft Upstream of the PC 0.626 0.016 40.144 <0.001 Passenger Vehicle without Trailer Base Condition Passenger Vehicle with Trailer -1.615 0.471 -3.432 <0.001 Existing Site Condition (No DSFS) Base Condition DSFS at PC with: Speed number only -0.681 0.302 -2.252 0.024 Speed number alternating with slow down message -1.076 0.310 -3.475 0.001 Speed number alternating with slow down message -0.879 0.323 -2.723 0.007 plus advisory speed panel DSFS at 255 ft upstream of PC with: Speed number only -0.544 0.329 -1.654 0.098 Speed number alternating with slow down message -1.472 0.343 -4.290 <0.001 Speed number alternating with slow down message -1.220 0.336 -3.628 <0.001 plus advisory speed panel DSFS at 450 ft upstream of PC with: Speed number and advisory speed panel -0.011 0.303 -0.038 0.970 Speed number alternating with slow down message -0.501 0.312 -1.604 0.109 plus advisory speed panel Single Unit Trucks and Buses (N=82) Parameter Estimate Std. Error t-value p-value Intercept 11.168 3.418 3.267 0.002 Speed 1000 ft Upstream of the PC 0.638 0.076 8.371 <0.001 Existing Site Condition (No DSFS) Base Condition With DSFS in Operation -1.541 0.964 -1.599 0.114 Tractor Trailer Trucks (N=55) Parameter Estimate Std. Error t-value p-value Intercept 26.306 4.147 6.343 <0.001 Speed 1000 ft Upstream of the PC 0.314 0.093 3.365 0.001 Existing Site Condition (No DSFS) Base Condition With DSFS in Operation -5.070 1.212 -4.183 <0.001 Response Variable = Speed at Point of Curvature (PC) R2 = 0.459 (Passenger Vehicles), R2 = 0.458 (Single Unit Trucks and Buses), R2 = 0.365 (Tractor Trailer Trucks) 64 4.3.1.1 Effect of DSFS position In terms of the sign position, the DSFS had the greatest effect on curve entry speeds for passenger vehicles when located 255 ft upstream of the point of curvature. As noted previously, this sign position corresponded to the minimum AASHTO braking distance to achieve the 30 mph curve advisory speed, plus a 1 second perception-reaction time. With the DSFS installed 255 ft upstream of the curve, depending on the messaging strategy, the curve entry speeds were 0.54 to 1.47 mph lower than when the DSFS was not present. The sign was only slightly less effective when positioned directly at the point of curvature, as the curve entry speeds were 0.68 to 1.08 mph lower than when the DSFS was not present. The DSFS did not have a significant effect on curve entry speeds when installed at the position that was furthest upstream of the curve (450 ft). Figure 13 provides a graphical representation of the DSFS marginal effects on curve entry speeds (passenger vehicles only) across the sign positions and messaging strategies. Figure 13 Mean curve entry speeds by DSFS position and messaging strategy (PSGR only) 65 4.3.1.2 Effect of DSFS messaging strategy Turning to the effects of the messaging strategy, the DSFS had the greatest effect on curve entry speeds when displaying the speed number combined with an alternating SLOW DOWN message for vehicles that were 10 mph over the advisory speed. The inclusion of an advisory speed panel with the DSFS did not substantively change this effect. However, the removal of the SLOW DOWN message, such that only the speed digits were displayed as feedback, significantly degraded this effect to the point where the curve entry speeds were only marginally lower than without the DSFS installed. 4.3.1.3 Effect of vehicle type The effectiveness of the DSFS at reducing curve entry speeds varied greatly by vehicle type. While passenger vehicles showed curve entry speeds that, depending on sign position and message strategy, were generally on the order of 0.5 to 1.5 mph lower with the DSFS installed at the site, the sign had an even greater effect on heavy vehicles, especially tractor-trailer trucks. For single-unit trucks and buses, curve entry speeds were, on average, 1.54 mph lower with the DSFS installed than without the DSFS. However, for tractor-trailer trucks, curve entry speeds were 5.07 mph lower when the DSFS was operating at the site. Due to the relatively small sample sizes of heavy vehicles, it was not possible to test the effects of DSFS messaging strategies or position with respect to the curve. Finally, the impact of the DSFS on curve entry speeds for passenger vehicles was not affected by whether the vehicle was towing a trailer. Figure 14 provides a graphical representation of the DSFS marginal effects on curve entry speeds across the three primary vehicle type categories. 66 Figure 14 Mean curve entry speeds by DSFS presence and vehicle type 4.3.2 Extreme curve entry speeds It was also important to assess the impacts of the DSFS on extreme curve entry speeds, which for purposes of this study, were defined as speeds that were in excess of 45 mph (e.g., 15 mph above the posted advisory speed of 30 mph). To simplify the interpretation of the results with respect to the DSFS condition, the logistic regression models were formulated to estimate the probability of a vehicle entering the curve below 45 mph. 67 Table 5 Binary logistic regression results for curve entry speed < 45 mph (Site 1) Passenger Vehicles (N=2,075) Parameter Estimate Std. Error t-value p-value Exp(B) Intercept 15.047 0.778 19.339 <0.001 3426329 Speed 1000 ft Upstream of the PC -0.290 0.015 -19.020 <0.001 0.748 Passenger Vehicle without Trailer Base Condition Passenger Vehicle with Trailer 0.719 0.511 1.405 0.160 2.052 Existing Site Condition (No DSFS) Base Condition DSFS at PC with: Speed number only 0.222 0.207 1.075 0.282 1.249 Speed number alternating w/slow down message 0.537 0.219 2.453 0.014 1.711 Speed number alternating w/slow down message 0.522 0.228 2.290 0.022 1.686 plus advisory speed panel DSFS at 255 ft upstream of PC with: Speed number only 0.133 0.237 0.562 0.574 1.142 Speed number alternating w/slow down message 0.933 0.271 3.450 0.001 2.543 Speed number alternating w/slow down message 0.485 0.240 2.020 0.043 1.624 plus advisory speed panel DSFS at 450 ft upstream of PC with: Speed number and advisory speed panel 0.266 0.210 1.268 0.205 1.305 Speed number alternating w/slow down message 0.335 0.225 1.490 0.136 1.398 plus advisory speed panel Single Unit Trucks and Buses (N=82) Parameter Estimate Std. Error t-value p-value Exp(B) Intercept 11.646 4.454 2.615 0.009 114233 Speed 1000 ft Upstream of the PC -0.206 0.091 -2.269 0.023 0.814 Existing Site Condition (No DSFS) Base Condition With DSFS in Operation 0.749 0.961 0.779 0.436 2.114 Response Variable = Speed <45 mph at Point of Curvature (PC) Note: all tractor-trailer trucks were below 45 mph at point of curvature and, as a consequence, were not analyzed. The parameter estimates in Table 5 indicate that the likelihood of a vehicle entering the curve below 45 mph was generally greater across all DSFS test conditions and vehicle types compared to the site without the DSFS. Vehicles entering the curve with the DSFS present were 1.14 to 2.54 times more likely to enter the curve at a speed below 45 mph compared to the site without the DSFS. In terms of the sign position, the DSFS had the greatest effect on extreme curve entry speeds when located either at the point of curvature or 255 ft upstream of the point of curvature. Further, the most effective DSFS message included the speed number combined with an alternating SLOW DOWN message for vehicles that were 10 mph over the advisory speed. 68 Considering vehicle type, the DSFS produced a greater reduction in extreme curve entry speeds when the vehicle was a single unit truck or bus. It was not possible to assess the effect on drivers of tractor-trailer trucks, as all possessed curve entry speeds that were below 45 mph. 4.3.3 Initial braking location The parameter estimates in Table 6 indicate that the presence of the DSFS generally led to a greater occurrence of braking between 200 and 600 ft prior to the curve, which was considered the optimal range from a driver comfort standpoint. Further discussion of the effects of sign location and vehicle type is provided in the sections that follow. Note that no consistent trends related to the sign message were observed. 4.3.3.1 Effect of DSFS position Not surprisingly, the DSFS position affected the occurrence of braking within this comfortable range, which was greatest when the DSFS was positioned 255 ft upstream of the curve. Specifically, when the DSFS was present at this location, drivers were 1.83 to 2.72 times more likely to begin braking 200 to 600 ft upstream than when the DSFS was not present. The DSFS had the weakest effect on braking location when positioned at the point of curvature. 4.3.3.2 Effect of vehicle type The effectiveness of the DSFS at influencing the initial braking location varied greatly by vehicle type. In general, the greater the vehicle size, the more effective the DSFS was at prompting braking within the 200 to 600 ft comfort range in advance of the curve. While drivers of passenger vehicles were 1.19 to 2.72 times more likely to initiate braking within the comfort range with the DSFS present than without, the sign had an even greater effect on heavy vehicles, especially tractor-trailer trucks. Specifically, drivers of single-unit trucks and buses were 1.75 times more 69 likely to begin braking 200 to 600 ft upstream of the curve when the DSFS was present, while drivers of tractor-trailer trucks were 3.30 times more likely to begin braking in that range. Due to the relatively small sample sizes of heavy vehicles, it was not possible to test the effects of DSFS position or messaging strategies. Table 6 Binary logistic regression results for braking 200-600 ft prior to the curve (Site 1) Passenger Vehicles (N=2,075) Parameter Estimate Std. Error t-value p-value Exp(B) Intercept -4.408 0.532 -8.280 <0.001 0.012 Speed 1000 ft Upstream of the PC 0.057 0.010 5.552 <0.001 1.059 Passenger Vehicle without Trailer Base Condition Passenger Vehicle with Trailer 0.423 0.296 1.427 0.154 1.526 Existing Site Condition (No DSFS) Base Condition DSFS at PC with: Speed number only 0.294 0.210 1.404 0.160 1.342 Speed number alternating w/slow down message 0.177 0.219 0.807 0.420 1.194 Speed number alternating w/slow down message 0.194 0.227 0.854 0.393 1.214 plus advisory speed panel DSFS at 255 ft upstream of PC with: Speed number only 0.603 0.220 2.742 0.006 1.827 Speed number alternating w/slow down message 1.000 0.218 4.588 <0.001 2.717 Speed number alternating w/slow down message 0.768 0.217 3.531 <0.001 2.155 plus advisory speed panel DSFS at 450 ft upstream of PC with: Speed number and advisory speed panel 0.318 0.210 1.512 0.130 1.375 Speed number alternating w/slow down message 0.750 0.206 3.634 <0.001 2.117 plus advisory speed panel Single Unit Trucks and Buses (N=82) Parameter Estimate Std. Error t-value p-value Exp(B) Intercept 0.695 2.137 0.325 0.745 2.004 Speed 1000 ft Upstream of the PC -0.047 0.049 -0.967 0.333 0.954 Existing Site Condition (No DSFS) Base Condition DSFS in Operation 0.560 0.637 0.878 0.380 1.750 Tractor Trailer Trucks (N=55) Parameter Estimate Std. Error t-value p-value Exp(B) Intercept -4.083 3.184 -1.283 0.200 0.017 Speed 1000 ft Upstream of the PC 0.053 0.069 0.770 0.442 1.055 Existing Site Condition (No DSFS) Base Condition With DSFS in Operation 1.193 1.133 1.054 0.292 3.298 Response Variable = Braking Initiated 200-600 ft Prior to Curve 70 4.4 Conclusion and Direction for Further Evaluation This initial evaluation was utilized to assess the impacts on driver speed selection and braking associated with various DSFS messaging strategies and longitudinal positions of the sign relative to the ramp curve. The evaluation was performed at a single ramp and included three DSFS messaging strategies (speed number only, speed number alternating with SLOW DOWN message, and inclusion of an auxiliary advisory speed panel), as well as at three sign positions (point of curvature, 255 ft upstream of the curve, and 450 ft upstream of the curve). Compared to the existing site condition, the DSFS reduced curve entry speeds and improved brake response across all test conditions, particularly for heavy trucks. Overall, considering the combination of both sign position and feedback messaging strategy, the greatest benefits to driver behavior were attained when the DSFS was positioned within approximately 250 ft of the curve and the feedback message included the speed number alternating with a SLOW DOWN message for vehicles that were greater than 10 mph over the curve advisory speed. This sign location corresponded to the minimum AASHTO braking distance to accommodate comfortable deceleration from the 85th percentile approach speed to the curve advisory speed (plus a 1 second perception-reaction time) upon entry to the curve. The inclusion of an advisory speed panel with the DSFS did not have a substantive impact on driver behavior. The results of this preliminary evaluation were useful for determining effective messaging strategies and longitudinal sign positions for subsequent evaluation at additional locations, particularly at high-speed freeway exit ramps and including system and service interchanges. It is also important to test the DSFS at locations where the freeway mainline runs adjacent to the ramp in order to determine the effect of mainline vehicles on sign activation and message display. 71 5. EVALUATION OF LONGITUDINAL SIGN POSITION, INTERCHANGE TYPE, AND TIME OF DAY 5.1 Study Design and Site Characteristics Following the encouraging results of the initial DSFS evaluation at Site 1, a series of subsequent field evaluations were conducted at three freeway exit ramps possessing significant horizontal curvature. This series of evaluations tested the effectiveness of the feedback sign across various sign positions (at the point of curvature vs. 350 ft upstream), interchange types (system vs. service), times-of-day (peak vs. off-peak), light conditions (daylight vs. darkness), and vehicle types (passenger vehicles vs. trucks). Three exit ramp locations were selected for this study, which included westbound (WB) I-96 to southbound (SB) I-69 (Site 2), northbound (NB) US-127 to Round Lake Road (Site 3), and eastbound (EB) I-96 to 36th Street (Site 4). The same 15-inch TraffiCalm sign that was utilized in the preliminary evaluation was also utilized at each of the three exit ramp locations. To further confirm the effect of sign position, the DSFS was positioned at the PC at Sites 2 and 3 and 350 feet upstream of the PC at Site 4. Furthermore, Site 2 was a system (freeway-to-freeway) interchange, while Sites 3 and 4 were service (freeway-to-crossroad) interchanges, which further allowed for the assessment of differences in behavioral effects between the interchange types. Finally, time-of-day and light condition effects were assessed by using a data collection trailer to measure curve entry speeds before and after installation of the DSFS at Site 3. Detailed information on the three study sites, DSFS positions, and feedback messaging strategy is provided in Table 7. Plan views of the three study sites, including the DSFS positions, are provided in Figures 15, 16, and 17, respectively. One aspect that remained consistent between the three locations was the sign messaging strategy. During all test conditions, the DSFS was programmed to display the speed of vehicles 72 that were approaching below 40 mph, alternating with a SLOW DOWN message at 0.5-sec intervals for vehicles traveling 40 mph and above. This messaging strategy followed the current MDOT draft special provision for DSFS and was found to provide the greatest impact on driver behavior in the initial field study (68). A standard advisory speed panel was also installed with the DSFS at Site 2. However, the use of an advisory speed panel was not found to impact the effectiveness of the DSFS during the initial field study. Table 7 Site characteristics and DSFS test conditions No Location Mainline Speed Interchange Feedback Message DSFS Limit/Ramp Advisory Type Position Speed (mph) 2 WB I-96 to 70/30 System Speed number with PC SB I-69 alternating “Slow Down” message and advisory speed panel 3 NB US-127 75/30 Service Speed number with PC to Round alternating “Slow Lake Rd Down” message 4 EB I-96 to 70/20 Service Speed number with 350 feet 36th St alternating “Slow upstream Down” message of PC Note: PC = point of curvature. Speed data collection trailer was utilized at Site 3. 73 Figure 15 WB I-96 Exit Ramp to SB I-69 (Site 2) Figure 16 NB US-127 Exit Ramp to Round Lake Rd (Site 3) 74 Figure 17 EB I-96 Exit Ramp to 36th St (Site 4) 5.2 Data Summary Data were collected from each study location in two phases: 1) under the existing site conditions without the DSFS present and 2) after the installation of the DSFS. Data were primarily collected using a series of pole-mounted cameras from the three sites. However, the speed trailer was deployed at Site 3 to continuously collect data for a longer period. 5.2.1 Speed data from cameras A series of three pole-mounted high-definition video cameras were installed at various points on the approach to and within the curve. The video data collected from three sites were combined, organized, and coded into a single file for detailed analysis. The data set was screened to include only vehicle observations with a headway of 3 seconds or greater to eliminate the effects of vehicle platooning. Cases where no feedback message was displayed on the DSFS for an 75 approaching vehicle were also removed from the dataset. The final combined data set included a complete record for 1,758 vehicle observations collected from three sites. It should be noted that brake light data could not be discerned from one data collection period due to a slight misalignment of the camera, which reduced the sample size for the braking data to 1,651. Table 8 shows the descriptive statistics of the data collected using cameras from three test sites. Table 8 Descriptive statistics: video camera data (Sites 2-4) Video Camera Data (N=1,758) Variable Min Max Mean Std Dev Speed 1000 feet upstream of the pc, mph 30.303 83.832 58.949 7.701 Speed at PC (i.e., curve entry), mph 17.045 57.273 38.388 5.496 Speed within 15 mph of advisory speed (1-Yes, 0-No) 0 1 0.656 0.475 Braking first initiated 200-600 ft prior to PC (1-Yes, 0-No) 0 1 0.321 0.467 Passenger vehicle without trailer 0 1 0.820 0.385 Passenger vehicle with trailer 0 1 0.018 0.132 Single unit truck/bus/RV 0 1 0.039 0.194 Tractor-trailer truck 0 1 0.123 0.329 Existing site condition (No DSFS) 0 1 0.557 0.497 DSFS at point of curvature - I-96/I-69 0 1 0.136 0.343 DSFS at point of curvature - US-127/Round Lake 0 1 0.133 0.340 DSFS at 350 ft upstream of point of curvature - I-96/36th 0 1 0.173 0.379 5.2.2 Speed data from speed trailer To obtain a comprehensive sample of curve entry speeds across various times of day and light conditions, a speed data collection trailer was positioned at Site 3 (NB US-127 to Round Lake Rd) just beyond the right shoulder near the point of curvature. The speed trailer remained in a fixed position and continuously recorded data for a 36 day period. Data were collected under the existing site condition, without the DSFS, for the initial 20 days. The DSFS was then installed at the point of curvature (in front of the speed trailer), and the speed trailer remained operational for an additional 16 days. Figure 18 displays the DSFS and speed trailer installed at Site 3. 76 Figure 18 DSFS and speed trailer at US-127 exit to Round Lake Rd To isolate free-flowing vehicles, all the speed-trailer data were screened to include only those intervals (30-second) that included a single vehicle. A total of 17,433 vehicles were included in the final dataset for further analysis. The dataset included 16,839 passenger vehicles and 594 heavy vehicles, which were defined in this study as vehicles longer than 32 feet. To account for changes in weather at the site, weather condition data were collected from a nearby National Weather Service station and combined with the speed data. Furthermore, sunrise and sunset data were also included to determine any differential effects of the DSFS on curve entry speeds under varied lighting conditions. The initial analysis found curve entry speeds to vary widely during poor weather conditions, including rain, sleet, ice, and snow. This is not surprising, but it is difficult to characterize the impacts on visibility and pavement surface during such conditions. Thus, the speed trailer data were limited to only those periods that occurred during clear, cloudy, or fair conditions. Using the timestamps of each data bin, the speed trailer data were also coded for further analysis based on times-of-day (peak vs. off-peak) and light conditions (daylight vs. 77 darkness). For this study, the peak period was defined as 7:00 – 9:00 AM and 3:30 to 6:00 PM. Daylight and darkness were coded based on sunrise (between 7:19 and 8:00 AM) and sunset (between 5:05 and 5:26 PM) for each day during the data collection period. The data were then classified for analytical purposes as follows: peak, daytime off-peak, and nighttime. Thus, all daytime off-peak occurred during daylight periods, all nighttime occurred during dark periods, while peak included a mix of daylight, twilight, and dark periods. In using this coding structure, it was possible to assess whether the DSFS had a variable effect on the behavior of peak (e.g. commuter) vs. off-peak drivers and during different lighting conditions. Table 9 presents the descriptive statistics of the combined dataset that 17,433 observations from the speed trailer. Table 9 Descriptive statistics: speed trailer data (Site 3) Speed Trailer Data (N = 17,433) Variable Unit Min Max Mean Std Dev Speed at point of curvature (i.e., curve entry) mph 25.100 54.900 39.039 5.459 Passenger vehicle 0 1 0.966 0.181 Heavy vehicle (longer than 32 ft) 0 1 0.034 0.181 Daytime Off-Peak (9:00 AM – 3:30 PM) 0 1 0.394 0.489 Peak (7:00 AM – 9:00 AM; 3:30 PM – 6:00 PM) 0 1 0.178 0.382 Nighttime (6:00 PM – 7:00 AM) 0 1 0.428 0.495 No DSFS 0 1 0.557 0.497 DSFS at point of curvature 0 1 0.443 0.497 5.3 Results and Discussions Several measures of effectiveness related to curve entry speed and brake response were analyzed to determine the effects of the DSFS as a function of sign position, interchange type, and time of day. Three primary analyses were performed using appropriate statistical procedures. The dependent variables for these analyses included: 78 • Speed at the PC (curve entry), • Probability of initial braking occurring between 200 and 600 ft upstream of the PC, and • Probability of a vehicle entering the curve within 15 mph of the advisory speed. Preliminary models suggested only minor differences in DSFS effectiveness between passenger vehicles and heavy vehicles and further analysis of vehicle-specific DSFS effects was not performed. For all models, the upstream approach speed was included as a covariate to control for general driving behavior. Curve entry speeds were analyzed using linear regression including a random effect (intercept) term. A site-specific random intercept term was utilized in the video speed data models to account for the unobserved differences between sites. As the speed trailer was only used at a single site, a date-specific random intercept term was utilized for these models to account for unobserved day-to-day differences. Binary logistic regression was utilized to analyze the binary response variables, which included the probability of braking 200-600 ft in advance of the curve and the probability of a vehicle entering the curve within 15 mph of the advisory speed. The linear regression results for curve entry speeds (e.g. at the point of curvature) obtained from the video data are displayed in Table 10, while the binary logistic regression results for excessive curve entry speeds and brake response location are displayed in Tables 11 and 12, respectively. 79 Table 10 Random intercept linear regression results for curve entry speed (Sites 2-4) Response Variable: Speed at Point of Curvature (N=1,758) Parameter Estimate Std. Error t-value p-value Intercept 14.983 1.175 12.746 <0.001 Speed 1000-ft Upstream of the PC 0.414 0.014 29.208 <0.001 Passenger Vehicle with No Trailer Base Condition Passenger Vehicle with Trailer -1.339 0.665 -2.015 0.044 Single Unit Truck/ Bus -1.969 0.463 -4.252 <0.001 Tractor Trailer Truck -3.290 0.309 -10.648 <0.001 Existing Site Condition (No DSFS) Base Condition DSFS at PC - I-96/I-69 -2.290 0.299 -7.665 <0.001 DSFS at PC - US-127/Round Lake Rd -1.975 0.331 -5.970 <0.001 DSFS at 350 ft upstream of PC - I-96/36th 0.967 0.283 3.415 <0.001 Table 11 Random intercept logistic regression results for curve entry speed within 15 mph of the advisory speed (Sites 2-4) Response Variable: Speed at Point of Curvature within 15 mph of the Advisory Speed (N=1,758) Parameter Estimate Std. Error t-value p-value Exp(B) Intercept 12.846 1.390 9.245 <0.001 379268 Speed 1000-ft Upstream of the PC -0.201 0.014 -13.936 <0.001 0.818 Passenger Vehicle with No Trailer Base Condition Passenger Vehicle with Trailer 1.959 0.977 2.006 0.045 7.092 Single Unit Truck/ Bus 1.650 0.486 3.395 0.001 5.207 Tractor Trailer Truck 2.387 0.411 5.809 0.000 10.881 Existing Site Condition (No DSFS) Base Condition DSFS at PC - I-96/I-69 0.548 0.296 1.851 0.064 1.730 DSFS at PC - US-127/Round Lake Rd 0.901 0.275 3.276 0.001 2.462 DSFS at 350 ft upstream of PC - I-96/36th -0.300 0.215 -1.395 0.163 0.741 Table 12 Random intercept logistic regression result for brake response location (Sites 2-4) Response Variable: Braking Began 200-600 ft Before Curve (N=1,651) Parameter Estimate Std. Error t-value p-value Exp(B) Intercept -2.357 0.500 -4.713 <0.001 0.095 Speed 1000-ft Upstream of the PC 0.026 0.008 3.155 0.002 1.026 Passenger Vehicle with No Trailer Base Condition Passenger Vehicle with Trailer -0.148 0.408 -0.364 0.716 0.862 Single Unit Truck/ Bus -0.379 0.309 -1.230 0.219 0.685 Tractor Trailer Truck -0.087 0.191 -0.458 0.647 0.419 Existing Site Condition (No DSFS) Base Condition DSFS at PC - I-96/I-69 0.368 0.155 2.375 0.018 1.445 DSFS at PC - US-127/Round Lake Rd 0.599 0.151 3.953 <0.001 1.820 DSFS at 350 ft upstream of PC - I-96/36th -0.120 0.149 -0.803 0.422 0.887 Note: Each model included a random intercept for site. Represents data collected from video cameras. 80 First, as expected, the speed of vehicles measured 1,000 ft prior to the curve was strongly correlated with each of the driver response variables. Specifically, this suggests that faster drivers tended to maintain such behaviors regardless of the DSFS presence at the site and is aligned with prior behavioral research. Further discussion of the primary variables of interest, including the effects of sign position, interchange type, and time of day on DSFS performance are provided in the sections that follow. 5.3.1 Effect of sign position The parameter estimates in Table 10 indicate that curve entry speeds were approximately 2.0 to 2.3 mph lower with the DSFS present at the two ramp locations where the sign was positioned directly at the point of curvature. The DSFS installed at the point of curvature also decreased the occurrence of excessive curve entry speeds, as indicated in Table 11. Specifically, compared to the existing site condition, drivers were 1.7 to 2.5 times more likely to enter the curve within 15 mph of the curve advisory speed with the DSFS installed at the point of curvature. Furthermore, Table 12 suggests that the DSFS also improved brake response when posted at the point of curvature. With the DSFS at this location, drivers were 1.4 to 1.8 times more likely to initiate braking between 200 and 600 ft upstream of the curve. It is also worth noting that the DSFS had a similar effect on driver response across all vehicle categories. These results are aligned with prior findings associated with DSFS use in similar freeway ramp settings in the earlier phase of this study. In contrast, the DSFS was found to be ineffective across all measures of driver response when posted 350 ft upstream of the point of curvature. At this particular ramp location, curve entry speeds were found to increase by approximately 1 mph with the DSFS present. The DSFS had no significant effect on reducing excessive curve entry speeds or improving brake response at this 81 particular site. These results are consistent with earlier phases of the research at an alternative ramp location, which found the DSFS to be less effective when positioned 450 ft upstream of the curve (Table 4). One possible reason for the diminishing effectiveness at increasing advance warning distances is that drivers may be more likely to disregard warning messages that are provided too far in advance of a hazard. 5.3.2 Effect of interchange type It was also of interest to compare the effects of DSFS on driver response across different interchange types, including freeway-to-freeway system interchanges and service interchanges. This effect was evaluated by comparing the DSFS results between the I-96/I-69 (system interchange) and US-127/Round Lake Road (service interchange) ramps. Each of these ramps was a cloverleaf-type interchange with a 30-mph ramp advisory speed. The DSFS was installed directly at the point of curvature in both cases, which, as previously noted, elicited improved driver response compared to the further upstream sign position. Considering curve entry speeds, the DSFS had a slightly (0.3 mph) greater speed reduction effect at the I-96 freeway-to-freeway interchange. However, the DSFS was somewhat more effective at reducing excessive curve entry speeds and improving brake response when used at the US-127 service interchange. These findings suggest that the DSFS is equally effective at system and service interchange exit ramps when installed at the point of curvature. Figure 19 provides a graphical representation of the site-specific marginal effects, which demonstrate the effects of the DSFS on curve entry speeds related to the sign position and interchange type. 82 Figure 19 Mean curve entry speeds by site and DSFS presence 5.3.3 Effect of time of day Analysis of the speed trailer data allowed for an assessment of the effects of the DSFS on curve entry speeds across various times of day and light conditions. Again, these data are specific to the US127/Round Lake Rd exit ramp, where the speed trailer, along with the DSFS for a portion of the 36 days, were positioned near the point of curvature. Separate regression models were generated for curve entry speeds based on time of day and lighting conditions, including peak, daytime off-peak, and nighttime. Using these separate models, it was possible to assess whether the DSFS had different effects on the behavior of peak (e.g. commuter) vs. off-peak drivers and during different lighting conditions. The linear regression results for curve entry speeds are displayed in Table 13. Again, to control for unobserved variations between each daily data collection period, a date-specific random effect (intercept) was included in each model. 83 Table 13 Random intercept linear regression results for curve entry speed by peak/off-peak period and light condition (Site 3) Daytime Off-Peak (9:00 AM – 3:30 PM) (N=6,869) Parameter Estimate Std. Error t-value p-value Intercept 40.176 0.301 133.566 <0.001 Passenger Vehicle Base Condition Heavy Vehicle -4.911 0.294 -16.727 <0.001 Existing Site Condition (No DSFS) Base Condition DSFS at Point of Curvature -0.914 0.442 -2.068 0.048 Peak (7:00 AM – 9:00 AM and 3:30 PM to 6:00 PM) (N=3,096) Parameter Estimate Std. Error t-value p-value Intercept 39.694 0.292 135.733 <0.001 Passenger Vehicle Base Condition Heavy Vehicle -5.597 0.517 -10.825 <0.001 Existing Site Condition (No DSFS) Base Condition DSFS at Point of Curvature -0.488 0.435 -1.123 0.270 Nighttime (6:00 PM – 7:00 AM) (N=7,468) Parameter Estimate Std. Error t-value p-value Intercept 38.249 0.241 158.946 <0.001 Passenger Vehicle Base Condition Heavy Vehicle -3.671 0.453 -8.110 <0.001 Existing Site Condition (No DSFS) Base Condition DSFS at Point of Curvature -0.286 0.359 -0.796 0.432 Note: Data were collected by the speed trailer at US-127/Round Lake. Random intercept was data collection date. Table 13 presents several interesting findings related to the effectiveness of the DSFS by time-of-day. Most notably, the DSFS has the greatest effect on reducing curve entry speeds during daytime off-peak periods. The magnitude of this DSFS speed reduction effect (0.9 mph) is nearly double that observed during peak periods (0.5 mph) and triple that observed during nighttime periods (0.3 mph). This finding suggests that the DSFS has the greatest effects on non-commuter drivers compared to commuter drivers, who are generally more familiar with the site. It should be noted that preliminary models did not show any discernable differences in the DSFS effects on curve entry speeds across the vehicle type categories. Graphical representations of the curve entry speeds by time-of-day and DSFS presence are provided in Figure 20. 84 It is important to note that, unlike the video data, it was not possible to track vehicles through the site, as the speed trailer only records a single speed measurement for each vehicle. Thus, it was not possible to correlate driver speeding tendencies upstream of the site with their curve entry speed. As a driver’s upstream speed was found to be highly correlated with curve entry speed during the analysis of the video data, the inability to control for general speed selection tendencies was an important limitation to any inference drawn from the speed trailer data. Figure 20 Mean curve entry speeds by DSFS presence and time of day 5.4 Conclusion and Direction for Further Evaluation This chapter described a series of field evaluations at three freeway exit ramps with significant horizontal curvature to assess the effectiveness of the 15-inch TraffiCalm DSFS sign across various sign positions (at the point of curvature vs. 350 ft upstream), interchange types (system vs. service), times-of-day (peak vs. off-peak), light conditions (daylight vs. darkness), and vehicle types (passenger vehicles vs. trucks). 85 Compared to the pre-DSFS site condition, the DSFS reduced curve entry speeds and improved brake response at the two ramp locations where the sign was positioned at the point of curvature, during which curve entry speeds were reduced by approximately 2 mph compared to the pre-DSFS conditions. These findings were consistent between the system- and service- interchanges and across all vehicle types. The feedback sign was found to have an ineffective impact on driver behavior when positioned 350-ft upstream of the point of curvature. Consistent with the prior phase of this study, it was concluded that drivers are more likely to disregard speed warning messages when provided too far in advance of the curve. The DSFS was also found to be more effective during daytime off-peak periods than during peak periods and during darkness. This finding suggests that the DSFS has the greatest effects on non-commuter drivers and may be reflective of such drivers being more likely to respond to the speed warning message due to a reduced familiarity with the ramp geometry. On the other hand, commuter drivers are less likely to respond to the DSFS due to familiarity with the site, which may also at least partially explain the lack of an effect for nighttime drivers. While the evaluations described within this chapter this study provided further confirmation that the DSFS is most effective when placed near the PC, further evaluation of DSFS at freeway exit ramps under an expanded set of conditions and at additional ramp locations is recommended. This includes evaluating whether the behavioral effects diminish over time, effect of sign size (both the feedback display and the sign itself), effect of the side of the roadway (right- side vs. left-side/gore area), and the optimal radar detection settings to maximize driver response. These aspects were included in the next phase of field evaluations described in the following chapters. 86 6. EVALUATION OF SIGN SIZE, RADAR DETECTION RANGE, AND LATERAL SIGN POSITION 6.1 Study Design and Site Characteristics Additional research was conducted at the NB US 127/Round Lake Road exit ramp to evaluate several additional aspects of the DSFS, including feedback message display size (15-inch vs. 18-inch), sign border type (yellow border vs. no border), sign lateral placement positions at the PC (right side-mount vs. forward-mount [i.e., gore area]), and vehicle detection range (normal detection vs. late detection). In addition to the 15-inch TraffiCalm sign, two additional full-matrix DSFS were included in this field evaluation, including an 18-inch TraffiCalm sign, and an 18-inch sign manufactured by All Traffic Solutions (ATS). Each sign was individually installed and tested at identical locations near the start of the exit ramp curve, in both the traditional right-side-mount and an alternative forward-mount within the exit gore area. While the two TraffiCalm signs possessed a yellow border with reflective sheeting, the ATS sign did not include a border. Furthermore, the ATS sign utilized a radar system that provided greater vehicle detection ranges compared to the two TraffiCalm signs, consequently causing the ATS feedback message to be initially activated at a greater distance upstream, which may impact driver response while approaching the ramp curve. This was deemed a crucial variable for this evaluation and, consequently, the location of the approaching vehicle when the feedback message was initially activated (i.e., the sign activation location) was assessed for each subject vehicle, in addition to the other behavioral and vehicle related characteristics. As previously noted, the NB US 127/Round Lake exit ramp was a service interchange with a mainline speed limit of 75 mph and a loop ramp with a curve advisory speed of 30 mph. This site possessed warning signage that was compliant with the MUTCD, including W13-6, E5-1a, 87 and W1-8R signs. A plan view of the existing signage layout and DSFS positions for this site is provided in Figure 21. From the site, speed data and message activation location were collected for vehicles approaching and entering the curve across the various sign test conditions using appropriate methods. Further information on the study design and collection of data is included later in this section. Figure 21 NB US-127 Exit Ramp to Round Lake Rd 6.1.1 Sign lateral installation position A primary condition that was evaluated during the earlier phases of this research project was to vary the longitudinal position of the DSFS relative to the exit ramp curve. As previously noted, during these initial evaluations, the DSFS was consistently found to be most effective when positioned at or near the PC of the ramp curve, with the sign losing effectiveness at distances of 88 350 ft or greater upstream from the curve. However, at many exit ramp locations, the typical right- side-mount DSFS installation near the PC is not feasible due to terrain or obstructions, such as a bridge abutment or guardrail, which would potentially block the radar and/or the motorists’ view of the sign. To that end, a primary variable for the present evaluation was to vary the lateral position of the DSFS with respect to the side of the ramp (68, 69). Specifically, in addition to the traditional right-side installation at the PC, the DSFS was installed and tested in the ramp gore area, which positioned the sign on the left side of the ramp, between the ramp and the mainline freeway near the green E5-1a Exit sign. In this position, the sign was approximately 40 feet beyond the PC, but directly in front of motorists while approaching the curve in the ramp auxiliary lane. This setup is referred to herein as the “forward-mount” setup, as displayed in Figure 22a along with the traditional right-side-mount setup (herein referred to as “side-mount”) example in Figure 22b. a) Forward-mounted DSFS b) Side-mounted DSFS Figure 22 Forward-mounted vs. side-mounted DSFS installation 89 A total of six DSFS conditions were evaluated and compared to the existing site condition without the DSFS. The sign test conditions and corresponding data collection periods were sequenced as follows: 1. Existing site condition (prior to DSFS installation) 2. Side-mounted 18-in TrafficCalm sign 3. Side-mounted 15-in TrafficCalm sign 4. Side-mounted 18-in ATS Sign 5. Forward-mounted 18-in TrafficCalm sign 6. Forward-mounted 15-in TrafficCalm sign 7. Forward-mounted 18-in ATS Sign 8. Existing site condition (2 months after DSFS removal) After collection of data under the existing site condition, the DSFS were then installed and evaluated according to the sequence shown above. The DSFS were mounted by MDOT maintenance crews on dual aluminum sign posts, with a 7-ft bottom mounting height from the pavement surface. The existing signage at the site was not modified in any way. The initial DSFS installation remained operational for seven days prior to initiating data collection to allow for dissipation of any driver novelty effects associated with the new traffic control device. For each subsequent change to the DSFS condition, a period of two days was allowed to pass prior to data collection. After completion of all DSFS test conditions, the sign was removed and data were again collected 2-months later under the pre-existing site conditions. 90 6.1.2 Feedback message activation location It is crucially important for the DSFS to be activated when vehicles are at an appropriate distance in advance of the curve to allow enough time and space to react and decelerate accordingly. As noted previously, the vehicle detection ranges varied between the two radar systems, which subsequently affected when the feedback message would activate for approaching vehicles. To assess the relationship between the message activation point and driver response, the location of each subject vehicle during the initial display of the feedback message was recorded utilizing an elevated video camera temporarily installed on the roadside during each data collection period. The videos were later reviewed to extract the location of the vehicle to the nearest 10 feet using a series of reference markers painted on the shoulder. For consistency purposes, all sign activation measurements were referenced to the point of curvature. The sign activation location data were then merged with the LIDAR speed data for each corresponding subject vehicle, and were included as a predictor variable in the subsequent analysis. However, prior to analyzing the data, the sign activation data were first categorized, as follows: • Less than 250 ft upstream of the PC (late activation); • 250 to 400 ft upstream of the PC (normal activation); and • Greater than 400 ft upstream of the PC (early activation). The 250 ft threshold was selected to represent the approximate braking distance necessary for a vehicle to comfortably decelerate (at 11.2 ft/sec2) from 60 mph (the approximate 85th percentile speed at this point) to the curve advisory speed of 30 mph. The 400 ft threshold was selected as it represented the typical passenger vehicle detection range of the TrafficCalm radar. 91 6.2 Data Summary Speed data and message activation location were collected for vehicles approaching and entering into the curve across the various sign test conditions. Vehicle speeds were continuously tracked using the two-person LIDAR hand-off method for the entire exit ramp lane, beginning from the start of the taper, and continuing to the curve entry point. The location of each vehicle upon DSFS activation was determined using an elevated video camera temporarily installed on the roadside during each data collection period. The speed profile data and DSFS activation location data collected for each test condition were joined, organized, and coded into a single file for a comprehensive statistical analysis. The final data set included complete speed profiles for 2,047 vehicle observations, including 1,983 passenger vehicles and 64 heavy vehicles. The descriptive statistics for this data set are shown in Table 14. Table 15 shows the vehicle frequency and percentages based on vehicle detection ranges for different test conditions that include sign types and sign positions. To check for any obvious trends in the data, sources for potential bias, and data distributions, graphical representations of the data were reviewed and descriptive statistics were compared across each data collection condition. Table 14 Descriptive statistics for ramp vehicles (Site 3) Parameter Minimum Maximum Mean Std. Dev. Speed 1,000-ft Upstream of PC, mph 37.606 86.000 66.513 5.916 Speed 400-ft Upstream of PC, mph 28.000 76.825 56.230 6.300 Speed 250-ft Upstream of PC, mph 27.000 71.434 52.460 5.900 Speed At PC, mph 22.000 58.720 41.192 5.011 Passenger Car 0 1 0.969 0.174 Heavy Vehicle 0 1 0.031 0.174 Without DSFS (prior to sign installation) 0 1 0.161 0.368 Side-mounted DSFS 0 1 0.323 0.344 Forward-mounted DSFS 0 1 0.428 0.353 Without DSFS (2 months after removal) 0 1 0.088 0.283 N = 2,047 vehicles 92 Table 15 Frequency distribution of vehicle detection ranges based on sign positions (Site 3) Parameters Activation Location (ft) Frequency Percentage Without DSFS NA 330 100.00% (prior to sign installation) (n=330) Side-mounted DSFS 15-in TrafficCalm (n=188) <250 127 67.55% 250-400 45 23.94% >400 16 8.51% 18-in TrafficCalm (n=151) <250 55 36.42% 250-400 53 35.10% >400 43 28.48% 18-in ATS Sign (n=321) <250 4 1.25% 250-400 37 11.53% >400 280 87.23% Forward-mounted DSFS 15-in TrafficCalm (n=200) <250 86 43.00% 250-400 51 25.50% >400 63 31.50% 18-in TrafficCalm (n=205) <250 80 39.02% 250-400 78 38.05% >400 47 22.93% 18-in ATS Sign (n=471) <250 137 29.09% 250-400 33 7.01% >400 301 63.91% Without DSFS (n=180) NA 180 100.00% (2 months after sign removal) N = 2,047 vehicles 6.3 Results and Discussion Several measures of effectiveness related to vehicle speed were analyzed to determine the effects of the DSFS as a function of sign size/type, lateral installation position, and activation location. The dependent variables for these analyses were selected to assess driver response to the feedback message and included: • Speed at the point of curvature (i.e., curve entry), • Speed 250-ft upstream of the point of curvature, and • Speed 400-ft upstream of the point of curvature. 93 Additionally, the number of heavy vehicles in the sample was small, which required combining across several test conditions, such that only the DSFS lateral installation position was assessed. The multiple linear regression results for speeds approaching and entering the exit ramp curve for passenger cars and heavy vehicles are shown in Table 16 and Table 17, respectively. To assist with the visualization of the results, Figure 23 displays the parameter estimates and 95 percent confidence intervals for “Speed at PC” across all DSFS test conditions. Table 16 Multiple linear regression results for speeds approaching and entering the ramp curve (passenger cars, n=1,983) Speed At PC Speed 250-ft Speed 400-ft Upstream of PC Upstream of PC Parameters Activation Estimate p-value Estimat p-value Estimate p-value Location (ft) e Intercept 19.461 <0.001 8.116 <0.001 0.936 0.342 Upstream Speed 0.367 <0.001 0.685 <0.001 0.841 <0.001 Without DSFS (prior to sign Base Condition installation) Side-mount DSFS 15-in TrafficCalm <250 -3.290 <0.001 -0.876 0.032 0.401 0.278 250-400 -4.698 <0.001 -2.431 <0.001 -1.100 0.053 >400 -3.594 <0.001 -2.443 0.018 -1.430 0.127 18-in TrafficCalm <250 -3.040 <0.001 -0.857 0.132 0.513 0.319 250-400 -4.299 <0.001 -1.600 0.006 -0.054 0.918 >400 -3.201 <0.001 -1.377 0.035 -0.407 0.493 18-in ATS Sign <250 -2.676 0.174 -1.703 0.385 -1.481 0.405 250-400 -4.554 <0.001 -2.065 0.003 -1.050 0.099 >400 -4.173 <0.001 -2.179 <0.001 -1.365 <0.001 Forward-mount DSFS 15-in TrafficCalm <250 -2.738 <0.001 -1.379 0.004 -0.956 0.029 250-400 -4.663 <0.001 -2.708 <0.001 -1.929 <0.001 >400 -3.648 <0.001 -2.065 <0.001 -1.566 0.003 18-in TrafficCalm <250 -3.538 <0.001 -0.602 0.219 -0.108 0.807 250-400 -5.247 <0.001 -2.495 <0.001 -1.859 <0.001 >400 -5.050 <0.001 -2.941 <0.001 -1.877 0.002 18-in ATS Sign <250 -0.500 0.211 0.009 0.982 0.042 0.908 250-400 -2.290 0.002 -0.913 0.227 -0.609 0.373 >400 -3.260 <0.001 -1.347 <0.001 -0.858 0.003 Without DSFS 0.180 0.625 0.000 0.999 0.344 0.300 (2 months after sign removal) 94 Table 17 Multiple linear regression results for speeds approaching and entering the ramp curve (heavy vehicles, n=64) Speed At PC Speed 250-ft Speed 400-ft Upstream of PC Upstream of PC Parameters Estimate p-value Estimate p-value Estimate p-value Intercept 8.116 0.086 -4.659 0.278 -8.737 0.039 Upstream Speed 0.516 <0.001 0.842 <0.001 0.956 <0.001 Without DSFS Base Condition Side-mount Setup -4.019 0.010 -0.927 0.509 0.919 0.500 Forward-mount Setup -3.519 0.016 -0.400 0.760 0.338 0.790 Figure 23 Linear regression parameter estimates for the reduction in curve entry speed (with 95% confidence intervals) Several interesting findings were observed. First, as expected, the speed of vehicles measured 1,000 ft prior to the curve was strongly correlated with speeds approaching and entering the curve, and this effect was stronger at greater distances upstream of the PC. Specifically, this 95 suggests that faster drivers tended to maintain such behaviors regardless of the DSFS presence at the site and is aligned with the prior field evaluation phases. It is also clear that the DSFS tends to have a stronger effect on drivers as they proceed towards the curve. Considering passenger vehicles, speeds measured 400-ft prior to the curve PC were typically only marginally lower with the DSFS in place. By 250-ft prior to the PC, the speeds had become consistently lower with the DSFS in place, particularly for the forward-mount setup. The DSFS showed the greatest effect on speeds measured at the PC (e.g., curve entry point), where speeds were, on average, 3.5 mph lower with a DSFS present. Although the sample of heavy vehicles was small, the DSFS were similarly effective in reducing the speeds of heavy vehicles at the curve entry point. It is also worth noting that vehicle speeds collected 2-months after the removal of the DSFS had returned to their pre-existing levels. Further discussion of the primary variables of interest, including the effects of lateral sign position, sign size and type, and sign activation location are provided in the sections that follow. Please note that hereinafter, the discussion will primarily focus on the results pertaining to speeds measured at the PC, due to the magnitude of the speed reduction effects at this point. 6.3.1 Effect of sign lateral position In general, both lateral DSFS installation positions elicited a similar effect on curve entry speeds. For passenger vehicles, the reduction in curve entry speed ranged between 3.5 and 4.2 mph for the side-mounted setup, and between 2.4 and 4.5 mph for forward-mounted setup, depending on the sign type and activation location. Although the sample of heavy vehicles was small, speed reductions of 3.5 and 4.0 mph were observed at the curve entry point for the forward- mounted and side-mounted installations, respectively. 96 6.3.2 Effect of sign border and display size An important aspect of this study was to also compare the sign border type (prominent yellow border vs. no border) and size of the feedback display (15-inch vs. 18-inch). First, considering the size of the feedback display, the 15-inch and 18-inch TraffiCalm signs had a similar effect on curve entry speeds. However, there was evidence of a slight interaction effect between the sign border and lateral position. Specifically, the 18-inch TraffiCalm sign had a slightly stronger speed reduction effect than the ATS sign when utilized in the forward mount position. This may be due to the TrafficCalm’s prominent yellow sign border attracting greater attention with the sign posted in the more visually cluttered gore area. Conversely, the ATS sign had a slightly stronger speed reduction effect than the TrafficCalm signs when used in the side- mount condition, which may be due to the more consistent upstream activation of the sign, which is explained in greater detail in the following section. 6.3.3 Effect of message activation location The message activation location was found to have the strongest relationship with curve entry speed. The greatest reductions in curve entry speeds were observed for cases where the feedback message initially activated when subject vehicles were within 250 to 400 ft of the curve. For such cases, the curve entry speeds were approximately 4.5 mph lower compared to cases without the DSFS present. Interestingly, the DSFS was found to be slightly, but consistently, less effective at reducing curve entry speeds when initial activation occurred greater than 400 ft upstream of the curve. Across all test conditions, the DSFS was least effective when the feedback message did not initially activate until the vehicle was within 250 ft of the curve. This was likely due to drivers 97 not being afforded adequate time to react and respond to the message. In such cases, curve entry speeds were 1.3 to 2.5 mph higher than cases where the sign activated for vehicles that were within 250 to 400 ft of the curve. Generally speaking, the effect of sign activation location was dampened when the sign was side-mounted compared to forward-mounted. Late message activation resulted in particularly poor speed reduction performance for the ATS sign when utilized in the forward- mount position, possibly due to the lack of a conspicuous sign border. 6.4 Conclusion and Direction for Further Evaluation This field evaluation phase tested several aspects of the DSFS, including physical characteristics (display size and border type), lateral installation position (side-mount vs. forward- mount), and vehicle detection range of the radar. The NB US-127 exit ramp to Round Lake Rd (Site 3) was singularly utilized for this entire evaluation phase to eliminate site-to-site heterogeneity that would potentially confound the analysis. Compared to the existing site condition, installation of a DSFS near the start of the exit ramp curve resulted in lower speeds of vehicles approaching and entering the curve. Generally speaking, the speed reduction effects of the DSFS increased as the vehicles proceeded towards the curve. The greatest effects were observed at the curve PC (e.g., curve entry point), where overall speeds were 3.5 mph lower, on average, with a DSFS present. Reductions in curve entry speeds were observed across all DSFS test conditions, ranging from 0.5 mph to 5.2 mph depending on the test condition. Although the sample of heavy vehicles was relatively small, DSFS were similarly effective for reducing speeds of heavy vehicles at the curve entry point. It is also worth noting that speeds collected 2-months after removal of the DSFS had returned to prior levels, which further supports the effectiveness of the DSFS as a speed reduction countermeasure at exit ramps. 98 Regarding the effect of the lateral installation position, both the side-mounted and forward-mounted DSFS provided a similar level of effectiveness in reducing curve entry speeds, which was consistent for both passenger vehicles and heavy vehicles. Similarly, considering the size of the feedback display, there was no discernable difference in the speed reduction effects between the 15-inch and 18-inch displays. In terms of the sign border, there was evidence of a slight interaction effect between the sign border and lateral position. Specifically, considering the forward-mount position, the prominent yellow sign border produced a stronger speed reduction effect compared to no border, perhaps because due to the greater need for conspicuity with the sign was posted in the more visually cluttered gore area. The most interesting results were related to the message activation location, which was found to have the strongest effect on curve entry speed across all of the sign-related variables considered. The DSFS were most effective when the feedback message was initially activated for vehicles that were between 250 and 400 ft upstream of the curve, and this finding was consistent across nearly all test conditions. For cases where the feedback message was initially activated within this range, curve entry speeds were approximately 4.5 mph lower compared to cases without the DSFS present. Interestingly, the DSFS was found to be slightly less effective at reducing curve entry speeds when initial activation occurred for vehicles further than 400 ft upstream of the curve. This finding is consistent with earlier phases of this research project, which found the DSFS to lose effectiveness when installed 350 ft or further upstream from the curve, as drivers are more likely to disregard speed warning messages when provided too far in advance of the hazard. Not surprisingly, the DSFS was least effective when the feedback message did not activate until the vehicle was within 250 ft of the curve. This diminished effectiveness was likely due to drivers not being afforded adequate time to react and respond to the message. 99 The findings from this field evaluation phase provide further evidence that dynamic speed feedback signs are an effective countermeasure for reducing curve entry speeds at freeway exit ramps. Regarding the optimal sign design, radar calibration, and installation characteristics, the DSFS should be positioned near the start of the curve, with the radar calibrated such that the message activates when vehicles are at least 250 ft in advance of the curve in order to provide adequate time for drivers to react and decelerate prior to reaching the curve. In terms of sign size, 15-inch and 18-inch signs were found to be equally effective and may be used interchangeably at freeway exit ramps. A prominent yellow reflective border around the sign is recommended to help improve conspicuity during cases when the sign is activated late and/or when the sign is located in a visually cluttered environment. When necessary, due to obstructions or terrain issues that would otherwise prohibit the traditional right-side mount, the DSFS will likely be similarly effective when installed within the gore area of the ramp (e.g., forward-mount setup) between the green Exit sign and the initial chevron. However, caution should be exercised when positioning the sign in the gore area, due to the increased likelihood that an errant vehicle may collide with the sign in this position. Thus, if possible, installing the sign in the traditional right-side mounting position is preferred. While the results of this evaluation were encouraging, a follow-up evaluation at additional locations was deemed necessary for further confirmation of the effects of both lateral and longitudinal sign installation positions. Furthermore, because the DSFS was often activated by vehicles traveling on the mainline freeway, it was also deemed important to assess the effects of the feedback sign activation on mainline vehicle speed. This follow-up evaluation is described in the following chapter. 100 7. EVALUATION OF SIGN POSITION AND IMPACT TO MAINLINE TRAFFIC 7.1 Study Design and Site Characteristics A field evaluation was performed at two additional freeway interchange ramp locations to further confirm the effects of both longitudinal sign position (at PC vs. upstream of the curve) and lateral sign position (right side-mount at PC vs. forward-mount within the gore area). This evaluation also considered the effect of the feedback sign on the speed of mainline vehicles, which were often observed to activate the feedback message. Two additional freeway exit ramp locations were selected for this phase of the study, including NB US-127 exit to Dunckel Road (Site 5), which was a service interchange, and EB I-96 to NB US-127 (Site 6), which was a system interchange. Both sites were selected due to the proximity of an overpass near the ramp curve, which would potentially require alternative sign installation locations. Both sites had a mainline speed limit of 70 mph and a ramp advisory speed of 25 mph. Both sites possess MUTCD compliant warning signage including W13-6, E5-1a, and W1-8R signs. Plan views of the study sites, including the various DSFS positions are provided in Figures 24 and 25 for sites 5 and 6, respectively. 101 Figure 24 NB US-127 Exit Ramp to Dunckel Rd (Site 5) Figure 25 EB I-96 Exit Ramp to NB US-127 (Site 6) 102 7.1.1 Sign test conditions The field evaluation utilized the 15-inch TraffiCalm sign and the 18-inch TraffiCalm sign, each of which is described in further detail in Chapter 3. The 15-inch TraffiCalm sign was installed and evaluated at Site 5 and the 18-inch TraffiCalm sign was installed and evaluated at Site 6. At both sites, the DSFS was first installed beneath the existing exit advisory speed sign (W13-6) and then moved to the point of curvature, as depicted in Figure 26 for Site 6. The advisory speed signs were 485 ft and 580 ft upstream of the PC at sites 5 and 6, respectively, which presented greater upstream distances than had been previously evaluated in this research. In addition to the traditional right-side-mount at the PC, the DSFS was also evaluated in the forward-mount position beneath the Exit sign within the gore area at Site 6 (Figure 26c). The DSFS was not moved to the forward-mount position at site 5 due to safety concerns raised by the MDOT maintenance team regarding frequent gore-area intrusions at this location. Data were collected for the following conditions: 1. Without DSFS; 2. DSFS upstream of the PC (mounted beneath the exit advisory speed sign); 3. DSFS at the PC, right-side-mount; and 4. DSFS at the PC, forward-mount in the ramp gore area (Site 6 only). a) DSFS at advisory speed sign b) DSFS at PC (side-mount) c) DSFS at PC (fwd-mount) Figure 26 DSFS installation locations (Site 6) 103 The feedback message was displayed for approaching vehicles in the same manner employed during the prior evaluation phases, as follows: • Display the measured speed if below 35 mph (i.e., advisory speed + 10 mph) and • Display the measured speed alternating with a “SLOW DOWN” message at 1 hz cycles at 35 mph and above. 7.1.2 Mainline vehicle speeds As mainline vehicles often activate the sign, it was of interest to evaluate the effect of DSFS messages on the speeds of mainline vehicles to determine any adverse impacts on behavior. In general, vehicles traveling in the right-most lane were more likely to trigger the DSFS because of their proximity to the sign. To test the effects of DSFS messaging on the speeds of mainline vehicles, this evaluation phase included collection of spot-speed data at the upstream exit advisory sign for free-flowing mainline vehicles traveling in the right lane with and without DSFS present. The spot speed data for mainline vehicles were collected at sites 5 and 6 using a single handheld LIDAR gun, separately for passenger vehicles and heavy trucks. Data were collected only for mainline vehicles that activated the DSFS during the period that the DSFS was present. 7.2 Data Summary The two-person LIDAR handoff method was utilized to collect vehicle speed profiles for exiting vehicles starting from the exit ramp taper to the PC. The speed profile data collected for each test condition were joined, organized, and coded into a single file for a comprehensive analysis separately for each site. The final data set included complete speed profiles for 456 and 737 vehicle observations from sites 5 and 6, respectively, with descriptive statistics shown in Tables 18 and 19, respectively. 104 Table 18 Descriptive statistics for ramp vehicles (Site 5) Site 5: (NB) US-127 to Dunckel Road (n=456) Parameters Min Max Mean Std. Dev Upstream Speed, mph 46.900 90.000 66.251 5.815 Speed at PC, mph 21.347 58.583 41.023 5.073 Passenger Vehicles 0 1 0.915 0.279 Heavy Vehicles 0 1 0.085 0.279 2 Average Deceleration Rate, ft/sec 0.663 3.799 1.810 0.425 Existing Site Condition (No DSFS) 0 1 0.430 0.476 DSFS at Advisory Speed Limit Sign 0 1 0.235 0.392 DSFS at PC-Side-mount 0 1 0.335 0.445 Table 19 Descriptive statistics for ramp vehicles (Site 6) Site 6: EB I-96 to NB US-127 (n=737) Parameters Min Max Mean Std. Dev Upstream Speed, mph 46.833 79 65.859 0.177 Speed at PC, mph 20 59 34.400 0.145 Passenger Vehicles 0 1 0.959 0.007 Heavy Vehicles 0 1 0.041 0.007 2 Average Deceleration Rate, ft/sec 1.141 3.763 2.318 0.015 Existing Site Condition (No DSFS) 0 1 0.263 0.016 DSFS at Advisory Speed Limit Sign 0 1 0.201 0.015 DSFS at PC-Side-mount 0 1 0.281 0.017 DSFS at PC-Forward-mount 0 1 0.255 0.016 7.3 Results and Discussions Two measures of effectiveness related to ramp vehicle speeds were analyzed to evaluate and further confirm the effects of the DSFS as a function of the longitudinal sign position relative to the PC and lateral sign position (site 6 only). The dependent variables for these analyses included: • Speed at the point of curvature (i.e., curve entry), and • Speed at the exit advisory sign (i.e., 485 ft [site 5] or 580 ft [site 6] upstream of the PC). 105 Preliminary models showed only minor differences in DSFS effectiveness between passenger vehicles and heavy vehicles and further analysis of vehicle-specific effects was not performed. For all models, the upstream approach speed was included as a covariate to control for general driving behavior. The linear regression results from sites 5 and 6 are presented in Tables 20 and 21, respectively. Table 20 Multiple linear regression results for curve approach and entry speeds (site 5) Speed at PC Speed at Exit Advisory Sign Parameter Estimate p-value Estimate p-value Intercept 16.534 <0.001 8.727 <0.001 Upstream Speed 0.387 <0.001 0.778 <0.001 Passenger Vehicles Base Condition Heavy Vehicles -5.551 <0.001 -4.395 <0.001 Existing Site Condition (No DSFS) Base Condition DSFS 485 ft Upstream of PC 0.073 0.876 -1.960 <0.001 DSFS at PC-Side Mounted -2.250 <0.001 -0.998 <0.007 Table 21 Multiple linear regression results for curve approach and entry speeds (site 6) Speed at PC Speed at Exit Advisory Sign Parameter Estimate p-value Estimate p-value Intercept 14.878 <0.001 2.483 0.147 Upstream Speed 0.315 <0.001 0.883 <0.001 Passenger Vehicles Base Condition Heavy Vehicles -3.789 <0.001 -1.833 0.003 Existing Site Condition (No DSFS) Base Condition DSFS 580 ft Upstream of PC -0.300 0.409 -1.535 <0.001 DSFS at PC-Side Mounted -1.939 <0.001 -1.172 <0.001 DSFS at PC-Forward Mounted -1.714 <0.001 -2.011 <0.001 Several interesting findings were observed. First, as expected, the upstream speed of vehicles prior to the curve was strongly correlated with speeds approaching and entering the curve, and this effect was stronger for speeds measured at the exit advisory sign. Specifically, this suggests that faster drivers tended to maintain such behaviors regardless of the DSFS presence at 106 the site and is consistent with the prior phases of this research. Further discussion of the primary variables of interest, including the effects of sign position longitudinal sign position from the PC, lateral sign position at the PC, and effect of the DSFS on mainline vehicle speeds are provided in the sections that follow. 7.3.1 Effect of longitudinal sign position The parameter estimates in Tables 20 and 21 show that curve entry speeds were approximately 1.9 to 2.3 mph lower with the DSFS positioned directly at the point of curvature for sites 5 and 6, respectively, compared to when the DSFS was not present. On contrary, when DSFS was installed at the advisory speed signs, no significant reduction in the curve entry speeds was observed. These findings further strengthen the argument for positioning the DSFS at or near the PC for the greatest reduction in curve entry speeds (68–71). The speeds measured at the exit advisory sign locations upstream of the PC were significantly lower for all DSFS installations compared to the site without the DSFS. But again, these effects were only sustained for cases where the DSFS was installed near the PC. Although one could argue that the DSFS positioned at the PC would not typically be able to detect vehicles at such great distances upstream of the radar (485 ft and 580 ft in this case). However, it was hypothesized that drivers of exiting vehicles may have been prematurely reacting to the activation of the DSFS caused by mainline vehicles. A graphical representation of the model results for speed at PC for the various longitudinal DSFS installation positions is shown in Figure 27. 107 45 40 Mean Speed At PC, mph 35 30 25 20 15 10 5 0 Site 5: NB US 127 to Dunckel Road Site 6: EB I-96 to NB US-127 Without DSFS DSFS At PC DSFS At Curve Advisory Speed Sign Figure 27 Mean curve entry speeds based on DSFS presence and longitudinal position 7.3.2 Effect of lateral sign position Both side-mounted and forward-mounted sign positions showed a similar reduction in curve entry speeds. Compared to the site without the DSFS, curve entry speeds were 1.9 mph and 1.7 lower with the side-mounted and forward-mounted DSFS present, respectively. However, the forward-mounted installation resulted in greater speed reduction at the upstream exit advisory sign compared to the side-mounted installation, likely due to the improved upstream visibility of the sign when positioned in the gore area. 7.3.3 Effect on speeds of mainline vehicles The mainline vehicle speed data are presented in Table 22. Again, while the DSFS was present, speeds were only measured for cases where the DSFS was activated by the subject vehicle. The average speed at site 5 for passenger cars with and without the DSFS was 73.6 and 73.9 mph, respectively. The same measures for heavy vehicles were 65.3 and 65.5, respectively. For site 6, 108 the average speed for passenger cars with and without the DSFS was 73.6 mph and 73.4 mph, respectively. For the heavy vehicles, the same measures were 64.7 mph and 64.5 mph, respectively. Two sample t-tests found these differences in average speeds to not be significantly different with respect to DSFS presence. Table 22 Mainline vehicle speed summary Mainline Vehicle Speed Summary by Vehicle Types, mph Site No Test Conditions Min Max Mean Std. Dev 5 Existing Site Condition (No DSFS)-PSGR (n=90) 67 86 73.633 3.752 DSFS at Curve Advisory Speed Sign-PSGR (n=85) 67 88 73.894 3.745 Existing Site Condition (No DSFS)-HV (n=19) 61 75 65.263 3.142 DSFS at Curve Advisory Speed Sign-HV (n=29) 60 70 65.483 2.459 6 Existing Site Condition (No DSFS)-PSGR (n=96) 63 82 73.635 3.440 DSFS at Curve Advisory Speed Sign-PSGR (n=73) 66 82 73.417 3.672 Existing Site Condition (No DSFS)-HV (n=39) 60 68 64.684 2.195 DSFS at Curve Advisory Speed Sign-HV (n=30) 58 68 64.517 2.530 Note: PSGR-Passenger Vehicle, HV-Heavy Vehicle 7.4 Conclusion and Direction for Further Evaluation This field evaluation phase further confirmed the effects of both longitudinal sign position (at PC vs. upstream of the curve) and lateral sign position (right side-mount at PC vs. forward- mount within the gore area). This evaluation also considered the effect of the feedback sign on the speed of mainline vehicles, which were often observed to activate the feedback message. The results found DSFS installed at the curve PC to produce significantly lower speeds for vehicles approaching and entering the curve compared to when the DSFS was not present. The average curve entry speeds were 1.7 mph to 2.3 mph lower with the DSFS installed at the PC compared to the existing site without the DSFS. However, positioning the DSFS at the upstream exit advisory sign proved ineffective, as curve entry speeds were not significantly different than when no DSFS 109 was present. This finding provides further support for the hypothesis that drivers are more likely to disregard warning messages that are provided too early with respect to the hazard. Collectively, these findings further confirm the results from the prior phases of this research, which found the DSFS to be most effective at reducing curve entry speeds when installed near the PC. Additionally, it was further confirmed that both the side-mount and forward-mount DSFS provided similar effects on driver behavior, with the forward-mount position providing slightly earlier speed reductions, likely due to the greater visibility of the sign when positioned within the gore area. However, the likelihood of collision with the sign may increase when positioned in the gore area, and further investigation of the long-term viability of DSFS installed in the gore area is recommended. Finally, the activation of the DSFS did not have any significant effect on the speeds of mainline vehicles. The results of this evaluation further confirmed the effects of sign positioning on the behavior of exiting drivers, and indicated no substantive impacts of DSFS on mainline traffic. However, one very important aspect regarding traffic control devices is to know whether faster drivers are impacted more than other drivers, as they are of particular concern due to the elevated risk of exceeding the safe curve speed. Thus, a follow-up evaluation was necessary to evaluate the impact of DSFS on drivers at different speed percentiles, particularly, that which included faster drivers. This follow-up evaluation is described in the following chapter. 110 8. EFFECTS ON DRIVERS AT DIFFERENT SPEED PERCENTILES 8.1 Study Design The effects of DSFS on drivers at different speed percentiles, particularly on the faster drivers, are of specific interest due to the greater likelihood of a speeding-related crash. Again, faster vehicles must decelerate in a shorter time-space dimension to make a safe maneuver as compared to slower vehicles. While prior studies have reported average speed reduction and percent reduction in speeding after installing DSFS, no prior research explored the impact of DSFS on drivers across different speed percentiles. To address this gap, this study was designed to compare the impact of DSFS on drivers at different speed percentiles (i.e., faster drivers, average drivers, slower drivers) at freeway exit ramps. LIDAR speed profile data from the NB US 127/Round Lake Road exit ramp (site 3) were utilized to conduct this analysis. As previously noted, the NB US 127/Round Lake exit ramp is a service interchange with a mainline speed limit of 75 mph and a loop ramp with a curve advisory speed of 30 mph. A plan view of the existing signage layout and DSFS positions for this site was provided previously in Figure 21. As described in Chapter 6, three DSFS were tested both at side-mounted and forward-mounted positions. The results showed a lower reduction in speed for the cases when the sign was activated late (within 250-ft of the PC). To control for the variability due to sign installation and activation location, only cases when the sign was initially activated for vehicles that were greater than 250-ft in advance of the PC and when DSFS was installed at the right side of the ramp (typically installation location) were included in this analysis. 111 8.2 Data Summary Vehicle speed profiles were collected before installing the DSFS and after the installation of each type of DSFS. Following the data collection, vehicle speed profile data for each data condition were joined and the final datasets include 847 vehicle observations. To analyze the effects of the DSFS on drivers approaching at different speeds, the speed profile datasets were partitioned into three groups based on the 15th and 85th percentile speeds measured at the upstream location. Speed at the upstream location for the free-flowing vehicles was assumed to be not influenced by the sign or curve geometry and represent the driver's general driving behavior. The categories were defined as 1. slower drivers (≤ 15th percentile upstream speed) 2. average drivers (>15th percentile upstream speed and <85th percentile upstream speed) 3. faster drivers (≥85th percentile upstream speed) Table 23 presents the descriptive statistics including minimum, maximum, mean, and standard deviation for drivers at different speed percentiles. Table 23 indicates that the average speed from the upstream location to the PC (with and without DSFS combined) decreased by 20.5 mph, 34.0 mph, and 38.6 mph for slower drivers, average drivers, and faster drivers, respectively. Figure 28 presents average vehicle speed profiles for drivers across the different speed percentiles for both test conditions (e.g., with active DSFS vs. without). A comparison of the slopes of the lines in Figure 28 indicates that fasters drivers reduced speed more (i.e., used greater deceleration) when DSFS was in operation compared to the slower and average drivers. Whereas the average difference between the faster and slower drivers at the upstream location was 17.4 mph, this difference had reduced to 7.8 mph at the curve entry point. 112 Table 23 Descriptive statistics for vehicle speeds for different driver types Parameter Minimum Maximum Mean Std. Dev. Upstream speed (1,100-ft in advance of the PC), mph For slower drivers 44.53 62.00 58.94 3.01 For average drivers 62.04 73.34 68.12 3.03 For faster drivers 73.52 86.00 76.29 2.45 Speed 250-ft upstream of the PC, mph For slower drivers 28.00 71.43 46.25 5.45 For average drivers 39.00 65.00 52.98 4.44 For faster drivers 42.50 71.12 58.38 4.86 Speed at the PC, mph For slower drivers 24.00 48.04 37.90 4.99 For average drivers 28.00 57.00 41.70 4.46 For faster drivers 34.89 58.72 45.70 4.82 Figure 28 Average vehicle speed profiles for drivers at different speed percentiles 8.3 Results and Discussion To evaluate the effects of different drivers' responses to the DSFS, separate models were developed for slower drivers, average drivers, and faster drivers in addition to a combined model 113 for all drivers. Speeds were estimated at two different critical locations and the dependent variables for the analysis include: • Speed 250-ft upstream of the PC, and • Speed at the PC (i.e., curve entry). The multiple linear regression results for speeds at these two locations are presented in Table 24. The table presents the model results for the slower drivers, average drivers, faster drivers, and all drivers combined. The table contains parameter estimates and their significance level (p- value). The effects of the sign on drivers can be estimated from the “With DSFS” parameter estimates where the change in speed compared to the without DSFS conditions is presented. For example, from Table 24, compared to the without DSFS conditions, speed at the PC of the freeway exit ramp curve was 4.0 mph, 4.5 mph, and 5.2 mph lower when the DSFS was in operation for the slower drivers, average drivers, and faster drivers, respectively. Table 24 Multiple linear regression results for speeds approaching and entering the freeway exit ramp curve Slower Drivers Average Drivers Faster Drivers All Drivers Speed 250-ft Upstream of the PC, mph Parameters Estimate p-value Estimate p-value Estimate p-value Estimate p-value Intercept 3.843 0.623 15.585 <0.001 8.169 0.513 10.796 <0.001 Upstream Speed 0.745 <0.001 0.569 <0.001 0.673 <0.001 0.638 <0.001 Passenger Cars Base Condition Base Condition Base Condition Base Condition Heavy Vehicles -4.707 <0.001 -0.795 0.545 0.155 0.953 -2.744 <0.001 Without DSFS Base Condition Base Condition Base Condition Base Condition With DSFS -1.500 0.081 -2.378 <0.001 -2.856 <0.001 -2.389 <0.001 Speed At PC, mph Parameters Estimate p-value Estimate p-value Estimate p-value Estimate p-value Intercept 20.185 0.005 22.818 <0.001 18.012 0.114 20.533 <0.001 Upstream Speed 0.355 0.003 0.314 <0.001 0.389 0.010 0.350 <0.001 Passenger Cars Base Condition Base Condition Base Condition Base Condition Heavy Vehicles -4.335 <0.001 -0.804 0.523 1.012 0.674 -2.220 0.004 Without DSFS Base Condition Base Condition Base Condition Base Condition With DSFS -3.966 <0.001 -4.453 <0.001 -5.183 <0.001 -4.562 <0.001 114 Several general trends can be observed across three different site types. First, the upstream speed was strongly correlated with the speeds approaching and entering the critical locations. The correlation was stronger at a greater distance upstream of the PC and aligns with the prior findings. Secondly, the speed of the heavy vehicles was consistently lower compared to the passenger cars as the heavy vehicle speed limit was 65 mph compared to the 75 mph speed limit for the passenger cars. Detailed discussions on the effect of the sign on drivers at different speed percentiles are presented in the section below. The results presented in Table 24 suggest that DSFS has a significant effect on vehicle speeds while approaching and entering the ramp curve. On average for all drivers, the speeds at 250-ft upstream of the PC and at the PC were 2.4 mph and 4.6 mph lower compared to the prior conditions without the DSFS, respectively. A more detailed analysis showed the magnitude of speed reduction to vary among the divers across the different speed percentiles. This difference in driver reaction began early, at 250-ft upstream of the PC. At this location, with the DSFS operational, the average drivers showed a speed reduction of 2.4 mph compared to without the DSFS. For the same conditions, the speed reduction with the operational DSFS was 1.5 mph for the slower drivers, and 2.9 mph for the faster drivers. This trend of higher speed reduction by the faster drivers continued to the curve PC, where DSFS was installed and the reduction in speed maximized. At the PC, following the installation of the DSFS, the average drivers' speed was reduced by 4.5 mph compared to the without DSFS condition. For the slower driver, this reduction was 4.0 mph and for the faster drivers, this reduction was 5.2 mph. The difference in speed reduction between the slower and faster drivers is 1.2 mph, which is similar to the difference observed at 250-ft upstream of the PC. All these results suggest that DSFS impacted faster drivers more compared to slower drivers and it started in advance of the PC. The results can be also 115 visualized in Figure 29, where speed reductions by different driver categories along with the 95 percent confidence intervals are presented. The figure also shows a clear trend of higher speed reduction by the faster drivers at both critical locations evaluated here. Figure 29 Linear regression parameter estimates for the reduction in speed at freeway exit ramps (with 95% confidence intervals) 8.4 Conclusions and Direction for Further Evaluation A detailed analysis was conducted to evaluate the effects of DSFS on drivers across different speed percentiles, particularly, the faster drivers because of their elevated crash likelihood while transitioning from a high-speed to a low-speed environment. Vehicle speed profiles collected before and after installing the DSFS at NB US-127/Round Lake Road were used. Observed drivers were categorized into three speed groups: slower drivers, average drivers, and faster drivers based on their upstream speed, which was prior to the influence of curve geometry or DSFS. Results showed the DSFS to have a significant speed reduction effect for motorists approaching and entering the curve, and these speed reduction effects were more pronounced for 116 faster drivers. At 250-ft upstream of the PC, with the DSFS installed, faster drivers reduced their speed by 1.4 mph and 0.5 mph more than the slower drivers and average drivers, respectively. At the PC, the speed reductions with the DSFS installed were 1.2 mph and 0.7 mph greater for the faster drivers than slower and average drivers, respectively. These results collectively provided encouraging evidence on the effectiveness of DSFS in reducing vehicle speed, particularly, for the faster drivers, who require the greatest reductions in speed to safely negotiate the ramp curve. The findings from this study suggest dynamic speed feedback signs to be an effective countermeasure for helping faster drivers to reduce speeds while transitioning from a high-speed to low-speed environment. While the evaluations described within this chapter provided very encouraging results on DSFS having a greater impact on faster drivers, it is also important to evaluate whether sign effects change over time. This is necessary since each of the prior evaluations has been of a short-term nature, having been performed over a matter of days or weeks, in most cases. To this end, a follow- up evaluation on the long-term effects of a permanent DSFS installation was performed, which is described in the next chapter. 117 9. TEMPORAL CHANGES IN DSFS EFFECTIVENESS 9.1 Study Design and Site Characteristics In the early stages of the project, the Michigan DOT permanently installed a DSFS at the WB I-96 exit ramp to SB I-69 (Site 2). With this permanent installation, the research team was able to assess whether the driver behavior effects associated with the DSFS diminished with time. The permanent DSFS was a 15-inch TraffiCalm sign that was identical to the test sign utilized during all phases of this research project. The sign was programmed to display only speed of the vehicles traveling below 40 mph (advisory speed + 10 mph) and alternate with a “SLOW DOWN” message at 0.5-sec intervals for vehicles traveling 40 mph and above. The sign was positioned at the PC and also included a 30-inch square advisory speed panel, mounted immediately below the DSFS. The sign was equipped with a solar panel for a continuous power supply. The research team periodically collected data at this location before installation of the DSFS, and during the initial 14 months following installation. 9.2 Data Summary Data were collected from the location in four phases: 1) prior to installation of the DSFS, 2) 3-months after the installation of the DSFS, 3) 9-months after the installation of the DSFS, and 3) 14-months after the installation of the DSFS. Data were collected using a series of three pole- mounted high-definition video cameras temporarily installed at specific points on the approach to and within the curve. The cameras were installed at the same locations and provided similar fields- of-view during each data collection period. The videos were reviewed to extract the relevant driver behavior data using procedures described in Chapter 3. The data were combined and coded into a single file for detailed analysis, and were screened to include only vehicle observations with a 118 headway of 3 seconds or greater. A small number of cases where no feedback message was displayed on the DSFS for an approaching vehicle were also removed from the dataset. The final combined data set included a complete record for 759 vehicle observations collected during four data collection periods. Table 25 presents the descriptive statistics of the combined dataset. Table 25 Descriptive statistics for evaluation of temporal effects of DSFS Descriptive Statistics (N=759) Parameter Minimum Maximum Mean Std. Dev Speed 1000-ft Upstream of the PC 32.468 83.832 55.978 7.637 Speed at the PC 19.461 52.395 35.963 5.676 Passenger Vehicle with No Trailer 0 1 0.665 0.472 Passenger Vehicle with Trailer 0 1 0.038 0.192 Single Unit Truck/ Bus 0 1 0.248 0.432 Tractor Trailer Truck 0 1 0.049 0.215 Without DSFS at the Site 0 1 0.510 0.500 DSFS at PC- After 3 Months 0 1 0.194 0.395 DSFS at PC- After 9 Months 0 1 0.211 0.408 DSFS at PC- After 14 Months 0 1 0.086 0.280 9.3 Results and Discussions Two measures of effectiveness related to curve entry speeds were analyzed over time to determine the temporal effect of the DSFS. Separate models were developed using appropriate statistical procedures. The dependent variables for these models included: • Speed at the point of curvature (i.e., curve entry), and • Probability of a vehicle entering the curve with a speed <45 mph. Preliminary models suggested only minor differences in DSFS effectiveness between passenger vehicles and heavy vehicles and further analysis of vehicle-specific DSFS effects was not performed. For all models, the upstream approach speed was included as a covariate to control for general driving behavior. The linear regression results for curve entry speeds (at the PC) 119 obtained from the video data are displayed in Table 26, while the binary logistic regression results for excessive curve entry speeds are displayed in Table 27. A graphical representation of the curve entry speeds over time is presented in Figure 30. First, the vehicle speed measured 1,000 ft upstream of the PC was strongly correlated with both driver response variables considered in this study. This indicates that faster driving vehicles largely maintain their behavior irrespective of the presence of DSFS, and was consistent with all other phases of the field evaluations. The following sections provide further discussion of the primary variables of interest. Table 26 Linear regression results for curve entry speed Response Variable: Speed at PC Parameter Estimate Std. Error t-value p-value Intercept 15.342 1.321 11.610 <0.001 Speed 1000-ft Upstream of the PC 0.403 0.022 18.648 <0.001 Passenger Vehicle with No Trailer Base Condition Passenger Vehicle with Trailer -1.076 0.674 -1.598 0.111 Single Unit Truck/ Bus -1.141 0.619 -1.841 0.066 Tractor Trailer Truck -3.026 0.370 -8.173 <0.001 Without DSFS at the Site Base Condition DSFS at PC- After 3 Months -2.675 0.351 -7.610 <0.001 DSFS at PC- After 9 Months -1.854 0.343 -5.405 <0.001 DSFS at PC- After 14 Months -2.116 0.481 -4.402 <0.001 9.3.1 Curve entry speeds The parameter estimates in Table 26 indicate a significant reduction in curve entry speed following the installation of the DSFS, and the effect remained relatively consistent over the entire 14-month study period. The parameter estimates in Table 26 show that the curve entry speeds were lowest immediately following DSFS installation (2.7 mph lower than before installation), and increased slightly after 9 months and 14 months of installation. These findings clearly suggest that the DSFS does not lose its effectiveness in reducing curve entry speeds with time. 120 50 40.10 38.25 37.99 40 37.43 Speed, mph 30 20 10 0 Without DSFS DSFS at PC- After 3 Months DSFS at PC- After 9 Months DSFS at PC- After 14 Months Figure 30 Average curve entry speeds before DSFS installation and during the 14-months after DSFS installation 9.3.2 Extreme curve entry speeds The DSFS also decreased the occurrence of excessive curve entry speeds following the installation of the sign, although this effect diminished slightly over time. The elasticity values calculated from the parameter estimates in Table 27 show that vehicles were 4.2, 2.6, and 2.0 times more likely to enter the curve with a speed lower than 45 mph after 3, 9, and 14 months of installation, respectively, compared to before the DSFS was installed. Table 27 Binary logistic regression results for curve entry speed <45 mph Response Variable: Speed at PC <45 mph Parameter Estimate Std. Error t-value p-value Exp(B) Intercept 15.265 2.165 7.052 <0.001 4259478 Speed 1000-ft Upstream of the PC -0.214 0.033 -6.420 <0.001 0.807 Passenger Vehicle with No Trailer Base Condition Passenger Vehicle with Trailer -0.295 1.095 -0.269 0.788 0.745 Single Unit Truck/ Bus 15.819 2714.632 0.006 0.995 7418324 Tractor Trailer Truck 15.635 1172.820 0.013 0.989 6167362 Without DSFS at the Site Base Condition DSFS at PC- After 3 Months 1.436 0.770 1.866 0.062 4.204 DSFS at PC- After 9 Months 0.960 0.765 1.254 0.210 2.611 DSFS at PC- After 14 Months 0.699 1.060 0.659 0.510 2.011 121 9.4 Conclusion and Direction for Further Evaluation This research phase sought to evaluate temporal changes in the effect of DSFS on driver behavior over the initial 14-months after installation of a DSFS at a single system interchange ramp. Driver behavior data were collected before installing the sign, and 3 months, 9 months, and 14 months after installation of the sign. Compared to the period prior to DSFS installation, the DSFS reduced curve entry speeds both initially, and throughout the 14-months after installation. The DSFS also reduced the number of vehicles entering the curve at speeds of 45 mph or above. Although these data were only collected at a single site, the results suggest that the DSFS does not lose effectiveness as a ramp speed reduction treatment over time. It is recommended that future research continue to monitor curve entry speeds both at this location and at future DSFS freeway exit ramp installations. 122 10. CONCLUSIONS AND RECOMMENDATIONS 10.1 Conclusions A series of field evaluations were performed at six freeway interchange ramps to assess the impacts of speed feedback signs on various measures of driver behavior, particularly speed on the approach and entry to the ramp curve. Three commercially available speed feedback signs, which varied in size, border type, and radar detection range, were utilized during the field evaluations. The field evaluations were performed across multiple phases, each of which assessed important aspects related to the design, operation, and/or installation of the DSFS, including: • sign messaging strategy, • longitudinal positioning of the sign relative to the ramp curve, • lateral positioning of the sign with respect to the side of the ramp, • sign dimensions and other physical characteristics, • radar activation range, • time of day, • interchange type, • drivers at different speed percentiles, and • temporal changes in driver behavior. Considering all phases of the field evaluation, it was concluded that dynamic speed feedback signs are an effective countermeasure for reducing speeds of vehicles approaching and entering horizontal curves on freeway exit ramps. The most critical aspect influencing the effectiveness of the DSFS as a speed reduction countermeasure was the longitudinal positioning of the sign relative to the ramp curve. Generally speaking, the DSFS was effective across all sites 123 and all test conditions as long as the sign was positioned within 250 ft of the start of the curve. Specifically, a DSFS positioned near the start of the curve resulted in curve entry speeds that were, depending on the condition, 1.5 mph to 5.3 mph lower than without a DSFS present. In contrast, the DSFS was consistently ineffective when positioned at greater distances upstream of the curve, perhaps due to drivers’ tendencies to disregard warning messages that are provided too far in advance of the hazard. Regarding the lateral sign position, the DSFS provided similar effects on driver behavior when installed in either the traditional right-side-mount or forward-mount positions, although the forward-mount contributed to speed reductions beginning further upstream. This was likely due to the greater visibility of the sign when positioned within the gore area, particularly for locations where a bridge overpass or other sight-obstruction immediately proceeds the ramp curve. The strongest sign-related effects were associated with the radar detection range. With the DSFS installed near the start of the curve, the lowest curve entry speeds were observed for cases where the DSFS was positioned so that feedback message was activated for vehicles that were within 250 to 400 ft of the start of the curve. Activation of the display panel for vehicles further than 400 ft upstream of the curve did not provide additional speed reduction benefits. Not surprisingly, the DSFS was least effective when the feedback message did not activate until the vehicle was within 250 ft of the curve. This diminished effectiveness was likely due to drivers not being afforded adequate time to react and respond to the feedback message. In terms of sign size, 15-inch and 18-inch display panels were found to be equally effective and may be used interchangeably at freeway exit ramps. A prominent yellow reflective border around the sign is recommended to help improve conspicuity during cases when the sign is activated late and/or when the sign is located in a visually cluttered environment. Interestingly, there was little 124 difference in the speed reduction effects between the various sign messaging strategies, although slight benefits were observed when the speed number was alternated with a SLOW DOWN message, perhaps due to increased conspicuity of the alternating message frames. However, including an advisory speed panel within the DSFS assembly did not have a substantive impact on driver behavior. Regarding DSFS effects on drivers across different speed percentiles, it was found that sign had a greater impact on reducing the speeds approaching and entering the curve for faster drivers compared to the slower and average drivers. Finally, while this research primarily evaluated the short-term effectiveness of DSFS, the speed reduction effects were sustained during the initial 14-months of operation for the lone permanent DSFS installation included in this study. Although the sample of heavy vehicles was somewhat limited across the field evaluations, the DSFS was similarly effective for heavy vehicles and passenger vehicles. In terms of interchange characteristics, the DSFS was equally effective irrespective of the mainline speed limit or ramp advisory speed. Additionally, the effectiveness of DSFS was similar between system interchanges and service interchanges. Finally, the DSFS did not show any significant effect on the speeds of mainline vehicles when activated. 10.2 Recommendations Based on the study findings, the continued use of DSFS as a speed reduction treatment at freeway exit ramp curves is recommended. A series of specific recommendations related to the sign characteristics, operational performance, and installation details are provided in the following list. These recommendations were developed on the basis of providing optimal DSFS performance towards reducing curve entry speeds, lane departures, and associated crashes, along with practical considerations. Further, these recommendations may be utilized by MDOT towards development 125 of guidelines for the use of DSFS at freeway ramps and other highway warning curve applications, which are not addressed in the current MDOT special provision for speed feedback signs. • Site Selection: Select the ramp sites to install DSFS considering the following conditions: o Frequent lane-departure events including run-off and rollover (reported and/or on- site evidence) o Posted ramp advisory speeds at or below 35 mph o Average curve entry speed 10 mph or above advisory o Ramp AADT of 1,000 or higher (15th percentile of MDOT loop ramps) o DSFS sign installation capability ▪ Roadside adjacent to the ramp near the PC can accommodate the installation of the sign; the gore area as an alternative location ▪ Clear visibility of the roadside within 20 feet of the traveled way for at least 600 ft in advance of the ramp curve (not necessary if ramp gore area is used) • Longitudinal sign installation position relative to the point of curvature: Install the DSFS as close to the point of curvature as practical, but not more than 250 ft upstream of the curve. • Lateral sign installation position: Install the DSFS on the right-side of the ramp. When necessary, due to obstructions or terrain issues that would otherwise restrict the traditional right-side mount, the DSFS may be installed within the gore area of the ramp between the green Exit sign and the initial chevron. However, caution should be exercised when positioning the sign in the gore area, due to the increased likelihood that an errant vehicle 126 may collide with the sign in this position. Thus, if possible, installing the sign in the traditional right-side mounting position is preferred. • Sign messaging strategy: Program the sign to display the following messages: o For speeds at or below the advisory speed + 10 mph, display the speed number; o For speeds exceeding the advisory speed + 10 mph, display the measured speed alternating with a “SLOW DOWN” message. The message frames should be alternated at 0.5 second to 1.0 second intervals. o No maximum cap for speed feedback message is recommended. o A minimum speed threshold of 15 mph is recommended for activation of the feedback panel to prevent activation from rain and small objects. o Do not flash the display or utilize the strobe beacon, as the MUTCD specifically prohibits the use of flashing displays (Paragraph 1 in Section 2L.04 of the 2009 MUTCD) and strobe effects (Paragraph 4 of Section 2A.15 of the 2009 MUTCD) on changeable message signs, which the FHWA officially interprets to include radar speed feedback signs. • Sign characteristics: The DSFS should include a full matrix amber LED feedback display capable of displaying characters that are a minimum of 15 inches in height. To help improve conspicuity, the sign should include a prominent border (i.e., warning plaque) that includes microprismatic reflective yellow sheeting with black “YOUR SPEED” text. A supplemental advisory speed plaque with a black legend and border on a yellow background is optional. • Sign activation range: Ensure that the feedback panel activates for approaching vehicles a minimum of 250 ft in advance of the point of curvature. 127 10.3 Limitations and Direction for the Future Research While this research provided substantial evidence of the effectiveness of DSFS as a speed reduction countermeasure at freeway exit ramps across a variety of contexts, a future evaluation should assess the effectiveness of DSFS in reducing the frequency/severity of ramp lane departure crashes. Furthermore, additional long-term evaluations should be performed to further confirm whether the speed reduction effects of DSFS remain consistent or diminish with time. The effectiveness of DSFS as a speed reduction countermeasure across other critical contexts, such as rural highway curves, speed transition zones, highway work zones, and school zones, should also be evaluated. 128 APPENDIX 129 STATE AGENCY SURVEY QUESTIONNAIRE FORM General 1. Does your state utilize Dynamic Speed Feedback Signs (DSFS) for curve warning on high speed roadways? Some examples are provided below, although other signs and/or messages (e.g., “REDUCE SPEED”, “TOO FAST”, etc.) are also acceptable. 2. If so, on what types of roadways (i.e., freeways, two-lane highways) are they used? a) Are they used at interchange exit ramp curves? Sign Specifications and Operation 3. Does your state have a specification or special provision for these signs used for curve warning? If so, please attach the specification/special provision or provide a reference number. Also, if available, please attach some example vendor products/spec sheets/installation photos of such signs that have been utilized in your state along with approximate costs. 4. For high speed curve warning, especially freeway ramps, what types of message(s) or warning(s) are displayed and what are the speed thresholds for vehicles to trigger the message(s)/warning(s)? 5. Do the signs have the ability to communicate with the operations center and/or are the signs integrated with other components, such as ice or pavement sensors? 6. Are you able to collect and analyze the speed data from the sign’s radar unit; alternatively, do you collect speed data using an add-on device? If so, what type of device? 130 Site Selection and Sign Installation 7. How are the curve sites selected? 8. Where is the sign positioned with respect to the curve (at the curve PC, XX ft upstream of PC, etc)? 9. For freeway exit ramp curve installations only - is there any particular location of the sign that best isolates exiting vehicles, while limiting detection of mainline vehicles in order to prevent false triggering of the sign? 10. How often do the signs need to be maintained or replaced due to vehicle strikes or other damage? 11. Do you have any feedback (formal or informal; written or anecdotal) regarding the benefits, drawbacks, lessons learned, driver behavioral impacts, etc regarding the use of the DSFS for curve warning? 131 SUMMARY OF SURVEY RESPONSES, BY STATE Table 28 Key findings from the statewide survey State DOT Key Findings DSFS Use Arizona • Started using temporary DSFS signs on mobile trailers for treatment at high crash locations on the state highway system • Started development of a statewide project to deploy the temporary DSFS signs on all types of roadways including ramps and two-lane highways California • Used DSFS on two-lane rural highway curves, but not too often Florida • Used DSFS on two-lane rural highway curves Illinois • Used DSFS on two-lane rural highway curves and freeway interchange ramps Iowa • Used DSFS and curve warning signs exclusively on two-lane rural highway curves as part of several research projects Mississippi • Used DSFS on two-lane rural highway curves and freeway interchange ramps Missouri • Used flashing beacons on overhead static warning signs on two-lane rural highway curves and freeway curves • Used height detection systems to detect trucks and activate flashers on truck tipping sign in advance of curves New Mexico • Used DSFS on two-lane rural highway curves Oregon • Used DSFS on two-lane rural highway curves and freeway curves • Used curve warning signs at the interchange ramps that display messages according to a weather responsive system (using temperature) indicating hazardous roadway conditions such as low visibility or slippery surface Pennsylvania • Used DSFS on two-lane rural highway curve and freeway to freeway interchange ramps South Dakota • Used DSFS on two-lane rural highway curves Texas • Used DSFS on two-lane rural highway curves and four-lane divided highways Wisconsin • Used DSFS on two-lane rural highway curves and freeway curves Sign Specifications and Operation Arizona • Sign is programmed in a way that • Speeds 1-10 mph above the advisory speed will blink at the driver • Speeds 11-20 mph above will also trigger a strobe light 132 Table 28 (Cont’d) California • Follows sign specifications provided in section 87.14 of STANDARD SPECIFICATIONS published by Caltrans and recommendations include • Comply with the California MUTCD, Chapter 2B • Light Emitting Diode (LED) character display must be capable of displaying the detected vehicle speed within 1 second and remain blank when no vehicles are detected within the radar detection zone • Have the option to flash the pre-set speed limit when the detected vehicle speed is 5 mph higher than the pre-set speed • Characters must be a minimum 15 inches in height and visible from a min distance of 1,500 feet and legible from a min distance of 750 feet • LEDs must be amber and have a wavelength from 590 to 600 nm and rated for a minimum of 100,000 hours • Radar unit must be able to detect up to 3 lanes of approaching traffic and have a speed accuracy of ±1 mph Illinois • Used full matrix LED displays with 18-inch display digit height • Mounted on an 18-foot aluminum pedestal and added a custom 48-in by 18- in plaque stating “YOUR SPEED” • Positioned the sign on the curve, but prior to the point where the curve radius sharpens • Displayed amber lighted “SLOW DOWN” message and the minimum threshold to display a feedback message is generally 5 to 10 mph over the advisory speed • No speed feedback is provided for speeds 20 mph over the advisory speed, and instead only the “SLOW DOWN” message is displayed Iowa • Used speed feedback sign capable of displaying approaching vehicle speed alternating with speed limit information and curve warning sign displaying SLOW DOWN feedback along with a curve sign for vehicles above a certain threshold • Signs were set to activate when an oncoming vehicle’s speed >50th percentile speed for the curve Mississippi • Used static “YOUR SPEED” sign with the dynamic speed feedback display whenever installed • Used curve warning sign and accompanying advisory speed sign on curves Missouri • Used truck tipping (if truck related) horizontal alignment arrows or a text message like “Watch Your Speed”. New Mexico • Follows special provision for speed feedback sign that include • LED display with attached sign or sign unit in combination with LED displays in compliance with the MUTCD • Programmable sign with the capacity to display violator alerts including speed numbers and 'SLOW DOWN' text • Sign display characters maybe 9", 12", 15", 18", and 22" in height • Red-Blue flashing bars or white LED flashing strobes 133 Table 28 (Cont’d) Oregon • Developed an ITS application that runs on the Advanced Transportation Controller and allows the controller to do a variety of ITS things including weather warning systems, queue warning systems, over height warning systems, and curve warning systems • Utilized a forward-facing radar unit and set custom speed thresholds to activate a blank-out sign, flashers, or activate a custom message on a variable message sign • Used a curve warning sign consists of a yellow arrow warning message with a “SLOW DOWN” feedback message underneath the arrow • Integrated radar gun with variable message sign (VMS) to provide curve waring on freeway curves using overhead VMS sign. Programmed the sign to display “SLOW DOWN, SHARP CURVE AHEAD” when approaching vehicles are traveling within the recommended advisory speed (50 mph) for the curve, display “SLOW DOWN, YOUR SPEED IS XX MPH” when a vehicle is traveling between 50-70 mph, display “YOUR SPEED IS OVER 70 MPH” when detected speed is over 70 mph • Used curve warning signs on freeway ramps indicating slippery or icy conditions via establishing communication with roadside RWIS and pavement sensors Pennsylvania • Used a curve warning sign consists of a yellow arrow warning message with a “SLOW DOWN” feedback message underneath the arrow. South Dakota • Follows provision for solar-powered speed feedback sign that include • Sign should have a 28-in by 33-in “YOUR SPEED” faceplate and consist of a white or yellow background with 4-in lettering • The height of the LED numeral on sign display should be 12-inch • Sign should be programmed not to flash at drivers at any time • Sign should provide a solid display of the driver speed followed by a solid display of “SLOW DOWN” as drivers exceed the speed limit of the area Texas • Follows provision for solar powered speed feedback sign that include • Furnish displays that flash the LEDs, when the detected vehicle speed exceeds the posted speed • Sign display capable of displaying “YOUR SPEED” or “SLOW DOWN” in two lines of min 4 in. height • The sign background shall be black and numeric speed display characters shall be 10 inches in height • Radar controller should be FCC compliant K band radar microwave vehicle detector integrated with the sign with a factory preset range of 600 feet • Speed range of at least 5 to 99 mph should be used and the trigger speed shall be adjustable from the DFSS control 134 Table 28 (Cont’d) Texas • Controllers should be capable of providing local control of the unit and shall provide: on/off toggle control of the sign, and a changeable message that reads: “YOUR SPEED” or “SLOW DOWN” which shall be toggle switchable or keypad adjustable • Used dynamic curve warning system to warn and guide motorists through a curve once activated with radar by directing the chevrons to flash sequentially. Follows provision for a dynamic LED curve warning system that includes • LED chevron must be capable of detecting a compact vehicle within 300 feet and it must occur within 112 milliseconds of the vehicle arrival • Signals should be wirelessly transmitted to sequential signs to trigger a predetermined flash duration • Has the capacity to monitor and control the LED chevron signs through a web-based system that allows for management of device settings such as solar and battery output, flash durations, and counting the number of activations Wisconsin • Used overhead curve warning sign for mainline freeway vehicles • Programmed the sign to activate the message “TOO FAST FOR CURVE” at different speed thresholds based on vehicles types • The speed threshold for message activation for car/motorcycle is 67 mph, van/pick-up truck is 65 mph, bus/truck is 55 mph, and semi-trailer/tractor- trailer is 50 mph on a 50 mph roadway with 45 mph curve advisory speed Site Selection and Sign Installation Arizona • Curve sites selected based on crash data filtered for single vehicle roadway departure crash with a possible secondary filter of roll-over crashes • Sign positioning will be dependent on the geometry but will be positioned to provide information to drivers as they enter the curve California • Segments are selected based on crash data • Warning signs should be placed so that they provide an adequate perception reaction time and should not be placed too far in advance of the condition, such that drivers might tend to forget the warning because of other driving distractions, especially in urban areas Illinois • Sites selected based on crash analysis, particularly considering overturn and run-off the road crashes at the curve • Sign installation position must be designed for each location • The sign must be visible at a point where the feedback is actionable by the motorist and should avoid providing feedback too early, where the motorist does not see the need to act • The detection area of the sign must be considered as well • Shielding a portion of the radar detection area can be utilized to limit mainline interference when installed on ramps 135 Table 28 (Cont’d) Iowa • Sites are selected based on crash history, suitability of location to install a sign and collect speed, and willingness of the agency to install the sign • Signs are placed at the same location a static curve advisory sign would be placed following the guidelines Mississippi • Sites are selected based on need and engineering study • Position the sign just in advance of the PC Missouri • Signs are installed to address specific concerns • Signs are placed at the same location a static curve advisory sign would be placed following MUTCD guidelines New Mexico • Signs are installed normally per request • Signs are positioned based on engineering judgment Oregon • Sites are selected based on crash history • Signs are positioned mostly after the static curve warning signs which are typically placed at or just before the point of tangent South Dakota • Sites are selected based on geometrics and crash history • Signs are positioned at the advance curve warning sign and advisory speed plaque according to the MUTCD. Texas • Sites are selected based on crash history, suggestions from law enforcement, and local compliant • Signs are positioned at the advance warning sign Wisconsin • Sites are selected on a case-by-case basis • Signs are positioned near the PC 136 BIBLIOGRAPHY 137 BIBLIOGRAPHY 1. Michigan Traffic Crash Facts: Data Query Tool. Michigan Office of Highway Safety Planning. http://www.michigantrafficcrashfacts.org/datatool/build.php. Accessed Jun. 12, 2017. 2. Strategic Highway Safety Plan. American Association of State Highway and Transportation Officials, Washington, D.C., 1998. 3. Neuman, T. R. Guidance for Implementation of the AASHTO Strategic Highway Safety Plan: A Guide for Addressing Collisions with Trees in Hazardous Locations. Transportation Research Board, 2003. 4. Governor’s Traffic Safety Advisory Commission, State of Michigan. State of Michigan Strategic Highway Safety Plan. , 2004. 5. Datta, T. K., P. T. Savolainen, T. J. Gates, J. J. Kay, N. B. Nicita, S. Parajuli, J. Finkelman, and Wayne State University. Evaluation of Non-Freeway Rumble Strips - Phase II. Publication RC-1627. 2015. 6. Savolainen, P. T., T. J. Gates, B. J. Russo, J. J. Kay, and Wayne State University. Study of High-Tension Cable Barriers on Michigan Roadways. Publication RC-1612. 2014. 7. Governor’s Traffic Safety Advisory Commission, State of Michigan. Update of 2008 Strategic Highway Safety Plan. , 2013. 8. Governor’s Traffic Safety Advisory Commission, State of Michigan. 2017-2018 State of Michigan Strategic Highway Safety Plan. , 2016. 9. Gates, T., P. Savolainen, J. Kay, J. Finkelman, and A. Davis. Evaluating Outcomes of Raising Speed Limits on High Speed Non-Freeways. Michigan. Dept. of Transportation. Office of Research Administration, 2015. 10. Gupta, N., M. S. Mahmud, H. Jashami, P. T. Savolainen, and T. J. Gates. Evaluating the Impacts of Freeway Speed Limit Increases on Various Speed Measures: Comparisons Between Spot-Speed, Permanent Traffic Recorder, and Probe Vehicle Data. Transportation Research Record, 2022, p. 03611981221106481. https://doi.org/10.1177/03611981221106481. 11. Mahmud, M. S., N. Gupta, B. Safaei, H. Jashami, T. J. Gates, P. T. Savolainen, and E. Kassens-Noor. Evaluating the Impacts of Speed Limit Increases on Rural Two-Lane Highways Using Quantile Regression. Transportation Research Record, 2021, p. 03611981211019732. https://doi.org/10.1177/03611981211019732. 138 12. Megat Johari, M. U., N. Megat Johari, P. T. Savolainen, and T. J. Gates. Safety Evaluation of Freeway Exit Ramps with Advisory Speed Reductions. Transportation Research Record, 2022, p. 03611981221099908. https://doi.org/10.1177/03611981221099908. 13. Governor’s Traffic Safety Advisory Commission. Action Plan 2013-2016. Michigan Traffic Safety Engineering Action Team. , 2015. 14. Hallmark, S. L., Y. Qiu, N. Hawkins, and O. Smadi. Crash Modification Factors for Dynamic Speed Feedback Signs on Rural Curves. Journal of Transportation Technologies, Vol. 05, No. 01, 2015, p. 9. https://doi.org/10.4236/jtts.2015.51002. 15. Federal Highway Administration (FHWA). Manual on Uniform Traffic Control Devices (MUTCD). , 2009. 16. The State of Vermont Agency of Transportation. Guidelines for the Use of Radar Speed Feedback Signs on the State Highway System. Jan, 2019. 17. Veneziano, D., Z. Ye, K. Westoby, I. Turnbull, and L. Hayden. Guidance for Radar Speed Sign Deployments. Transportation Research Board, 2012, p. 1. 18. Montana Department of Transportation. Guidelines for the Use of Vehicle Speed Feedback Signs on the State. 2012. 19. Ross, H. E., Ed. Recommended Procedures for the Safety Performance Evaluation of Highway Features. Transportation Research Board, National Research Council ; National Academy Press, Washington, D.C.: Washington, D.C, 1993. 20. American Association of State Highway and Transportation Officials. Manual for Assessing Safety Hardware. 2016. 21. ENTERPRISE Transportation Pooled Fund Study TPF-5 (231). Planning Guidance for Intelligent Transportation System (ITS) Devices: Final Report. 2014. 22. City of Bellevue Transportation Department. Stationary Radar Sign Program 2009 Report. , 2009. 23. Garber, N. J., and S. T. Patel. Effectiveness of Changeable Message Signs in Controlling Vehicle Speeds in Work Zones. Virginia Transportation Research Council, 1994. 24. Nicholas J. Garber and Surbhi T. Patel. Control of Vehicle Speeds in Temporary Traffic Control Zones (Work Zones) Using Changeable Message Signs with Radar. Transportation Research Record, Vol. 1509, 1995, pp. 73–81. 25. McCOY, P. T., J. A. Bonneson, and J. A. Kollbaum. Speed Reduction Effects of Speed Monitoring Displays with Radar in Work Zones on Interstate Highways. Transportation Research Record, 1996, p. 8. 139 26. Pesti, G., and P. T. McCoy. Long-Term Effectiveness of Speed Monitoring Displays in Work Zones on Rural Interstate Highways. Transportation Research Record: Journal of the Transportation Research Board, Vol. 1754, No. 1, 2001, pp. 21–30. https://doi.org/10.3141/1754-03. 27. Thompson, B. Evaluation of Radar Activated Changeable Message Sign for Work Zone Speed Control. Maine. Dept. of Transportation, 2002. 28. Wang, C., K. K. Dixon, and D. Jared. Evaluating Speed-Reduction Strategies for Highway Work Zones. Transportation Research Record: Journal of the Transportation Research Board, Vol. 1824, No. 1, 2003, pp. 44–53. https://doi.org/10.3141/1824-06. 29. Sorrell, M. T. Use of Radar-Equipped Portable Changeable Message Sign to Reduce Vehicle Speed in South Carolina Work Zones. 2007, p. 16. 30. Mattox III, J. H., W. A. Sarasua, J. H. Ogle, R. T. Eckenrode, and A. Dunning. Development and Evaluation of Speed-Activated Sign to Reduce Speeds in Work Zones. Transportation research record, Vol. 2015, No. 1, 2007, pp. 3–11. 31. Medina, J. C., R. F. Benekohal, A. Hajbabaie, M.-H. Wang, and M. V. Chitturi. Downstream Effects of Speed Photo–Radar Enforcement and Other Speed Reduction Treatments on Work Zones. Transportation Research Record: Journal of the Transportation Research Board, Vol. 2107, No. 1, 2009, pp. 24–33. https://doi.org/10.3141/2107-03. 32. Hajbabaie, A., J. C. Medina, M.-H. Wang, R. (Ray) F. Benekohal, and M. Chitturi. Sustained and Halo Effects of Various Speed Reduction Treatments in Highway Work Zones. Transportation Research Record, Vol. 2265, No. 1, 2011, pp. 118–128. https://doi.org/10.3141/2265-13. 33. Roberts, C. A., and E. J. Smaglik. Driver Feedback on Monetary Penalty and Its Impact on Work Zone Speed. Transportation Research Record: Journal of the Transportation Research Board, Vol. 2272, No. 1, 2012, pp. 27–34. https://doi.org/10.3141/2272-04. 34. Thapa, R., S. Hallmark, O. Smadi, and A. Goswamy. Assessing Driving Behavior Upstream of Work Zones by Detecting Response Points in Speed Profile: A Naturalistic Driving Study. Traffic Injury Prevention, Vol. 20, No. 8, 2019, pp. 854–859. https://doi.org/10.1080/15389588.2019.1663348. 35. Huang, Y., and Y. Bai. Effectiveness of Graphic-Aided Portable Changeable Message Signs in Reducing Vehicle Speeds in Highway Work Zones. Transportation Research Part C: Emerging Technologies, Vol. 48, 2014, pp. 311–321. https://doi.org/10.1016/j.trc.2014.09.007. 36. Huang, Y., and Y. Bai. Driver Responses to Graphic-Aided Portable Changeable Message Signs in Highway Work Zones. Journal of Transportation Safety & Security, Vol. 11, No. 6, 2019, pp. 661–682. https://doi.org/10.1080/19439962.2018.1463336. 140 37. Gambatese, J., and A. Jafarnejad. Evaluation of Radar Speed Display for Mobile Maintenance Operations. Publication OR-RD-16-09. Oregon Department of Transportation, Oregon, 2015. 38. Zhang, F., and J. A. Gambatese. Highway Construction Work-Zone Safety: Effectiveness of Traffic-Control Devices. Practice Periodical on Structural Design and Construction, Vol. 22, No. 4, 2017, p. 04017010. https://doi.org/10.1061/(ASCE)SC.1943-5576.0000327. 39. Anderson, S. M., J. R. Cunningham IV, J. Liang, and E. J. Fitzsimmons. Evaluation of the Effectiveness of Dynamic Speed Feed Back Signs in Work Zones on High-Speed Kansas Roadways. Presented at the 100th Transportation Research Board Annual Meeting, Washington, D.C, 2021. 40. Tribbett, L., P. McGowen, and J. Mounce. An Evaluation of Dynamic Curve Warning Systems in the Sacramento River Canyon. Western Transportation Institute, Montana State University Bozeman, 2000. 41. Ullman, G. L., and E. R. Rose. Evaluation of Dynamic Speed Display Signs. Transportation research record, Vol. 1918, No. 1, 2005, pp. 92–97. 42. Bertini, R. L., C. M. Monsere, C. Nolan, P. G. Bosa, and T. A. El-Seoud. Field Evaluation of the Myrtle Creek Advanced Curve Warning System. Oregon. Dept. of Transportation. Research Unit, 2006. 43. Shauna L. Hallmark, Neal Hawkins, and Omar Smadi. Evaluation of Dynamic Speed Feedback Signs On Curves: A National Demonstration Project. Publication FHWA-HRT- 14-020. Center for Transportation Research and Education, Iowa State University, 2015, p. 210. 44. Montella, A., F. Galante, F. Mauriello, and L. Pariota. Low-Cost Measures for Reducing Speeds at Curves on Two-Lane Rural Highways. Transportation Research Record: Journal of the Transportation Research Board, Vol. 2472, No. 1, 2015, pp. 142–154. https://doi.org/10.3141/2472-17. 45. Mahmud, M. S., A. Bamney, M. U. Megat Johari, H. Jashami, T. J. Gates, and P. T. Savolainen. Evaluating Driver Response to a Dynamic Speed Feedback Sign at Rural Highway Curves. Transportation Research Record, 2022, p. 03611981221112401. https://doi.org/10.1177/03611981221112401. 46. Sandberg, W., T. Schoenecker, K. Sebastian, and D. Soler. Long-Term Effectiveness of Dynamic Speed Monitoring Displays (DSMD) for Speed Management at Speed Limit Transitions. 2006, p. 14. 47. Cruzado, I., and E. T. Donnell. Evaluating Effectiveness of Dynamic Speed Display Signs in Transition Zones of Two-Lane, Rural Highways in Pennsylvania. Transportation Research Record, Vol. 2122, No. 1, 2009, pp. 1–8. https://doi.org/10.3141/2122-01. 141 48. Hallmark, S. L., N. Hawkins, and S. Knickerbocker. Use of DSFS as a Speed Transition Zone Countermeasure in Small, Rural Communities. Presented at the 2015 IEEE 18th International Conference on Intelligent Transportation Systems, 2015. 49. Mahmud, M. S., M.-U. Megat-Johari, A. Bamney, H. Jashami, T. J. Gates, and P. T. Savolainen. Driver Response to a Dynamic Speed Feedback Sign at Speed Transition Zones Along High-Speed Rural Highways. Transportation Research Record, 2022. 50. Walter, L., and J. Broughton. Effectiveness of Speed Indicator Devices: An Observational Study in South London. Accident Analysis & Prevention, Vol. 43, No. 4, 2011, pp. 1355– 1358. https://doi.org/10.1016/j.aap.2011.02.008. 51. Ardeshiri, A., and M. Jeihani. A Speed Limit Compliance Model for Dynamic Speed Display Sign. Journal of Safety Research, Vol. 51, 2014, pp. 33–40. https://doi.org/10.1016/j.jsr.2014.08.001. 52. Karimpour, A., R. Kluger, and Y.-J. Wu. Traffic Sensor Data-Based Assessment of Speed Feedback Signs. Journal of Transportation Safety & Security, 2020, pp. 1–24. 53. Karimpour, A., R. Kluger, C. Liu, and Y.-J. Wu. Effects of Speed Feedback Signs and Law Enforcement on Driver Speed. Transportation Research Part F: Traffic Psychology and Behaviour, Vol. 77, 2021, pp. 55–72. https://doi.org/10.1016/j.trf.2020.11.011. 54. Ash, K. G. Increasing Speed Limit Compliance in Reduced-Speed School Zones. Brigham Young University - Provo, 2006. 55. O’Brien, S. W., and C. L. Simpson. Use of “Your Speed” Changeable Message Signs in School Zones: Experience from North Carolina Safe Routes to School Program. Transportation Research Record, Vol. 2318, No. 1, 2012, pp. 128–136. https://doi.org/10.3141/2318-15. 56. Williamson, M. R., R. Fries, and H. Zhou. Long-Term Effectiveness of Radar Speed Display Signs in a University Environment. Journal of Transportation Technologies, Vol. 06, No. 03, 2016, pp. 99–105. https://doi.org/10.4236/jtts.2016.63009. 57. Bloch, S. A. Comparative Study of Speed Reduction Effects of Photo-Radar and Speed Display Boards. Transportation Research Record: Journal of the Transportation Research Board, Vol. 1640, No. 1, 1998, pp. 27–36. https://doi.org/10.3141/1640-05. 58. Chang, K. N., M. Nolan, and N. L. Nihan. Measuring Neighborhood Traffic Safety Benefits by Using Real-Time Driver Feedback Technology. Transportation Research Record, Vol. 1922, No. 1, 2005, pp. 44–51. https://doi.org/10.1177/0361198105192200107. 59. Gehlert, T., C. Schulze, and B. Schlag. Evaluation of Different Types of Dynamic Speed Display Signs. Transportation Research Part F: Traffic Psychology and Behaviour, Vol. 15, No. 6, 2012, pp. 667–675. https://doi.org/10.1016/j.trf.2012.07.004. 142 60. Churchill, A. E., and S. Mishra. Speed Feedback Signs as a Tool to Manage Demand for Lower Residential Speeds. Presented at the TAC 2016: Efficient Transportation - Managing the Demand - 2016 Conf. and Exhibition of the Transportation Assoc. of Canada, 2016. 61. Jue, M. J., and J. T. Jarzab. Long-Term Effectiveness of Radar Speed Feedback Signs for Speed Management. Institute of Transportation Engineers. ITE Journal; Washington, Vol. 90, No. 5, 2020, pp. 40–44. 62. Wu, M., K. El-Basyouny, and T. J. Kwon. Before-and-After Empirical Bayes Evaluation of Citywide Installation of Driver Feedback Signs. Transportation Research Record: Journal of the Transportation Research Board, Vol. 2674, No. 4, 2020, pp. 419–427. https://doi.org/10.1177/0361198120912243. 63. Jeihani, M., A. Ardeshiri, and A. Naeeni. Evaluating the Effectiveness of Dynamic Speed Display Signs. National Transportation Center, Morgan State University, 2012. 64. Santiago-Chaparro, K. R., M. Chitturi, A. Bill, and D. A. Noyce. Spatial Effectiveness of Speed Feedback Signs. Transportation research record, Vol. 2281, No. 1, 2012, pp. 8–15. 65. Zhao, X., L. Huang, W. Guan, and J. Rong. Evaluation of Effects of Warning Sign Position on Driving Behavior in Horizontal Sharp Curves. Advances in Mechanical Engineering, Vol. 7, No. 2, 2015, p. 971290. https://doi.org/10.1155/2014/971290. 66. TraffiCalm Systems. SafetyCalm Software Operating Instructions. TraffiCalm Systems, 2013. 67. All Traffic Solutions. TraffiCloud User Manual. All Traffic Solutions, 2016. 68. Gates, T. J., M. S. Mahmud, A. J. Ingle, M. Motz, T. Holpuch, and P. T. Savolainen. Evaluation of Alternative Messages and Sign Locations on Driver Response to a Dynamic Speed Feedback Sign on a Freeway Interchange Ramp. Transportation Research Record: Journal of the Transportation Research Board, Vol. 2674, No. 12, 2020, pp. 530–541. https://doi.org/10.1177/0361198120959076. 69. Mahmud, M. S., M. Motz, T. Holpuch, J. Hankin, A. J. Ingle, T. J. Gates, and P. T. Savolainen. Driver Response to a Dynamic Speed Feedback Sign on Freeway Exit Ramps Based on Sign Location, Interchange Type, and Time of Day. Transportation Research Record: Journal of the Transportation Research Board, 2021, p. 036119812110152. https://doi.org/10.1177/03611981211015250. 70. Mahmud, M. S., T. J. Gates, B. Safaei, and P. T. Savolainen. Driver Response to a Dynamic Speed Feedback Sign at a Freeway Exit Ramp Considering the Sign Design and Installation Characteristics. Transportation Research Record, 2022. 71. Gates, T. J., M. S. Mahmud, P. T. Savolainen, D. Zhao, A. Zockaie, and M. Ghamami. Evaluation of Dynamic Speed Feedback Signs on Freeway Interchange Ramps. Publication SPR-1704. Michigan Department of Transportation, Lansing, Michigan, 2022. 143