I NJ'M "" 13‘ I .L AT- “T" l' ..... I ‘1‘ 'VLLH: 1LTKL! wwwhflfiWLLH’L #. “1,21,“...“1 L 1111 |, 1%;11511‘1Lk'f 2‘ .. 31L LLB: ' IL I *5 ' |. :1m 5: . . .. ., LL LL: .531 I“ 11:1 W "11 Ii-‘L‘ 111: M L .. , . 11-; “HI j|r||'|L '1 LL.|1 |. '.',. 4,1 ;, 1 13111 Wyn £51 1 1111_|| .| _. F ..': vTL‘1' ‘ 1' u L '1‘1’11‘ 111L‘ "0‘ “12054? 311 ‘1" -‘t"- - ' 11111.11111E‘1111'F11‘F £14". ”11; '11 1L1111111 I111"? ':|I111 11:11]:|'.|| '.|".v‘11 511m}.£?121:151:i -;E 1.1%]; - . 1 I; I am p—J \a A .- "5171'- - .7: :4.- .. A .4_ .. A WM I I “11.. I111}. 114 111-. 11’9” | L . LLLLLLLLL LL :"L‘ ‘ “1 L11. :1 11 *1 1 .g1dkfiflfiflfi11mggwnmu|th L1flqsl ”1'111V11E'- ‘ | - ififi‘x ; 511 511143 ‘IIL'1."|L‘ 1"”L|L |111|1m1|1 51;“ : . 1 | | L 31311111111111" L1 L LLL ~ L: L-TLéI‘h‘iLI?” |3~11'1£L‘1;1L‘1""1. . 5% | 11:1 |1|’ 1‘1}?11113111111é11z1115‘1111 i W— ‘3. I :1||11J5l|11.11 ‘ ‘ |;_M’¢§'1 .1233" C " “L I11 15.111; 121.1111 LEW”? (h - a; |‘,‘ L 9‘. m"” +-L ' ‘ L . ' 1' .‘|’:’ 1: 3-“ |is I ‘1'91< ‘1 'l- '13.“? H ‘1 ,| I ' ‘1 '- ‘18-: bkpfigt I. .|' .""QR=WL3 " 'L ’ a": "a: a -' 33:7.‘13‘12‘L‘i-EJEL, L (me L «m ELL i 11.111111 1:1 . . ,ul. { “:31 ‘ \ 5‘ ' i ' y -m.- V A ”.2! 5E:E£% 4 m .- u o l nn—uv—-o 'mmu—u ' '1 . I .‘ u 1 L IJ EL Mk» I. ‘I‘ L if! :;~1:.'":L1L::U*1 I L@m%WL#fifl$flflwfi ~ WL$n1W4L1W1ww~ L . :1 LLL|LEL;L'L1 13'1'L;1."1.1511|I“I.”1 ' . ~ L ' “'1'11:11U1011:5’6 II1111‘1HILI’111 'H 5 in? . .'1111351153111’51’1111"1LE?1*‘=‘111111’151L111J1“:["53“ ”i: ‘ i LLLLLLLLLLL'LLL i 1?" LLLLL’LLL ” . 1LLILLLLL'LLLLLLLLLLL- 1‘ -_ :1": L ‘ .-|_‘L|||..“'~3,fi‘l-I ,I‘I .‘ ' " '. 1Sfi¥§111QhHPXU1I . LLLMWMLW . 113‘; 1313".“ 31:: 1*:‘5 .11- ' . L‘|||:|r|1ggt|hLiLI'L'LL 7. 11,1 E9151||| W| L 1?,“ '1111’1 1.3-7" -‘ - 2?. Ytn1¥ ‘i‘fi: 3% ”131550" 3 ' .3: d“ . . 3:17] .gIN 3“”:1” ”‘ x“ 3 ': a‘p' .5' i :71; .rtgflj 5m: 1115:5311?! 1351-? 11.133111 id'hhi‘zzs’flg‘g; é; 'v'|:::,|i£-‘ I 9'73; :51“ “"21"!“ HAL" JUL“ 1Q)?!“ PIKE»??? “03" I F.” 4 $ 33"‘33‘1 ‘LHL :2" WHILE“; 5115:1551 55%? 61?"; if“ ‘9me | I 9...:{331f‘w'1}; 1d "111:9; (1r L L .L L L L L, ~ ,LLLLLL «a: «L . .TLLIE li’L}d{&lfik‘ 3.11; “K1515. ‘. ..; -'." u-i" 73%|? :1:|:|LTLE.’| '|'-| ‘9' a , II ‘j‘igl' LEI WM”LWLL’ ”’ka$%“” LH I 31%}: “.4“. I"|J§:‘. I Q-LL‘ 3"..Q, 4;:5‘5-‘3' V-f - "'P;.i' <..a pi" ' ~ ath LLLLLnL; L1. 11: LLLL LL; L,. :lciji " : . . . x. rfiJa-‘mrmr‘a 1 : 3. o 1 \ 1 i; I. I I: ‘35:“ F: 1;:‘1‘151 .. LI‘.‘I:!:‘\‘I')'5 ‘ i111; ‘1 1511' 1" h! :1 ‘1’";qu :3" “L; A“): 11.;' 1;!” “real I I ,, L .ELJIL'E LILLL'LL' ‘: 1“!le ILL'LLIJLLI 'L 1"1*'L'111"1J"1 ’ I)“: My LIBRARY Michigan State University This is to certify that the dissertation entitled EVALUATION OF RAMP METERING STRATEGIES AT LOCAL ON-RAMPS AND FREEWAY-TO-FREEWAY INTERCHANGES USING COMPUTER SIMULATION MODELLING APPROACH presented by Abdul-Rahman Ibrahim Hamad has been accepted towards fulfillment of the requirements for Major professor Date MUV /3/ /yfl MSU is an Affirmative Action/Equal Opportunity Institution 0-12771 mnmmmmmnmm 2301063 2101 RETURNING MATERIALS: PVIESI_J Place in book drop to LIBRARJES remove this checkout from w your record. FINES will be charged if book is returned after the date stamped below. EVAUEKTION OF RAMP METERING STRATEGIES AI LOCAL.ON-RAMPS AND FREEWAY-TO-FREEHAY INTERCHANCES USING COMPUTER SIMULATION MODELLING APPROACH BY Abdul-Rahman Ibrahim Hamad A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Civil and Environmental Engineering 1987 sy De 8 we E? <'_) 5—) 0;: C’) 1' ,~. ABSTRACT EVALUATION OF RAMP METERING STRATEGIES AT LOCAL ON-RAMPS AND FREEWAY-TO-FREEWAY INTERCHANGES USING COMPUTER SIMUIATION MODELLING APPROACH By Abdul-Rahman Ibrahim Hamad Ramp metering is a strategy of freeway operations designed to improve the flow of freeway traffic by controlling the rate at which additional vehicles are allowed into the traffic stream. The primary goal of ramp metering is the efficient use of the highway system. Many large urban centers have installed freeway ramp metering systems to help reduce the congestion on their urban freeways (e.g., Detroit, Los Angeles, Chicago, and Houston). The problem is that these systems do not show the expected benefits when there are freeway-to-freeway interchanges in the urban freeway system. This is because these interchanges are not metered and the large volumes travelling between the freeways tend to interrupt the smooth flow that is supposed to be achieved from the ramp metering strategy. This study utilized the Integrated Traffic Simulation (INTRAS) model, which is a microscopic freeway simulation model, to defixua the optimal strategy for metering flow onto the freeway and to evaluate the benefits of such strategy. The evaluation was conducted on the portion of the Ford Freeway (I-94) within the Detroit city limits. vol‘ mod. ten achi free , Abdul-Rahman Ibrahim Hamad Field data including volume, speed, vehicle mix, and volume/capacity ratio were used to calibrate and validate the INTRAS model. The results of the study indicated that significant benefits, in terms of reduced delay and increased speed on the freeway, can'be achieved by introducing ramp metering to both local on-ramps and freeway-to-freeway interchanges. His men: COKE of a and C035 fer data ACKNOWLEDGMENTS All praise and thanks are due to Allah, Lord of the Universe, for His merciful divine direction throughout my study. I am indebted.to all.those persons who helped and encouraged me in the conducting and completion of this dissertation. Sincere ap- ‘preciation and gratitude to Dr. Thomas Maleck, my advisor and committee chairman, for his valuable time, assistance, and encourage- ment. His understanding and consideration have been incentive for the completion of this dissertation. My appreciation and gratitude are extended to the other members of my guidance committee, Professor William Taylor, Dr. Richard Lyles, and Professor Roy Erickson, for their contributions, advice, and constructive comments to the study. Finallgr, thanks are due to Michigan Department of Transportation for supporting this research and to all studentS‘who assisted in.the data-collection phase of this study. ii mm: or CONTENTS ' LIST OF TABLES LIST OF FIGURES Chapter 1. INTRODUCTION AND STATEMENT OF PROBLEM 1.1 INTRODUCTION 1.2 STATEMENT OF PROBLEM LITERATURE REVIEW 2.1 CONTROL STRATEGIES FOR FREEWAY OPERATIONS 2.2 RAMP METERING SYSTEMS 2.2.1 Detroit 2 Houston Chicago 3 4 Los Angeles 5 San Diego 2.2. 2.2. 2.2. 2.2. 2.3 COMPUTER SIMULATION MODELS 2.4 SUMMARY 2.5 OBJECTIVES OF THE STUDY THE INTRAS MODEL 3.1 GENERAL 3.2 PROGRAM PURPOSE AND CAPABILITIES 3.3 NETWORK IDEALIZATION AND MODELLING CONCEPTS 3.3.1 Network Presentation 2 Geometric Features Traffic Flow Patterns 3.3. 3.3.3 3.3.4 Signal and Sign Control iii Page vi viii 25 27 28 28 29 29 29 36 38 39 4.! .011 5.1 5.2 5.3 5.1. 5.5 - API 6.1 6.2 3.3.5 Traffic Descriptive Features 3.3.6 Freeway Traffic Responsive Control 3.4 SIMULATION AND PROGRAMING METHOD 3.5 ASSUMPTIONS AND LIMITATIONS 3.6 SIMULATION DEVELOPMENT 3.6.1 The Car Following Model 3.6.2 Lane Changing Process 3.6.3 Vehicle Generation METHODOLOGY 4.1 IMPLEMENTATION OF INTRAS 4.2 STUDY AREA 4.3 DATA COLLECTION 4.3.1 Data Elements 4.3.2 Field Data Collection Procedures 4.4 DATA REDUCTION 4.4.1 Building The Grid 4.4.2 Sample Size 4.4.3 Data Collection 4.4.4 Calculating The Speeds . CALIBRATION AND VALIDATION 5.1 THE CALIBRATION DATA 5.2 CHOOSING THE APPROPRIATE VARIABLES 5.3 THE CAR FOLLOWING MODEL 5.4 CALIBRATION PROCEDURE 5.4.1 The Computer Runs 5.4.2 The Significance Level Test 5.5 THE VALIDATION APPLYING THE CONTROL STRATEGIES 6.1 PROBLEMS WITH INTRAS USER'S MANUAL 6.2 EVALUATING THE PRESENT CONTROL STRATEGY iv Page 40 41 43 44 46 46 47 51 53 53 54 56 56 59 63 63 64 67 69 78 78 81 84 87 87 87 92 96 96 97 e basic run (no-metering) plying the present control strategy scussion of results asoum haNHu misin—I U> H1353 E T NG NEW STRATEGIES AT LOCAL ON-RAMPS 6.3.1 Applying the Strategies 6 3 2 Discussion of Results 6.4 TESTING NEW STRATEGIES AT LOCAL AND FREEWAY ON-RAMPS 6.4.1 Applying the Strategies w/ Existing Geometry 6.4.2 Applying the Strategies w/ Modified Geometry 6.4.3 Discussion of results 7. SUMMARY AND CONCLUSIONS 7.1 SUMMARY 7.2 CONCLUSIONS APPENDICES A. NETWORK DATA AND NODE-LINK DIAGRAM B. EXAMPLES OF INTRAS OUTPUTS REFERENCES Page 98 98 98 100 100 104 106 106 107 111 115 115 117 107 118 162 189 H) 1‘) Table 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. LIST OF TABLES Title Summary of Ramp Control Cases Summary of Freeway Simulation Models INTRAS Geometric Limits Selected Locations: Date, Time, and Duration of Data Collection Input Data Required for INTRAS Calculations of Standard Deviation Mt. Elliot Observed Data File Data Reduction Fortran Program Mt. Elliot Output Data File Summary of Average Speeds and Volumes Sensitivity Factor Values Observed and Model Speeds for Mt. Elliot Paired Comparison of Selected Parameters The Validation Results Comparing Measures of Effectiveness For One Peak Hour No-Metering vs. Present Strategy Comparing MOE's of Different Metering Strategies At Local On—Ramps Only (All Network) Comparing MOE's of Different Metering Strategies At Local On-Ramps Only (Freeway Only) Comparing MOE's of Different Metering Strategies At Local On-Ramps Only (Ramp and Surface Links) Comparing MOE's of Freeway-To-Freeway Control Strategies (All Network) vi Page 18 26 45 58 6O 66 7O 71 72 73 86 88 91 95 99 101 102 103 108 Tabll 20. 21. Table Page 20. Comparing MOE's of Freeway-To-Freeway Control Strategies (Freeway Only) 109 21. Comparing MOE's of Freeway-To-Freeway Control Strategies (Ramp and Surface Links) 110 vii 10. ll. 12. 14. 19, LIST OF FIGURES Figure Title 1. Sample Physical Freeway—Frontage Road Network 2. Representation of Sample Network 3. Typical Freeway Link Configuration 4. Typical Metered Ramp Geometry 5. Platoon Behavior: PITT Algorithm-One Second Interval 6. Platoon Behavior: PITT Algorithm-Three Second Interval 7. Lane Changing Vehicles 8. Study Area Boundaries 9. Selected Locations for Sampling 10. Data Elements: Sources and Flow Through the Study 11. Placement of Grid on the Screen 12. Choosing the Sample Size 13. Speed-Volume Plots for Sampled Locations 14. Mt. Elliot Sub-network 15. Speed-Volume Curve for Mt. Elliot 16. Observed and Model Speeds for Mt. Elliot 17. The Confidence Interval Test 18. Comparing Average Speeds of Different Metering Strategies at Local On-Ramps Only 19. Comparing Average Speeds of Different Metering Strategies at Local On-Ramps and Freeway Interchanges 20. The Network Link-Node Diagram viii Page 31 32 34 42 48 49 50 55 57 61 65 68 75 79 80 90 93 105 112 120 th pe in th St ES 8X1 1110‘ am CHAPTER 1 INTRODUCTION AND STATEMENT OF PROBLEM 1.1 INTRODUCTION Urban freeways are expected to move large volumes of traffic throughout the day, but particularly so during the peak traffic volume periods which may occur two or more times a day at some sites. Urban freeways usually operate satisfactorily during the early years after they have been opened to public traffic. Often, though, in the later stages of their design life, the operation of urban freeways, especially in freeway and ramp merging areas, deteriorate to such an extent that they become highly congested, unstable, and ineffective in moving high volumes of traffic at the very time the demand is heaviest and the need the greatest. Ramp metering is a strategy of freeway operations designed to improve the flow of freeway traffic by controlling the rate at which additional vehicles are allowed into the traffic stream. The primary goal of ramp metering is the efficient use of the highway system. The freeway on-ramp is the interconnecting roadway between the freeway and the adjoining highway or street that provides vehicle access to the freeway. Freeway ramp control systems are used to control the flow of vehicles onto the freeway and, thereby, maintain freeway operations at an acceptable service level. 1 ona sine: this ramp used 1 adver traffi the a ObstrL consid Contro locate: 1ane, ramp Ve green Until t 550501-1 the Che: Yellow subseque Ramp control systems can be implemented on individual on-ramps or on a sequence of on-ramps. Different types of systems have been used since the early 19603, including: 1. ramp closure, 2. pre-timed or fixed-time control, 3. gap-acceptance control, and 4. traffic responsive or real-time control. The most rudimentary ramp control system is a ramp closure. For this type of control, vehicle access to the freeway for a given on- ramp is prohibited during the peak periods. This type of control is used where a downstream one-way constriction, called a bottleneck, adversely restricts the free movement and flow of traffic because the traffic demand exceeds the available freeway capacity. By eliminating the additional on-ramp flow via a ramp closure, the traffic obstruction at the bottleneck may be prevented. Ramp closure is considered by many traffic engineers to be the least desirable type of control. Pre-timed ramp control systems utilize one or two traffic signals located on the ramp upstream of the beginning of the acceleration lane. For many applications the traffic signal rests in red until a ramp vehicle arrives at the traffic signal, at which time it is turned green to allow vehicle passage. The traffic signal remains green until the vehicle is detected by an inductive loop, called a check-out sensor, which is located just downstream of the traffic signal. When the check-out sensor detects a vehicle the traffic signal is turned yellow for a short period of time until it again displays red. If a subsequent vehicle is waiting at the traffic signal, it is detained for a pre-timed interval to allow separation between ramp vehicle releases. In this way pre-timed ramp control limits or meters vehicle access to the freeway. At some locations, instead of single vehicle metering, two or more vehicles are permitted access to the freeway when the traffic signal is turned green. In gap-acceptance ramp control, the release of ramp vehicles from the ramp-side traffic signal is coordinated so that both acceptable gaps on the freeway and ramp vehicles arrive at the merge area at the same time. An acceptable freeway gap is an opening in the right lane freeway traffic that exceeds a predefined time separation below which ramp drivers are not able or willing to make a merge. Drew (1967) defines an acceptable gap as "one equal to or larger than the critical gap," where the critical gap is "that gap for which an equal percentage of ramp traffic will accept a smaller gap as will reject a larger one." The gap-acceptance type of ramp control differs from the pre- timed control in that with pre-timed control the release of ramp vehicle is not coordinated in any way with the acceptable freeway gaps. For gap-acceptance control two freeway inductive sensors are used by a mini-computer to determine the size of the freeway gap and the speed at which the gap is traveling towards the merge area. The actual release time of the ramp vehicles is computed so that both the acceptable gap and the ramp vehicle arrive at the merge area simultaneously. Gap-acceptance control systems also employ a maximum waiting time from the moment a ramp vehicle activates the check-in sensor. If the computer does not find an acceptable gap within that time same \ sectic rate condit is ap; to pre systen flow 6 condit For air the to capaCIt time period, the vehicle is released from the traffic signal in the same way as in pre-timed ramp control. In real-time ramp control systems, traffic is monitored along a section of the freeway for the purpose of adapting the ramp metering rate or flow in accordance with the existing freeway traffic conditions. With a traffic responsive system, when the freeway flow is approaching downstream capacity, the flow from the ramp is reduced to prevent a breakdown in the freeway flow. A traffic responsive system also permits an increase in the ramp flow whenever the freeway flow decreases. Typically, the monitoring of freeway traffic conditions is a function of either a volume or an occupancy measure. For either measure, the ramp traffic flow is based upon maintaining the total demand to a value equal to or less than the downstream capacity in order to maintain a given service level. 1.2 PROBLDI STATEMENT Many large urban centers have installed freeway ramp metering systems to help reduce the congestion on their urban freeways (e.g., Detroit, Los Angeles, Chicago, and Houston). The problem is that these systems do not show the expected benefits when there are freeway-to-freeway interchanges in the urban freeway system. This is the case in the City of Detroit. In Detroit there are three major freeway-to-freeway interchanges along the Ford Freeway. The interchange ramps are not metered, and the large volumes travelling between the freeways tend to interrupt the smooth flow that is supposed to be achieved by the ramp metering strategy. The interchanges are about one mile apart, which means that the 1 traff Free1 on-ra diffe freel those spee inter 61) 1 Linw: imprc SCale the weaving areas between the interchanges are very limited and many traffic conflicts are expected to occur. In addition, the Lodge Freeway interchange, which is located in the middle, has a left-hand on-ramp and off—ramp interchange beside the regular right-hand ramps. Field data collected along the Ford Freeway showed large differences in the average speed of traffic between the parts of the freeway around the interchange areas and the parts that are outside those areas when the ramps were metered. For example, the average speed at the Van Dyke on-ramp, which is located outside the interchange areas, increased after ramp metering by about 15% (53 to 61) during the morning peak hour, while the average speed at the Linwood on-ramp, inside the interchange areas, did not show any improvement. The research discussed here was addressed to an examination of the ramp metering operation of the Surveillance Control and Driver Information (SCANDI) system in Detroit. The objectives were to determine the effects of ramp metering on the Ford Freeway and adjacent surface streets (that is defining the queue lengths behind the metering signal and their spill-back onto the surface streets), and to evaluate new strategies that can be used to increase the benefits of the system. Of special concern was an evaluation of the strategy of metering part or all of the three major freeway-to-freeway interchanges along the Ford.Freeway (I-94) within the Detroit city limits. The three freeways that intersect with the Ford Freeway are the Jeffries (I-96), the Lodge (US-10), and the Chrysler (I-75). One of the most important tools used to evaluate potential large- scale changes in traffic systems is simulation modelling. If a traffic system is represented on a computer by means of a simulation model, it is possible to predict the effects of traffic control and traffic management strategies on the system's operational performance. The use of simulation allows the testing of metering strategies other than the one in operation to determine if those strategies can be more effective than the one now in use. For this dissertation the Integrated Traffic Simulation (INTRAS) model (Wicks and Lieberman, 1977) was used to evaluate the ramp metering system in Detroit. prl frc bei a9? 2.1 fre: CHAPTERZ LITERATURE REVIEW There are several areas that need to be discussed before presenting the research itself. These include: control strategies for freeway operations, characteristics of ramp metering, and the reasons behind using simulation computer models instead of empirical approaches. 2.1 CONTROL STRATEGIES FOR FREEWAY OPERATIONS Travel demand continues to increase, especially on urban freeways, which causes the congestion on those facilities to increase. Past studies (e.g., Wattleworth, et al. 1967, and Newman, et al. 1970) have demonstrated that this increase in congestion can be slowed by exercising some type of traffic control strategy. Limiting access to a highway usually results in improved operations and safety for many motorists at a cost to a few motorists. This is part of the reasoning behind building limited access highways. Typically, the more limited the access, the better the level of service offered to the user. Therefore, the operation of roads that are already access limited, such as freeways and expressways, can be further improved by further regulating the access points. This is called ramp control, and has been the topic of many studies in the past I and h develc of the that, freew; traff Closi1 freew confi closu- and D past twenty-five years (e.g., Wattleworth, et a1. 1967, Gervais 1964, and Newman, et a1. 1970). This review chronicles the major developments in ramp control strategies. Ramp closures were investigated prior to ramp metering, because of their relative simplicity. The reasoning behind ramp closure was that, if higher volumes of traffic could leave the center-city by freeway in a given period of time, there would be a less of backlog of traffic in the central business area in the evening peak periods. Closing selected on-ramps would reduce the volume and density on the freeway, thereby increasing speed, as well as eliminating traffic conflicts at merge areas. Many cities experimented with on-ramp closures, with favorable results [e.g., Houston (Pinnel, et al. 1965), and Detroit (Gervais, 1964)]. In the mid 19603, ramp metering began to replace ramp closures as a means to control freeway volumes. The first meters (May, 1964) were fixed-time meters, releasing cars at constant intervals. Many early experiments with metering (Gervais, 1964) employed a policeman with a clock controlling the rate of access of vehicles from ramps. The policemen were eventually replaced by signals and new metering strategies were employed. Demand-capacity metering and gap-acceptance metering were two of the new strategies (both were discussed earlier in the introduction). Both relied on loop detectors to gather information on the freeway flow. The major drawback of gap-acceptance metering is that gaps are not stable, and may disappear after they are identified. Munjal, et al. (1973) compared fixed-time and gap-acceptance metering on the Long Island Expressway in New York City in 1969. They found that while gap-acceptance metering is only slightly more accurate than fixed-time metering in finding acceptable gaps, that when the gaps are found, the merge is much smoother. It was also suggested that prohibiting lane changing into the outside lane between detector location and ramp location would cut the gap-acceptance failure rate roughly in half. Another way of approaching the gap instability problem was the use of a pacer system. Tignor (1975) tested such systems in Boston in 1970. Typically, the system consisted of a band of lights which represent a gap in freeway traffic. The band travels along a guardrail type signal, changing length and speed as does the gap. The pacer system consisted of a series of green lights similar to conventional traffic lights which flashed on ahead of the driver to lead him into the gap. The system used seven sets of loop detectors in the right lane of the freeway to detect gaps. Public response to the system was good. 2.2 RAMP METERING SYSTEMS Chicago, Detroit, and Los Angeles each have a centralized traffic control center from which they receive data from television and electronic sensors, provide incident detection and motorist aid services, and control ramp metering. In addition, other cities use computers to optimize a series of metered ramps along a freeway corridor in order to minimize the travel time on the corridor. Five case studies, the Lodge Freeway in Detroit, Michigan, the Gulf Freeway in Houston, Texas, the Dan Ryan Expressway in Chicago, Illinois, the Harbor and Hollywood Freeways in Los Angeles, Cali one o inclt early the e Freew; they in 19 Close. times, EXper; and 11 ShOUId Choic. clOSUrl Surpri, °f 510i 0n the in the Pu ism no the 10 California, and the Inland and Helix Freeways in San Diego, California, are presented in the following sections. 2.2.1 Detroit The John C. Lodge Freeway surveillance project in Detroit had as one of its objectives the development of a traffic control system, including ramp metering. Two experiments were carried out in the early 19603 concerning ramp closure (Gervais, 1964). The first tested the effectiveness of "don't enter" signs above a ramp on the Lodge Freeway. This study showed that while motorists read these signs, they did not always obey them. For the second experiment, conducted in 1963, nine entrance ramps on a three mile study section were closed, either individually or in various groups, during varying times, although always during peak flows. Data on the effect of this experiment were collected through TV surveillance, roadway sensors, and license plate surveys. Although no specific conclusions were made regarding which ramps should be closed or for how long, it was determined that an effective choice of ramp closures will improve freeway flow. During ramp closure times, the freeway volume and average speed increased, as did, surprisingly, lane changes. The latter may suggest a more fluid state of flow. The number and severity of traffic stoppages dropped. Flow on the surface streets was not analyzed as thoroughly, but no change in the level of service was observed. Public response to the experiment was largely favorable, although when no barriers or police enforcement was present at a closed ramp, the violation rate was about 30-35%. After all the ramps were reope study trafi Freer syste inter: ramps by mar to me for a destin 11 reopened, surface street volumes remained higher than before the study, suggesting that many drivers learned from the experience. Wattleworth, et al. (1967) studied the effect of ramp metering on traffic flow when the first ramp meters were installed on the Lodge Freeway in July, 1967. An inventory of the freeway and surface street system and capacities was taken. Traffic was counted to determine intersection capacities, and loop detectors were installed on some ramps to determine demand. For other ramps, volumes were inventoried by manual counting or aerial photography. Sonic detectors were used to measure freeway volume. These data collectors provide information for a freeway input-output study. In addition, some origin- destination surveying was done via questionnaires. Eight ramps .were metered: one with a pre-programmed signal cycle (i.e., based on fixed-time control strategy); And the others with gap-sensitive cycles (i.e., based on gap-acceptance control strategy). In addition, the surface street intersection signal network was revised. Since data were collected for only a few days after the meters were installed, the results were not significant. Travel time (both total and average) dropped on the freeway (about 200 vehicle-hours saving was estimated), and on the surface streets as well, although the latter is likely a result of the improved signalization network. Average speed on the freeway increased by about 15 mph during the busiest hours . U. ar 1’6 mor mil me: OH' 12 2.2.2 Houston The first studies of ramp control on the Gulf Freeway in Houston were done in 1964 (Pinnell, et al., 1965). These studies indicated that by controlling the inbound entrance ramps during the morning peak period, the inbound level of service could be significantly improved and the total travel time greatly reduced. As a result of this research, five ramps were controlled; four by closure and one by manual metering by policemen. Pinnell, et al., conducted a second study during the first three months of 1965. In this study, nine ramps were controlled along a 5.3 mile study section. Three ramps were closed, five were manually metered, and one was metered by a conventional overhead traffic signal on the service drive. The initial control period was two weeks, and in that time total daily vehicle-hours in the study area dropped from 1244 to 873 on the freeway and frontage road vehicle-hours increased only from 190 to 201. On the manually metered ramps, the stationed officer decided whether to release one or several cars at a time; it was concluded that single vehicle releases are preferable. Public response to this study was very favorable, with 65% of the respondents in favor of continuing the ramp controls. Between March, 1966 and July, 1967 several gap-acceptance ramp meters were installed along the Gulf Freeway. Buhr, et al. (1969) concluded that gap-acceptance metering was generally more desirable than demand-capacity metering, ramp geometry and traffic flow permitting. Gap-acceptance meters require a computer controller to interpret signals from loop detectors and operate the meters accordingly. In VOl ex; 2.2 met obj Var I651 13 The controllers for the first ramp meters were analog computers. In July, 1967, a digital computer was first used on the Gulf Freeway (Buhr, et al., 1969). It was hypothesized that digital computers would be ideal for optimizing a system of ramps, but would be too expensive to control a single ramp. 2.2.3 Chicago The Chicago area expressway surveillance project included a ramp metering study on the northbound Dan Ryan Expressway with the objective of gaining better knowledge of the relative effects of various geometric design features on ramp control strategies. "An interplay between the expressway and the frontage street resulted in the generation of exceedingly high entrance ramp demands at the points where the expressway curves away from the frontage street. Congestion was triggered by high volumes force-merging with a near-capacity expressway, while the frontage street, through its discontinuity, was directly involved in sustaining the cause of congestion and delaying the local recovery from congested operation" (Fonda, 1969). A ramp control strategy was introduced to the expressway in the form of fixed-time "one-vehicle-at-a-time metering utilizing manually operated portable equipment" (Fonda, 1969). The control strategy was implemented on four successive entrance ramps with the objective of adjusting the merge demands to a level that could be accommodated. The results of the study were that the congestion was not eliminated, but the extent and duration were significantly reduced. "The severity of congestion was reduced such that individual motorist save dail duri trav VEhll 2.2.4 [GSpl (app: Opera in us l4 saved up to five minutes in traversing the 3.6-mile study section. A daily average of 627 vehicle-hours of expressway travel time was saved during control, while the peak-period vehicle-miles of expressway travel increased by 5 percent." (Fonda, 1969). The increase of delay surface streets around the controlled section was negligible, but the waiting time for vehicle in queue to enter the expressway, through the metered ramps, reached 7 minutes as a maximum. Even with long waiting time the "compliance with the one- vehicle-at-a-time scheme averaged 90 percent." (Fonda, 1969). 2.2.4 Los Angeles The California Department of Transportation has the responsibility of handling the operations, operational analysis (appraisal and interpretation of traffic flows), and planning of operational improvements. By 1969, three ramp metering systems were in use (Newman, et al. 1970 and Russell, 1969). The most documented of these is the Harbor Freeway project. This project, conducted.on.a.5-mile section of the southbound lanes of the Harbor Freeway in September, 1968, evaluated the effectiveness of ramp control strategy. Speed, volume, travel time, and density were measured.befOre and.after ramp control was implemented in the form of one ramp closure and.five metered ramps. Three of the ramps released single cars at set intervals, the other two released platoons of vehicles. Aerial photography was used to take an inventory of the layout and demand on the freeway. Density contour maps were used to identify trouble spots. Origin-destination surveys were used to identify alternative routes and decide which ramps to meter or close. Denar treat turns to 40 freeu and SI of th treat: front of vei "1 lpeak.1 and ir I"? 15 Demand-capacity analyses were also used to ensure that the controlled ramp volumes would not push freeway volumes above capacity. As a result of these studies it was decided to give preferential treatment to buses, allowing them to bypass the meters and make left turns where other traffic could not. Freeway speeds increased from 20 to 40 mph. Daily savings were approximately 1000 vehicle hours on the freeway, with a resultant loss of only 130 vehicle hours on the ramps and surface streets. Newman, et al. (1970) concluded that the success of this project is due in part to unusual strategies: preferential treatment to buses, timing of surface intersection signals to allow frontage road queues to cross intersections, and two-abreast release of vehicles at one ramp. The Hollywood Freeway project consisted of one metered ramp and one ramp which is closed by barricades during the peak period. The metered ramp is operated by a pretimed signal. Travel time savings were about 450 vehicle hours per day, consisting of 500 hours saving for freeway users upstream of the bottleneck less 50 hours for loss to diverted or delayed ramp traffic, (Russell, 1969). 2.2.5 San Diego In 1969 a study was conducted on a 3.2 mile section of the Inland Freeway in Chula Vista. In that study four ramps were controlled, "Peak-hour input from these four ramps was reduced by 580 vehicles, and input to the mainline upstream of the control section was increased by 540 vehicles. Speed on the freeway increased from 26 mph to 43 mph for the higher volume. Traffic on parallel streets increased 225 vehicles." (Russell, 1969). the C9} to- nun ram 600 of 200( Ian: majc some Rout has cOnt EYEE: COntI DiEgc 16 In San Diego there are several freeway-to-freeway interchanges that are being controlled by ramp metering. Those are part of a central control strategy that involve more than 75 local and freeway- to-freeway ramps in San Diego (Wherry 1987). According to Wherry, the central control strategy does not take in account the fact that the ramp is coming from a local street or a freeway. Instead the ramps are categorized in regard of the volumes using each ramp, and the number of lanes on each ramp is also dependent on the volumes on each ramp. Those categories are: one lane for ramps with volumes less than 600 vph, one or two lanes for ramps with 600-1000 vph, two lanes (one of them is a high occupancy vehicle (HOV) lane) for ramps with 1000- 2000 vph, and for ramps with volumes over 2000 vph there are three lanes (one of them is an HOV lane). In the last two categories the majority of the ramps are freeway-to-freeway ramps but there are also some local ramps. One example of a freeway-to-freeway ramp metered location is the Route 94 Freeway connector to the westbound Helix Freeway. This ramp has 2 lanes for single occupancy vehicles (SOV) and one HOV lane. The control strategy on the ramp is fixed-time in which 2 vehicles per green per lane are allowed to enter the freeway, which means that 6 vehicles-at-a-time can enter the Helix Freeway. This ramp has been controlled in this way since 1978. The success of this strategy resulted it being used on other freeway-to-freeway locations in San Diego. According to Wherry (1987), the average volumes on the ramps during the morning peak-period are 1600 vph on the SOV lanes and 525 vph on the HOV lane, and the average maximum delays are 5 minutes and 45 sec signa 6 per: the s descri violat the di freeva contr. litera' 17 45 seconds, respectively. The violation rate for running the red signal is about 15 percent, and the violators in HOV lanes were about 6 percent of the total vehicles on the ramp. The public reaction to the strategy of controlling the freeway-to-freeway ramps can be described as fair (Wherry, 1987). This noted by the low percentage of violations on these ramps and from the long queues on the ramps, where the drivers are willing to wait 5 minutes to get onto the other freeway instead of diverting to other routes. Table 1 summarizes some of the features and results of the ramp control systems, or the case studies, that were included in the literature review above. 2. 3 comm SIMULATION MODELS "Simulation is essentially a working analogy. It involves the construction of a working model presenting similarity of properties or relationships to the real problem under study. Simulation is a technique which permits the study of a complex traffic system in the laboratory rather in the field. In a more general sense, simulation may be defined as a dynamic representation of some parts of the real world achieved by building a computer model and moving it through time." (Buhr, et al., 1968). Simulation consists of using an analog or digital computer to trace time paths. "An. analog computer is one in which computation is performed by varying the state of some physical element in which the variables are continuous." (Gerlough, 1964).- "A digital simulation TABLE 1. HARBOR HOLLYK Chula VLF Dan Ru 18 TABLE 1. SUMMARY OF RAMP CONTROL.CASES SECTION No. OF RESULTS LENGTH RAMPS V Veh. Hours FREEWAY SIT! mi mph Elmer—legal HARBOR Los Angeles 5 6 +20 -1000 +130 HOLLYWOOD Los Angeles - 2 - - 500 + 50 Chula Vista San Diego 3.2 4 +16 - - GULF Houston 6 9 +16 - 360 + 23 LODGE Detroit 3.2 8 +15 - 200 - Dan Ryan Chicago 3.6 4 - - 627 - is cha1 compute paralle after a TI steps: 19 is characterized by the use of a digital computer. Whereas the analog computer must handle all elements of the simulation simultaneously (in parallel), the digital computer handles elements of the simulation one after another (in series)." (Gerlough, 1964). The simulation of any system normally requires the following steps: 1. Definition of the problem that need to be solved. 2. Formulation of a model, or choosing a model that fits the needs of the problem. 3. Preparation of the computer "program" which will implement the model. 4. Conducting experimental runs of the simulated system or in other words calibrating and validating the model to define the parameter values to be used. 5. Interpretation of results. The model is a statement of the problem with only important features of the system under study included. "Characteristics of a system should be stated by mathematical equations when possible. If data are not known or a suitable mathematical statement is not possible, the behavior of the system is described in words. There may be parts of the system which involve random or stochastic variables. These are treated by what are known as Monte Carlo techniques." (Gerlough, 1964). The traffic flow in any given network with a specific set of rules of conduct and controls can be simulated using the previous jprocedures. Then, the effect of any change in the network variables, like the control devices, can be observed if a random sample of the traffic random 5 empirica The a desert provide compute] four-Ian foliovin 1. iStriCtl 51' 1984) 20 traffic flow is introduced into the network. The preparation of these random samples can be done by empirical data only or a combination of empirical data and some theoretical assumptions (Gerlough 1964). The formulation of a model for freeway traffic flow must include a description of system behavior in terms of rules of the road and provide methods for the implementation of these rules within a computer. Gerlough (1964) defined one possible set of rules for a four-lane divided freeway in a section without interchanges as the following: 1. Each vehicle proceeds in either the right or left lane at its desired speed or the maximum allowable speed until it encounters another vehicle in the same lane. 2. ‘The encountering vehicle, if it is in the right lane, examines the lane to its left. If the encounterring vehicle is in the left lane, it examines the lane to its right. A lane change is made if it is safe to do so. If it is not safe to change lanes the encounterring vehicle decreases its speed to that of the encountered vehicle. 3. During each time increment, all vehicles in the left hand look for opportunities to move to the right. 4. During each time increment, all vehicles traveling at speed less than their desired speed look for opportunities to increase their speeds. The simulation approach is far more appealing and practical than a strictly empirical approach for the following reasons (Goldblatt, et a1, 1984): 1. It is less costly, by far. Fu designe his/he; iGEntif Co more C} dramati. first a c°mPUt61 ramp Co develop in the f 21 2. Results are obtained in a fraction of the time required for a field experiment. 3. The data generated by simulation include many MOEs that cannot, practically, be obtained empirically. 4. Disruption of traffic operations is completely avoided. 5. Many designs requires significant physical changes to the facility, such changes cannot be implemented for experimental purposes. 6. .Analysis addressing the operational impact of projected traffic demand patterns or of new facilities must be conducted by simulation or equivalent tool. Furthermore, these models produce information which allows the designer to focus his/her thinking, to identify the weaknesses in his/her concepts or designs, and therefore to provide the basis for identifying the optimal form of his/her candidate approach. Computers have been used to simulate freeway traffic flow for more than thirty years. Ihmtfids time, the models have changed dramatically. According to the Federal Highway Administration (FHWA), "the first actual documented simulation was performed in 1955 on an analog computer" (Ross and Gibson, 1977). Since the late 19603 computer use has become more widespread in ramp control strategies. Several simulation models have been developed to model ramp metering. The object of this is to eliminate the cost of implementing and evaluating different metering strategies in the field. Lev duplica includin statistf The anaI relatio: acceptan primari? configur By wk on interseq hovever’ fledels. advance (Ger-IOU. reEat'dir elemtints 22 Levy, et al. (1961) developed a digital simulation model which duplicated traffic flow'on.a 17,000 foot section of a freeway, including two on-ramps and two off-ramps. This model was derived from statistical.analysis of traffic data collected from several freeways. The analyses performed included development of a volume-speed relationship, and investigation of traffic lane distribution, gap acceptance levels, and behavior of exiting vehicles. This model was primarily used to determine the effects of different interchange configurations and spacing. By 1964, the Highway Research Board (HRB) recognized published work on different types of traffic simulation, including freeway, intersection, network, and tunnel simulations. Most of this work, however, dealt with simulation theories and techniques, not actual models. Levy's model was lauded in the HRB report as one which "may advance techniques to the point of usefulness for design purposes." (Gerlough, 1964). In that report also some guidelines were given regarding the important elements of any simulation model. Those elements are: 1" Statement of the behavior of each of the components and inputs of the system. This also include the probability distribution of any random phenomena. 2. Selection of the measures of effectiveness by which the performance of the system will be judged. 3. Statement of any particular assumption or simplifications of the model which may be necessary to permit the adaptation of” the model to a particular computer. to the its ade Br develop types simulal logic 111 an: capaci1 Tl Buhr n for ur Vehicl iflvolv traffi Differ Model, and tr and {h Inami °ff‘ra Side 0 in the ’11 "f ePar 23 At this point, the digital computer was determined to be superior to the analog computer for the purpose of traffic simulation, due to its adaptability and ability to handle more diverse input data. Buhr, et al. (1968) at the Texas Transportation Institute (TTI) developed a microscopic model to analyze the effects of the different types of ramp metering. This model was similar to Levy's in that it simulated only a short section of freeway with up to six ramps. The logic was based on more recent studies of gap acceptance done by the TTI and was designed to replicate the results of fixed-time, demand- capacity, and gap-acceptance metering, as well as no metering. The simulation logic for stepping vehicles through the system, in Buhr model, was divided into "three classifications: (a) flow logic for unimpeded vehicles, (b) car-following logic for platooned vehicles, and (c) maneuvering logic for vehicles executing maneuvers involving more than a single stream of traffic.” (Buhr, et el., 1968). Sinha and Dawson (1970) developed a microscopic model based on traffic behavior equations listed in the 1965 Highway Capacity Manual. Different freeway traffic situations can be simulated, using this model, by merely inputing the descriptive geometric characteristics and traffic data including speed distributions, total traffic volume, and the percentage of commercial vehicles in the stream. The Sinha and Dawson model had the capacity for simultaneous, dynamic analysis of traffic flow on 5 freeway lanes, 4 on-ramps, and 6 off-ramps. The ramps may be located on the right-hand or left-hand side of the freeway. The program logic for the processing of vehicles in the system was divided into five parts. Separate routines were prepared for the processing of vehicles on through lanes, on-ramps, '1 91' Si .5“ ‘1. 0f 24 acceleration lanes, deceleration lanes, and off-ramps. This model was ”validated at both microscopic and macroscopic levels. Several different macroscopic comparisons were made between simulated phenomena and data collected on sections of the Eisenhower Expressway' in Chicago and a Long Island parkway, and data reported in the 1965 Highway Capacity Manual. The comparisons were consisted and reasonable.” (Sinha and Dawson, 1970). In 1977 the FHWA prepared a review of network simulation models. Nineteen models were discussed in three classes: single road, single intersection, and network. Ten models were considered obsolete, six models were considered suitable for current computer use, and the simulation portions of three signal optimization models (i.e. , TRANSYT, SIGOP II, and CORQlC) were examined. The report (Ross and Gibson, 1977) discussed the operating principles and unique features of each model, as well as the validity and usefulness of the output. The computer language, type of machine needed, core requirements, auui execution speed were listed, if known, for each model. During 1970-80 the range of work utilizing traffic flow simulation increased. Studies by Sakashita, et a1. (1971) and Posner (1976) included determination of optimal motorist-aid strategies and the economic impacts of high-occupancy-vehicle lanes. The evolution of the simulation model continued as well, with models such as INTRAS and FREFLO capable of modelling larger freeway segments and incorporating accident and accident response simulation. Previous measures of effectiveness such as travel time and delay time were joined by new measures such as fuel consumption and pollutant emission (Payne, 1979). Tl summa: traffic 2.4 1 Tl contro meteri1 is con: Tl is als used 11 Nt Controj local $0me i mathemé °f the 25 There are many traffic simulation models available, and table 2 summarizes the features of those models which simulate freeway traffic. 2.4 SUMMARY The literature review reveals that the use of ramp metering as a. control strategy for urban freeways is widely used. Different ramp metering strategies are used, but the the fixed-time metering strategy is considered most reliable and simplest to implement. The strategy of metering a freeway-to-freeway interchange ramp is also discussed in the literature review, the technique has been used in San Diego successfully. Nothing was found that addressed simulation of freeway-to-freeway control strategies in conjunction with the usual control strategies at local on-ramps by means of computer models. Gordon (1972) developed some ideas, but they were theoretical and based upon developing mathematical equations to calculate the delay and queue at the on-ramp of the interchange. 'The use of traffic simulation models, as a reliable approach, to evaluate the effectiveness of ramp metering operations is also documented in the literature review. é :2 l INTRAS FREFLO FRECON coeo CORCog TRAFLO parts C 26 TABLE 2. SUMMARY OF FREEWAY SIMULATION MODELS MODEL MODELLTYPE MQQ§L_EQRPOSE TRAFFIC FLOWS INTRAS stochastic incident detection vehicle-specific microscopic and evaluation of time-stepping control strategies simulation. FREFLO deterministic simulate freeway conservation equation macroscopic 1-direction dynamic speed density FRECON macroscopic simulate freeway modified from FREFLO l-direction FREQ macroscopic simulate freeway H.C.M. (speed-volume and evaluate curve) priority lanes. CORQ macroscopic queueing in step-function freeway corridor. travel time. CORCON macroscopic queueing in step-function freeway corridor. travel time. TRAFLO microscopic all networks FREFLO parts of this table were taken from Aerde, et al. (1987). 2.5 01 ll guidel: differc t0'fIEl T? model, the 01 evalua impleml 27 2.5 OBJECTIVES OF THE STUDY The objective of this dissertation is to develop operational guidelines to measure the efficiency of flow on the freeway through different ramp metering strategies, including the metering of freeway- to-freeway interchange ramps. The study conducted in this dissertation utilizes the INTRAS model, which is a microscopic freeway simulation model, to determine the optimal strategy for metering flow onto the freeway, and to evaluate the effect of the different types of strategies that can be implemented. 3.1 mod be ten an: iea 5101 St; CHAPTER3 THE INTRAS MODEL 3.1 GENERAL From the beginning of this study it was apparent that computer modeling would be required since the number of variables required to be analyzed precluded a hand calculation approach. The problem that remained was finding a software package that could perform the desired analyses. INTRAS was selected to meet this need because of its features and capabilities which were summarized in table 2 in chapter 2, and are discussed in more details in this chapter. Released in 1980 by the FHWA (Wicks and Andrews, 1980), INTRAS is an acronym for Integrated Traffic Simulation. INTRAS has a number of features that make it suitable for the system analysis required in this research. INTRAS allows for an unprecedented level of detail in the modelling of an urban freeway system. It is a vehicle-specific time— stepping simulation designed to represent traffic and traffic control strategies in a freeway and surrounding surface street environment. 28 sir Lie 118' err re: 51 is re Va de 511 29 3.2 PROGRAM PURPOSE AND CAPABILITIES INTRAS has been developed for use in studying freeway incident detection and control strategies. It is based on knowledge of freeway operations and surveillance systems and incorporates detailed traffic simulation logic developed and validated for this purpose, (Wicks and Lieberman, 1977). To allow simulation of freeway control policies, including ramp metering and diversion, the capability of modelling the off-freeway environment (i.e. , the ramps and the surface streets that service the ramps) is included in INTRAS. 3.3 NETWORK IDEALIZATION AND MODELLING CONCEPTS The representation of a "real world" traffic system in the terminology of INTRAS is the most important task a user faces. The simulation results cannot reflect the actual traffic system unless it is accurately represented to the model. The model's concept of the real network is built upon the data supplied (i.e., measurements of various network features and characteristics). A familiarity with definitions of these features is, therefore, required for the user to successfully utilize INTRAS. 3 . 3 . 1 Network Representation The geometric representation of a roadway system for the INTRAS model is comprised of links (one-directional roadway segments) and nodes (intersections or geometric discontinuities). The logical division of a road system into links may correspond to the natural segmentation caused by cross streets or ramp junctions. Figures 1 and 30 2 represent a typical roadway system and its network representation, respectively. If analysis of a natural segment indicates different characteristics on one portion than on another, it may be desirable to further subdivide the segment. For example, if it is observed that on the upstream portion of a segment, traffic always travels at a slower speed than on the downstream, the actual segment may be represented in the model inputs by two links with differing free-flow speeds. A change in grade is also sufficient reason for segmentation. Implementation of this type of characteristic would be accomplished by the insertion of an additional node. For example, link (8,9) in Figure 2 might be partitioned into two links, (8,15) and (15,9), by the insertion of an intervening node 15. To permit appropriate logical treatment for roadway sections of diverse characteristics, and to realize some computer storage economy, three link types are defined for INTRAS. A ”surface" link is defined as a non-freeway roadway segment servicing one direction of traffic. The nodes at each end represent at-grade intersections. Each surface link extends from the upstream stopline to the downstream stopline. Vehicles traversing an INTRAS surface link are moved at constant time intervals. The method properly replicates (Peat, et a1. 1973) the dynamics of traffic on urban networks. A "freeway" link is defined as a one-way roadway segment, of a controlled-access highway, and is characterized by generally constant geometric characteristics (grade, curvature, number of through lanes). The extremities of a freeway link correspond to either ramp junctions or significant geometric changes. 31 Roma—v mxmtccc new msu_3 "wetnem uuohuoz peel unannouhehniouuh Huuumhah odnlwm .H unawah @ vac: ...onu::.... Ir“ 11111111111111.1111 ‘11 11.1.1 0 111111 /1 x /111\ llllllll llllllrlllel 1 llll . A a >c3:;LL m :oEoow1VA1c 2.0395110 :oEuoleLm cowuuoqld cogumm 32 Acmo_. mxmtuc¢ new meow: "wuc=om avenue! oanlnm we neuueunonounom .N shaman .9 a... xuuco moouusm ooouusm .. e . 0-..... e >uuco ooeuusm . 0 o 02.9 a U m < >33ooum «A auuco ooouunm uwxo UUfluuam a u. 33 Each freeway link may contain up to five through lanes and two auxiliary lanes. Each auxiliary lane may be described as: a. acceleration- A lane which extends from the upstream extremity of a freeway link to some mid- link position b. deceleration- A lane which extends from a mid-link 'position t01the downstream extremity of a freeway link cw full-AX lane which extends for the full length of a freeway link with at last one end connecting to an on or off-ramp Auxiliary lanes may occur on either the left or right-hand side of the roadway. Typical freeway links are illustrated in Figure 3. Vehicles traversing freeway links move in accordance with the logic of car following, lane-changing and vehicle generation component models developed for INTRAS (Wicks and Lieberman, 1977). A ”ramp" link is defined as a one-way non-freeway roadway segment which connects directly to a freeway link. Ramps may be one or two lanes in width. Ramp links are further characterized as either on or off-ramps indicating the end of the link which connects to the :freeway. The same logic is applied to move vehicles on ramp links as on surface links. \ m 34 Change Ramp Ramp Change in in Grade Junction Junctign Curvature Node 5 Node i Node 3 Node 1 / A A m _ _9_. _ _.\ ______ ——-p—————————n——— ——-b—-—:—————-—— re— Freeway Link—fl, ‘s—Freeway Link —'1 ' (103) (1.1) Hreeway Link—D1 . 1-¢——J-§reeway Link (1.3) (id) Node j Node i Node 3 Node 1 // fl—fi % 4.11.. w/ j ———- .__ ____.__£._:1 ——' ————————— "4:" -—- ————————— 7—‘-—— "“‘1:::: ————— 7: i) ,r’ 1-q——Freeway Link—’1 g/n—Freeway Link—w (id) (id) 522: A a Acceleration Lane D a Deceleration Lane F - Pull Auxiliary Lane Figure 3. Typical Freeway Link Configuration Scarce: flicks and AndFEuE (198G? 35 Because the simulated network is just a portion of some real world traffic system, special links have been devised to handle conditions at the network extremities. These links serve to process vehicles in to and out of the simulated study network. Links handling incoming traffic are called entry links. The INTRAS model allows both freeway and surface entry links. They are coded on input cards and processed the same way as interior freeway and surface links but are subject to a few additional requirements. For freeway entries, auxiliary lanes may not be specified, nor pockets on surface entries. It is not necessary for these links to exactly replicate geometry of their real world counterparts. What is important is that the incoming volumes, distributions of vehicle types, and incoming lane distribution (for freeway entries) be accurately specified. The performance of vehicles on these special purpose links are not included in the network totals of the output reports. Traffic leaving the network is said to move on to exit links. These links are never coded on input cards and therefore not included in link data arrays or processed explicitly. The notion of exit links only exists in connection with traffic movements. If an interior link Specifies a node on the periphery as a destination for some traffic movement (i.e. , left-turn, thru or right-turn), then an exit link is implied and traffic may leave the network by it. Nodes on the periphery of the network (i.e., upstream nodes of entries and downstream nodes of exits) are identified to INTRAS by node numbers greater than 699. Values from 700 to 799 are reserved 36 for freeway peripheral nodes, while those from 800 to 899 represent nodes associated with surface entries and exits. 3.3.2 Geometric Features To model a roadway system in sufficient detail to replicate "real world" traffic statistics, it is necessary to accommodate those geometric features which significantly affect traffic performance. These features included in the INTRAS design are: Intersections - Each intersection is identified by a unique node number. Links are identified by the ordered pair of node numbers which identify their upstream and downstream extremities. There may be up to four links approaching, and four links departing, at a given intersection (node). Vehicles on each approach link to an intersection may have up to three destinations (receiving links) upon passing through that intersection. Each of these receiving links is entered by performing the associated traffic maneuvers: left turn, through movement or right turn. left Unnmrs seek gaps in opposing traffic; right turners slow’befOre turning, etc. Freeway-Freeway and Freeway-Ramp Interconnections - The lane alignment of freeway links and on-ramp links with the next downstream freeway link is defined by two input specifications. First, the number and type (through, auxiliary) of lanes which comprise each lard: is specifiede Second, the lane in the downstream link which receives 37 traffic from the right-most through lane of the upstream link must be identified. Freeway links are logically connected to downstream off- ramps by specifying the number of ramp lanes, and whether it is a right-hand or left-hand off-ramp. The outside lanes on the designated side of the freeway are then internally assigned as connecting to the off-ramp. Grade Specification - INTRAS has been designed to accept link-specific grade as input. Thus, it is proper to define a continuous section of roadway (containing a significant change in gradient, usually 1%) as two contiguous links, with a node defined at the point where the grade changes. Curvature - A change in horizontal curvature is sufficient reason to segment a roadway section into two links. Two methods of limiting vehicle performance on horizontal curves are available in the INTRAS design. First, a lowered value of desired free-flow speed may be defined for an affected link. Although easy to apply, this method presumes some pre-analysis on the part of the user. Second, radius of curvature, super elevation and pavement condition may be defined. An internal table is referenced to determine friction coefficient from pavement condition. The basic equation for vehicle operation on a curve is then used to generate an upper bound for desired free-flow speed. V - JI5RZe+f) 38 where, e - rate of roadway superelevation, foot per foot f- friction coefficient for given pavement condition radius of curve in feet as I .< I vehicle speed, miles per hour The simulation model applies the minimum of the input free- flow speed, and the curvature dictated upper bound, to traffic on the subject link. lane Separation - The typical freeway often contains sections changing where lane changes are physically prohibited by virtue of barrier curbs or traffic islands. These restrictions are designed to segregate through traffic from weaving traffic, or, to guide vehicles around some obstruction (bridge abutments, etc.). INTRAS is designed to accept physical barriers of this nature on a link-specific basis. 3.3.3 Traffic Flow Patterns Examination of the flow of traffic through a "real" traffic system is necessary to set up traffic flow patterns through a network. Turning movements (as percentages or counts) must be defined by the user in the model input. Lane channelization and early‘warning signs provide the model with information needed to guide vehicles into the proper lanes to negotiate these prescribed maneuvers. 39 The early warning sign capability of INTRAS allows the user to define the point on the roadway at which drivers begin to react to an upcoming off-ramp. As simulated vehicles pass an early warning sign, they are assigned to either turn or remain on the freeway at the indicated off-ramp. Their desired lane thereafter reflects this downstream movement. If an early warning sign is not specified for a. particular off-ramp, then vehicles do not exhibit lane preferences (due to the off-ramp) until they enter the freeway link which connects directly to the ramp. 3.3.4 Signal and Sign Control Each intersection in a simulated study network requires a control policy to establish the right-of-way for approaching vehicles. INTRAS has the ability to simulate both fixed-time signal control and sign control. Provision has been made for the modular inclusion and referencing of the specially coded subroutines to model traffic responsive signal control. Ramp metering and freeway traffic diversion procedures (described in later sections) utilize this provision. Fixed-Time Signal Control - Intersections of an INTRAS simulated network may be controlled by fixed time signals of up to six control intervals each. During each interval, one of the following standard signal configurations is applied to control each of the approach links: 40 Amber Green Red Red with Green Right Arrow Red with Green Left Arrow Red with Right Turn after Stop No Turn - Green Through Arrow Red with Left and Right Green Arrows No Left Turn -Green Through and Right The duration of each control interval is user-specified. In this research the network no surface street intersections were modelled due to time and data limitations. Sign Control - Each intersection not controlled by a fixed-time signal is controlled by either stop or yield signs. The user must specify which approaches face such signs. For the common situation, where no control of any kind is present, the INTRAS user needs to specify yield signs for one approach direction to indicate the minor street . 3.3.5 Traffic Descriptive Features Each driver-vehicle pair in a traffic stream behaves as an individual entity having different motivations and standards of performance. This quality is modelled in INTRAS to achieve the proper stochastic variation in individual vehicle performance. To accomplish this, the INTRAS design provides for five vehicle types, each possessing its own family of vehicle characteristics (length, speed acceleration.profile, etc.). These characteristics may be revised as 41 an option, so that the particular vehicle types chosen for the basic INTRAS model do not constitute a limitation on the user. The vehicle- types chosen for the basic INTRAS model are: Low Performance Passenger Car Intercity Buses Single Unit Trucks Trailer Truck Combinations Variations within vehicle types are attributed to differences in driver performance. Decile distributions of these characteristics (variation about mean free-flow speed, queue discharge headway, etc.) are implemented in the INTRAS model. 3.3.6 Freeway Traffic Responsive Control Vehicles entering the freeway via on-ramps may be subjected to a ‘variety of control techniques. In parallel to, or independent of on- ramp control, diversion of freeway vehicles to a parallel service facility may be simulated. In this dissertation the diversion option was not used because there are no continuous service roads parallel to the Ford Freeway where this study was conducted. On-Ramp Controls: Four methods of on-ramp control can be implemented in the INTRAS model. A typical geometric configuration of a metered on-ramp site is shown in Figure 4. 1. Clock Time metering: To simulate clock-time control of on- ramp, one fixed metering rate (vehicles per minute) is specified at each such node. A countdown clock is assigned to each associated on-ramp and the signal is set to "green" 42 Frontage Road I | 1 1 fl I ”—1 l l A 1H Freeway I l l I l l ”i- —~L-— l r l , l l I I I | B C E U___—_ ——— 1521: Auxiliary Lane of Freeway Link B,A Left Turn Pocket of Surface Link E.D Figure 4. Typical Metered Ramp Geometry Butter; 5111.: din.‘ hlull'i‘u‘. (195W 43 each time the clock returns to zero. The signal is maintained at "green" until a vehicle is discharged, and is then set to ”red". 2. Demand-Capacity ramp metering: In this ramp metering strategy, vehicle headway on the ramp is set at a rate dependent on the available capacity of the freeway. The level of available capacity is established by comparing the number of vehicles on the link with a given capacity value set by the user. 3. Speed control ramp metering: For this control option, a vehicle is released if the speed on the freeway is above a user-specified threshold value. 4. Cap acceptance merge control: Under this ramp control option, a vehicle is released when an acceptable gap exists on.the freeway receiving link. The acceptable gap is specified.as a minimum required headway between two vehicles on the freeway link. 3.4 SIMULATION AND PROGRAMING METHOD The INTRAS simulation model employs a time stepping procedure for ‘moving discrete vehicles through the simulated traffic network. Each time step all vehicles in the network are processed in accordance with their desired speeds and destinations inhibited by the immediate traffic and control environment. A description of the various traffic and network characteristics modelled by INTRAS was presented in Section 3.3., and more discussion about the different parts of INTRAS simulation logic is presented in section 3.6. 44 3.5 ASSUMPTIONS AND LIMITATIONS Certain restrictions on network geometry and parameter values were required before the programming of INTRAS could begin. Reasonable values for these restrictions were chosen by considering the mission of the program and then determining limitations which could not reasonably be considered to interfere with the expected applications. These design limitations are the subject of this section. The geometry of the simulated network is restricted as to link lengths, maximum allowable lanes on each link, and number of approaches to each node. These restrictions are identified in Table 3. Similar in nature to the geometric restrictions is the limitation on signal control intervals. A maximum of six such intervals are permitted for signal controlled intersections. In the calibration of INTRAS (Wicks and Lieberman, 1977) assumptions were made as to the degree of detail required to accurately represent the dynamic characteristics of freeway traffic. A maximum of five vehicle types were defined. The first two types are allowed to exhibit different acceleration characteristics in the freeway and non-freeway environments. Five grade categories are provided. The first of these categories is assumed to represent a negative gradient, and so, no limitation on desired speed (due to grade) is designed for this category. 45 TABLE 3. INTRAS Geometric Limits Definition Limitation Number of through freeway lanes per link 5 5 Number of auxiliary freeway lanes per link 5 2 Number of ramp lanes per link 5 2 Number of surface lanes per link (including pockets) 5 5 Number of right turn pockets per surface link~ S 1 Number of left turn pockets per surface link 5 1 Length of freeway links 5 9800 feet Length of surface and ramp links 5 3265 feet Number of approaches to surface intersection (node) 5 4 surface links or s 3 surface links and 1 ramp link Number of approaches to freeway intersection (node) or IA 1 freeway link 1 ramp link 46 3. 6 SIMU'IATION DEVELOPMENT 3.6.1 The Car Following Model A fail-safe car-following model is the process of determining a vehicle's speed and position given that its leader has a speed and position that has already been calculated for the current time scan. Generally, the output of the model is the acceleration of the following vehicle. A fail-safe model has two elements. First, there~ is the car-following model which calculates the follower's behavior based on some prescribed desired speed. Secondly, there is an overriding,collision prevention model which is based on the following vehicle being able to avoid a collision when the leader undergoes its most extreme deceleration pattern. The algorithm used in INTRAS for the car-following model is called "The PITT Algorithm". The primary car-following relationship in the algorithm is that a following vehicle will attempt to maintain. a space headway that is calculated by the following equation: Space headway - L + kv + 10 feet. Where, L is the length of the leading vehicle, v is the speed of the leader, and k is a calibration parameter which is a function of driver type. The full car-following formula and its derivation can be found in Wicks and Lieberman (1977). An initial operational test was applied to this car-following model through simulating the car-following behavior in a single lane. Platoons of two vehicles and five vehicles were run down the lane at a 47 constant speed. An artificial velocity disturbance was applied to the leading vehicle, and the behavior of the followers was examined. Figures 5 and 6 show the behavior of five-vehicle platoon traveling at 60 feet/second, with either a one-second or three-seconds scanning interval. The velocity of the leader was varied by applying an acceleration of -6 feet/second/second for 6 seconds, a zero acceleration for 3 seconds and an acceleration of 6 feet/second/second for 6 seconds. The figures illustrate the velocity response of the third and fifth vehicles in the platoon. The results of the test are excellent, as can be seen in the figures, with the following vehicles demonstrating good oscillatory behavior, while remaining fundamentally stable. The behavior at the longer scanning interval was reasonably consistent. Overall, under the simple operational test, the PITT model consistently showed satisfactory behavior, Wicks and Lieberman (1977). 3.6.2 Lane Changing Process The development of the lane changing component in INTRAS was given a great deal of attention since it is an essential requirement that the model satisfactorily perform lane changing and merging at high volumes. It is also essential that the lane changing component of INTRAS be fully integrated with its the car-following component. Figure 7 shows the lane changing process in INTRAS. Where "a vehicle wishing to change to another lane, vehicle 3, looks at the gap available in that lane and carries out the following checks: 48 .iiai. c...msm_s as. msu_, ”motzcm Hu>uoucH ocooom ecolenuauomad aanm “uofi>mnom cooumam .n ousmfim Amocoomm. oEaa ov on ON OH F b b m. macaco> II:.III.|II.:II m. oaodno> a. oaoano> 39985 (ft./sec.) 49 :2»: cmnuwam: new $35 "9.25m Hm>uoucw ocooom mouneuenuwuooam aaHm ”uofi>mswm cooumam .o muowfim Amocoowmv mafia cc on om OH 2 303m; .I...III n. maca£m> H¢ 0H0w£o> Speed (ft./sec.) 5C) .mkow. cmutwnmig tam maoiz "motaom moaoflno> ocwmcmnu mama one .h musowh 51 1. Does the lead headway to the gap leader, vehicle 1, satisfy the car-following rules? 2. Does the lag headway to the gap follower, vehicle 4, satisfy the car-following rules? If the answer to both is yes, then the vehicle can move to the new lane." (Wicks and Lieberman, 1977). The default value, in INTRAS, of the acceptable gap in the target lane is 3.1 seconds, and it is applied deterministically. This default value was used in this research after several sensitivity runs with different values for the acceptable gap showed that 3.1 is the best value to be applied in this situation. For a full review of the lane changing process in INTRAS see Wicks and Lieberman (1977). 3.6.3 vehicle Generation The Vehicle generation in INTRAS takes place on an entry link (i.e., a dummy link) which feeds the first link of the freeway or the first link of the surface road. The vehicles are generated using a negative exponential gap distribution. The vehicle characteristics are randomly generated (i.e., driver type, vehicle type, desired lane, and desired speed), Wicks and Lieberman (1977). The headway between the generated vehicles is "checked through the car-following equation and ,if too short, is adjusted upward to the minimum safe following position. The speed and position of the new vehicle are thus determined. Vehicles are generated such that each lane of the dummy link always has at least two vehicles in it unless an excess of vehicles has already been generated. In this way, 52 each generated vehicle has time to respond to the car-following rules and be operating normally by the time it enters the simulated freeway." (Wicks and Lieberman, 1977). For more details on the vehicle generation logic of INTRAS see Wicks and Lieberman (1977). CHAPTER4 METHODOLOGY 4.1 IMPIEMENTATION OF INTRAS Since the INTRAS model has not been released for public use yet, a copy of it was obtained directly from FHWA for the study. The FHWA copy of the model was written for IBM mainframe, and the first step in the implementation was to convert the model so it could be run on the MDOT Burroughs mainframe. The conversion process was a lengthy one since the Burroughs has an old Fortran compiler version, and its random access memory (RAM) is too small to handle all the subroutines and large arrays of the model. These limitations of the Burroughs mainframe led to many changes in the source code of the model. The major changes were the elimination of the fuel consumption and the plotting subroutines from the model, and the reduction in sizes of many storage arrays. Since the INTRAS simulation model requires a lot of input data with many variables involved, the second step in the implementation process was to break those variables into two categories: control variables, and fixed parameters. Control variable: The primary goal of this dissertation was to improve the freeway traffic operation by optimizing the metering rate on the on-ramps. This specific scope led to the choosing of only one 53 54 control variable, which was the timing of the signals, which control the metering rate of the ramps. INTRAS has the ability to model this metering rate 1111four different ways, because it has four types of ramp metering control methods: Clock time metering, speed control metering, demand/capacity' metering, and gap acceptance merge control. In this study, the control method that was used to model the control variable was the "Clock Time Metering". That decision was based on the conclusion, from the literature review, that it is the simplest method to implement and the most reliable of the four methods. Also the literature review (Munjal, 1973, Buhr, et al., 1969, and Buhr, et a1. , 1969) revealed that the other three control methods have high failure rates and are not stable because their implementation depends totally on accurate and continuous operation (which is usually hard to achieve) of implemented detectors in the pavement. Fixed parameters: The rest of the potential variables (both network variables and model parameters were kept fixed during the study. 4.2 STUDYAREA The evaluation was conducted on the portion of the Ford Freeway (1-94) within the Detroit city limits. That is, where the ramp meters are installed on that freeway. The study area boundaries are shown in Figure 8. The Ford Freeway runs about 15 miles inside the City of Detroit, where it has three major freeway-to-freeway interchanges. All the 55 8335.8 no.2 63¢ .a 853m ya h- u a i. ,.._//a_. ...\. Tran. .. I..- J. _.._ x [i m \ . 3 . a ....... 56 non-freeway entrance ramps have ramp metering signals to control vehicle entry to the freeway at a rate of one car per green interval. Because of the size of the network, it was not feasible to collect data at each location of interest along the freeway. Instead, a set of locations were chosen in a way to cover all different operational situations (i.e., the merging areas after local entrances in different locations along the freeway). Figure 9 shows the selected locations. Later in the study each location will be referred to by the immediately proceeding ramp. The choice of the locations was planned in a way that: (1) both directions of the freeway will be covered, (2) ramps in between freeway-to- freeway interchanges will be sampled, and (3) ramps at the outskirts of the freeway will be sampled. Another consideration in the sampling was to cover, in most locations, both the morning and the evening peak hours. The following table ,Table 4, shows the selected locations, date, time, and duration of data collection. 4.3 DATA COLLECTION 4.3.1 Data Elements There are two types of data collected for this study: first, data for building the network; and second, data needed for calflorating and validating the model. 57 Human-um Hem ...—caucus Gounod-m .o 055.: EDGE—:4: e e o \ 3:; s n a N) E SBIUJSEN‘ 8313A8H3 n 395:: \1/ WWI _ _ {d A... MU as: a. kg? .m czowav 58 TABLE 4. SELECTED LOCATIONS: DATE, TIME, AND DURATION OF DATA COLLECTION MTION DIL D_ATE TIME DURATION METER LINWOOD EAST 09/03/86 8:00-5:30 10 min/hr on LINWOOD EAST 09/10/86 8:00-5:30 10 min/hr off MT.ELLIOT WEST 09/17/86 9:00-5:30 10 min/hr on MT.ELLIOT WEST 09/18/86 9:00-4:30 10 min/hr off VAN DYKE WEST 09/17/86 8:30-noon 10 min/hr on VAN DYKE WEST 09/18/86 8:30-noon 10 min/hr off JOHN R. WEST 10/15/86 3:30-5:00 15/30 min on JOHN R. WEST 10/16/86 3:30-5:00 15/30 min off TRUMBULL WEST 10/28/86 3:30-5:30 15/30 min on TRUMBULL WEST 10/29/86 3:30-5:30 15/30 min off GRATIOT EAST ll/l8/86 3:30-5:00 15/30 min on GRATIOT EAST 11/19/86 3:30-5200 15/30 min off MT.ELLIOT WEST 04/16/87 7:30-3:00 5/30 min off 59 1. Model building data: Table 5 summarizes the geometric and operational data elements needed for running the INTRAS model. The geometric data were taken from the Ford Freeway design plans, and the operational data were collected from the computer outputs of the main frame computer that controls the SCANDI system. These data include: total volume on the freeway, volumes on each lane on the freeway, and volumes on entrance and exit ramps. 2. Field data: These data includes vehicle speeds, vehicle mix, and volumes on both the main freeway and the on-ramp at each sample location. The procedure that was used to collect the field data is discussed in section 4.3.2. Figure 10 summarizes the data elements needed for this study and their sources. 4.3.2 Field Data Collection Procedures The collection of data was done in two steps: first, pilot data were collected during the period between April, 1986 and July, 1986. These data were used to check both the ability of the model to operate correctly, and the ability of the students involved in the data collection phase to operate with consistency and accuracy. Second, the final data were collected at the selected sampling locations between August, 1986 and April, 1987. The dates for collecting the final data were chosen to represent normal traffic operations and volumes for the City of Detroit (i.e., schools are open). Furthermore, the data were collected only during normal weekdays (i.e., Tuesday, Wednesday, and Thursday) to avoid any 60 TABLE 5. INPUT DATA REQUIRED FOR INTRAS M1319 Links defined by upstream, downstream node numbers. Link lengths. Number of lanes. Turn pockets. Grade. IBAEEIQ VOLUMES On all entry links nodes stratified by vehicle type (up to 5 types) Link-specific turn movements. C CONTRO PECIFICATIONS Stop and yield signs. Turn restrictions. Traffic signals. Traffic control may be fixed-time or traffic-actuated. Route diversion specifications. R V ' D OPE ION CHARACTERISTICS Driver's response mechanisms: free-flow speed, sensitivity, discharge headway. Link-specific mean speed for free-flowing traffic. Vehicle-type operational characteristics: acceleration, deceleration. 6l M VOLUME W ggomszagg ONT 0 IGNS, 11) LLLQLZL. MK (3) (3) \ INTRAS CALIBRATION 1"” VA I ATION Ta— I 1 EVALUATE THE PRESENT SYSTEM TEST NEW STRATEGIES (1) Field data (2) SCANDI computer data (3) Design plans FIGURE 10. DATA ELEMENTS: SOURCES AND FLOW THROUGH THE STUDY 51' pe da f o lo. tht COn SCAJ 62 abnormal traffic situations due to the start of the week (i.e., Monday), or due to the weekend traffic fluctuation on Friday afternoon hours. Efforts were made to collect the data during dry weather days so the weather condition effect on driver behavior will not affect the evaluation process. A video camera, that has a built-in timing clock, was used to collect the data. This was done by placing the camera on the bridge that crosses over the freeway following the intended sampling location. Pavement taping marks were placed at 50 foot intervals on the shoulders of each selected location prior to filming, and used when performing data reduction. For the locations where both peak hours were to be sampled, the data collection procedure was to film a 10 minute period of each hour for the whole day (i.e., between 8:00 a.m., and 5:30 p.m.). For the location where only the evening peak hour's data was to be collected, the procedure was to film the whole peak hour period (i.e., between 3:00 p.m., and 5:30 p.m.). Special arrangements were made to collect data needed for the study, since the ramp meters are already installed and in operation. These arrangements were made in coordination with SCANDI operation control engineers, since the ramp meters are controlled from the SCANDI operation room. These special arrangements consisted of the following: a. Coordination of a timetable for collecting the data at each location. tI wi' USI Co] are be: red 4.4. moniz 63 b. The collection of data at each location was done during a period of two consecutive days. c. On the first day, data were collected with ramp meters in operation (i.e., all ramp meters ON). d. In the second day, for the same location, the ramp meters were turned off and data were collected in the same manner as the first day (i.e., all ramp meters OFF). e. The data collection procedure was performed on two days per week at the maximum to avoid any false traffic diversion due to changes of ramp metering status. Through the SCANDI office, state police operational reports were obtained for same periods of time that filming took place to verify that no incidents occurred which might affect freeway flow. 4.4 DATA REDUCTION The Mt. Elliot entrance ramp onto the west bound Ford Freeway will be used as an example to illustrate the data reduction procedure used throughout this project. The data used for this example were collected September 18, 1986, between 10:00 a.m. and 10:10 a.m., and are considered to be a representative sample for the one hour period between 10:00 a.m. and 11:00 a.m. The following steps summarize the reduction procedure: 4.4.1 Building the Grid Using the pavement marks that were placed on the shoulder, a grid consisting of two lines that are 100 feet apart was drawn on the monitor screen that is showing the data, as in Figure 11. 64 4.4.2 Sample Size Two procedures were used to reach a decision on the sample size required for the study: 1. Statistical approach: The following equation was used to determine the sample size: n - [ (Za/ZY" * (5)2 / d2 where: n - the sample size; d - tolerable margin of error of mean value; 8 - standard deviation of sample distribution; and Z - standard normal statistic (table value). A pre-study was conducted to calculate the values needed for the equation, and the results were found as follows: d - i 2.0 mph was found to be a reasonable assumption. This was decided by comparing the estimation of the different persons collecting the data with the speeds of pilot vehicles with known speed appearing on the screen along with the regular traffic. S - 6.67, this was the average value of the values of standard deviation from different data sets. These values are shown in Table 6. Z - 1.96, assuming 95% confidence level 65 oooooooooooooooo ooooooooooooooooo ............ ooooooooooooooooo oooooooooooooooo oooooooooooo oooooooooooooooo oooooooooooooooo ooooooooooooo ooooooooooooooo ooooooooooooooo oooooooooooooooo oooooooooooooo oooooooooooooooo ooooooooooooooo ooooooooooooooo ooooooooooooooo ooooooooooooo ooooooooooooooo ooooooooooooooo oooooooooooooo oooooooooooooo .............. ooooooooooooo ooooooooooooo oooooooooooooo .............. ............. oooooooooooooo oooooooooooooo ooooooooooooo oooooooooooooo oooooooooooo oooooooooooooo ooooooooooooo ooooooooooooo oooooooooooo ooooooooooooo ooooooooooooo ccccccccccccc oooooooooooo ooooooooooooo I ........... OOOOOOOOOOOOO OOOOOOOOOOOO ............ ............ 000000000000 OOOOOOOOOOOO OOOOOOOOOOOO ........... ............ ........... OOOOOOOOOOOO IIIIIIIIIII OOOOOOOOOOO ........... ........... ........... ........... 0000000000 ........... .......... g r I d I I n e s °:3:3:3:3:3:3:1:3:3:1 OOOOOOOOOO .......... .......... ......... .......... ......... OOOOOOOOOO ......... 000000000 ......... ......... OOOOOOOOO ......... IIIIIIII ......... ........ ......... 00000000 ........ ........ ........ OOOOOOOO ........ ....... ........ ....... ........ ....... ....... ...... ....... ...... ....... ...... ...... ...... ...... ...... e lo c k —* o: la: 09: 20 Figure 11. Placement of Grid on the Screen TABLE 6. CALCULATION OF STANDARD DEVIATION 66 MION _J)_ATE TIME AVE . SPEED S , D , I 00 9/3/86 13:00 59.25 7.23 14:00 58.24 7.34 16:30 61.33 5.78 8:00 38.63 9.98 9:00 49.22 7.40 10:00 54.85 5.56 11:00 57.57 7.84 12:00 54.94 5.45 MI, ELLIOT 9/17/86 8:00 57.80 7.26 9:00 60.03 6.70 10:00 63.41 6.03 11:00 58.78 5.26 12:00 61.09 9.22 M LIOT 9/18/86 9:00 35.21 5.50 10:00 58.91 7.54 11:00 60.72 6.32 12:00 58.83 6.09 VAN D E 9/17/86 8:30 59.05 7.24 9:30 60.54 7.94 10:30 55.06 5.03 11:30 53.74 6.10 VAN YKE 9/18/86 8:30 27.58 5.51 9:30 53.15 6.11 10:30 61.24 6.76 11:30 58.17 5‘21 axe. S.D. -6.67 t1” an 15a: thl 6‘25 V0: 67 Using the above values in the equation gives n - 43 vehicles. 2. Empirical approach: a. A sample of volume (i.e., 15 vehicles) was chosen from the collected field data, and the average observed speed was plotted as indicated by point 1 in Figure 12. b. A second sample of the same size was selected, and the average speed of those two samples was plotted as point 2. c. This procedure was continued until a stable average speed (81 - 58.90) was reached at point 4. d. The volume associated with that value (i.e., 60 vehicles) was considered the sample size that will assure stable measures by students estimating the speeds. As a result of the two approaches, the sample size n for this study was taken as n - 60 vehicles per data set. 4.4.3 Data Collection To achieve balance in collecting the 60 vehicles and to satisfy the assumption of independence for the sample units (i.e., vehicles), an equal number of vehicles was taken from each lane of the freeway (i.e. , 20 veh. per lane). The choosing of vehicles in each lane was random. To avoid any bias in the calculated overall average speed of the three lanes of the freeway due to lane volume differences, the average speed of each lane was weighted according to the percentage of 'volume of that lane to the total volume on the freeway before calculating the overall average speed. 68 '° ’1 a: 5%} 80 .. 8 3 60 1. 31 C2J o A (55.90) 40 _ 20 - I I I I l >i 15 30 45 60 80%;": =9 Figure 12. Choosing the Sample Size 69 The data collected from the screen included: What lane the vehicle was in, time when the vehicle was at the first grid line (start time), time when the vehicle was at the second grid lixua (stop time), and type of vehicle. The vehicles were categorized in to four types (i.e., low performance passenger cars, high performance passenger cars, single unit trucks, and trailer trucks.) that INTRAS can simulate. The fifth type INTRAS can simulate, which is the intercity buses, is not simulated here because of its low percentage on the network (less than 1%). The difference between a low performance and a high performance passenger car is in the acceleration rate assigned to each type by the model. High performance cars are assumed to have higher acceleration rates (like sport cars). Distinguishing between these two type during screen data reduction was rather difficult, instead an assumed percentage (10%) of high performance cars was used in the model based on direct observation in the field. Table 7 illustrate the raw data collected for Mt. Elliot location on 9/18/86 between 10:00 and 10:10 a.m. 4.4.4 Calculating the Speeds a. A special Fortran program was used to read.the collected data and to convert it to speed data. Tables 8, and 9 show the Fortran program used for the data reduction, and the output file for the speed data at the Mt. Elliot location. b. The same procedure was used to calculate the speeds on all the selected locations. Table 10, and Figure 13 were prepared to summarize the results at all locations. 70 eoHHJm.ez Honequoa ooHoH HmzH» cm\mH\oo “mesa Hm.nu ma.om m H oo.mn mm.om n u mn.nH no.6: H H an.Hm mo.om u a we. oH nu.oouo H H mm.m me. He m H om.cH oo.mHHo H H no.nm co .nm m H om.om om.mu m H mm.Hn mm.on m H on.H¢ on. oe m H mo.HH Hm.ocuu H H on.on nm.om w H no.1H oo.u "n H H mo.¢n oo.nm m H cH.oo om.¢oum H n c¢.mc om.n¢ n H ca.nm n¢.m. u a mH.m¢ no.n¢ m H om.nH mn.u n H ma.oH me. ooHH H H n..nm mm.Hm m H oH.no 6H.ooH¢ H H Hm.mn mu m H n.m Ho.HHHn H H mn.oc mo.m¢ m a 51. H1 cm.om u H m¢.nm AH.¢m m H 1m. on 16.1n u a .no om.ceuo H m Hm.\1 mH.an m H m1.¢H ma.n u H «a.mo om.soum H m 11.1- mm.ea m H nm.oc Hc.mouo H H «1.9n su.mm m u om.no mH.ooHo H m 11.9: .H.o¢ m m “1.11 on.Hn m H n .me n¢.H¢ m m .om HH.on u H no. 10 sm.moHn H m oo.mc «5.69 m H «a.mo oH.noHn H H no.ne mo.H¢ m H oo.1c «a.monm H H :n.oe on.on m H m¢.nu u.uu m H Hm.nn AH.¢m m 1 um.AH \u.oH n H me. no Hm.noHH H H 11.mH nu.nH u H :3. 0H n1.ocue H m nH.om H¢.om n H ¢o. so 1.1oHn H m 1o.1u oo.¢u n H 6H.mH no.cH n H nm.cm o\.11 m H an.nH en.uH a H oo.wo mo.uoHc H m mzH» mzHe mzHe mzHe mzH» mzHe mapm emaem mzcu ma>e anew eaaem mzaJ u1>e acem emaem mzqa ma>e oHHm nun: eo>uonao uoHHHm .u: .h oHaua 71 Table 8. Data Reduction Fortran Program 00! .06 ( $0 '9 6. )6. 7. 22) 3|. .9. C. dd See .40 9V. 9‘. Stud-I .VMLVSIS ocean.“ "IOC‘IM (mil INIIGEI SLIOIII.(L1OOII.MIOOI.IIIOOOontleeI .E-Ot webbinawtmoou IEl-L 101.04».SIO.SSO.IM.!MJM.M.M6.MC,m,o~za,o=~.oc (mucus-I: Ian: OPEN! I..FlL(-'O£'fl0|" I 090“ 20.3 ILIG'CASEI ‘ . S'G'USI' 'NEU‘ I Item ".222th U“."\2..JIOIIM URIIE! 2...!” 00 SI- "1.0“ RE-ooc I|.$5$.ENO-99IM I I.SLIII.ELI I IJIIII J21 I I COHINUE IFI-I 00 IO. I’IJi leH I 11c.” 6010 6. VI lI-I ".../(InII-IIIIIII-IO...“3600.0/3280.” URI!“ 20.¢44I |.UIII.III II conmuc lI-l2 Hues-1‘ NW- 0 NI!“ -6 DO 7. l-I .IZ IF(SLIII.CO.2I new 6010 200 (L5: IHSLIIIJZOJI man 0010 3“ (NO If IOWTOUAOUI I I NIMIINOI COIO 7O VOID-101000! I I NU‘IO-NL'W0 I GOTO 7. VM- volt-w I I NtI1C-NUIC'I COHINUC own—1W Mime/111110 Govt-I out/111m: VIII-YMoTOUGOYM NWWONU‘IIOMHC can-twine" ”slussbflolh-wnoz CINIINUC SID-son" 550/: I I-I I I WIIEIZI.999IG~U.STD U‘HICI 20.9'IIW.WD.WC S'OP IOIauIcfvn Iowans-1.73.11 IOmsIleIII HImuIcII,II,II.Io.Iu .lNflo-IIII."- ',Io.)I Imus-01211.10.“ '.IJ.IN.‘~(. S'CtIh'.Ia.2.ltl.'VIn.Iu-I1 '.Iu lo~1s1lII/.251,'Avlu06( SPIIW'.':.2.II.251.'SN"'II SIU.- .‘oJ' "Imus-1.1540! I S'tto-‘J3.2.SI.'LAN£ 2 SPtttb'.FS.2.3-.'w~1( J Oé'IIOO' . 5.2! (NO 72 Table 9. Mt. Elliot Output Data File CARI 1 AVE. SPEED: 61 43 VEH.TYPE: 2 CARI 2 AVE. SPEED: 61.98 VEH.TYPE: I CARI 3 AVE. SPEED: 61.98 VEH.TYPE: l CARI 4 AVE. SPEED: 55.89 VEH.TYPE: l CARI 5 AVE. SPEED: 58.28 VEH.TYPE: 1 CAR! 6 AVE. SPEED: 63.72 VEH.TYPE: 1 CARI 7 AVE. SPEED: 56.82 VEH.TYPE: 2 CARI 8 AVE. SPEED: 60.34 VEH.TYPE: l CARI 9 AVE. SPEED: 70.29 VEH.TYPE: 2 CAR" 10 AVE. SPEED: 50.88 VEN.TYPE: 2 CARI 11 AVE. SPEED: 68 18 VEH.TYPE: 1 CAR" 12 AVE. SPEED: 68 18 VEH.TYPE: l CARI 13 AVE. SPEED: 53 69 VEH.TYPE3 2 CARI 14 AVE. SPEED: 64.94 VEH.TYPE: l CARI 15 AVE. SPEED: 68.18 VEH.TYPE: 1 CAR" 18 AVE. SPEED: 58.28 VEH.TYFE: 1 CARD 17 AVE. SPEED: 50.88 VEH.TYPE: l CARI 18 AVE. SPEED: 72.53 VEH.TYPE: 2 CARI 19 AVE. SPEED: 50.88 VEH.TYPE: 1 CAR! 20 AVE. SPEED: 52 45 VEH.TYPE: 1 CAR! 21 AVE. SPEED: 61.43 VEH.TYPE: 1 CAR“ 22 AVE. SPEED: 49.77 VEH.TYPE: 2 CARI 23 AVE. SPEED: 59.81 VEH.TYPE: 1 CAR! 24 AVE. SPEED: 65.56 'VEH.TYFE' 4 CARI 25 AVE. SPEED: 37.06 VEH.TYPE 1 CAR! 28 AVE. SPEED: 54.11 VEH.TYPE 1 CAR! 27 AVE. SPEED: 61.98 VEH.TYPE: 2 CAR” 28 AVE. SPEED: 55 43 VEH.TYPE 2 CAR! 29 AVE. SPEED: 61.98 VEH.TYPE l CARI 30 AVE. SPEED: 66.20 VEH.TYPE: 1 CAR! 31 AVE. SPEED: 55.43 VEH.TYPE 2 CAR! 32 AVE. SPEED: 60.34 VEU.TYPE 1 CAR! 33 AVE. SPEED: 66.20 VEH.TYPE l CARO 34 AVE. SPEED: 54.99 VEU.TYPE l CARI 35 AVE. SPEED: 55.43 VEH.TYPE 2 CAR! 36 AVE. SPEED: 60.34 VEH.TYPE 1 CAR! 37 AVE SPEED: 53.69 VEH.TYPE 1 CAR“ 38 AVE SPEED: 61 98 VEH.TYPE 1 CAR! 39 AVE. SPEED: 63 72 VEH.TYPE l CARI 40 AVE. SPEED: 55 43 VEH.TYPE 1 CARI 41 AVE. SPEED: 55 43 VEH.TYPE 2 CAR! 42 AVE. SPEED: 61.98 VEH.TYPE 2 CARI 43 AVE. SPEED: 72.53 VEH.TYPE 1 CAR! 44 AVE. SPEED: 63.72 VEH.TYPE 1 CAR! 45 AVE. SPEED: 58.78 VEH.TYPE 1 CAR“ 46 AVE. SPEED: 60.34 VEH.TYPE 1 CAR“ 47 AVE. SPEED: 61.43 VEH.TYPE 1 CAR! 48 AVE. SPEED: 60.34 VEH.TYPE 1 CAR! 49 AVE. SPEED: 54.99 VEH.TYPE 1 CAR! 50 AVE. SPEED: 58.28 VEH.TYPE 2 CAR” 51 AVE. SPEED: 53.27 VEH.TYPE 1 CAR! 52 AVE. SPEED: 49.77 VEH.TYPE 1 CAR! 53 AVE. SPEED: 56.82 VEH.TYPE 1 CAR! 54 AVE. SPEED: 60.34 VEH.TYPE. l CARI 55 AVE. SPEED: 37.88 VEH.TYPE. 2 CAR! 56 AVE. SPEED: 54.99 VEH.TYPE: l CARI 57 AVE. SPEED: 66.20 VEH.TYPE 1 CAR” 58 AVE. SPEED: 48.70 VEH.TYPE 1 CAR“ 59 AVE. SPEED: 61.98 VEH.TYPE 2 CAR” 60 AVE. SPEED: 66 20 VEH.TYPE 1 AVERAGE SPEED=58.91 SAMPLE STD.= 6.99 LANE 2 SPEED=59.17 LANE 3 SPEED=63.87 LANE l SPEED=53.69 73 TABLE 10. SUMMARY OF AVERAGE SPEEDS AND VOLUMES LOCATION DATE TIME VOLUME (VPH) AVE. SPEED RAMP EFREEWAY* (MPH)** METER LINWOOD 9/3/86 8:00 306 5736 38.63 ON 9:00 240 5166 49.22 ON 10:00 198 4266 54.85 ON 11:00 186 4170 57.57 ON 12:00 252 4236 54.94 ON 1:00 210 4440 59.25 ON 2:00 222 4674 58.24 ON 4:00 120 4812 61.33 ON LINWOOD 9/10/86 8:00 318 6138 31.05 OFF 9:00 150 5148 53.20 OFF 10:00 138 4536 56.18 OFF 11:00 216 4028 58.19 OFF 12:00 210 4356 57.39 OFF 1:00 210 4182 $9.50 OFF 2:00 186 4578 56.39 OFF 4:00 126 5346 54.80 OFF VAN DYKE 9/17/86 8:30 413 5478 59.05 ON 9:30 377 4884 60.54 ON 10:30 372 3816 55.06 ON 11:30 468 3690 53.74 ON VAN DYKE 9/18/86 8:30 --- 5394 27.58 OFF 9:30 432 4782 53.15 OFF 10:30 432 4068 61.24 OFF 11:30 456 3703 58.17 OFF MT. ELLIOT 9/17/86 9:00 408 5370 60.03 ON 10:00 336 3978 63.41 ON 11:00 414 3894 58.78 ON 12:00 414 3960 61.09 ON 2:00 486 4248 56.99 ON 3:00 504 5106 56.12 ON 4:00 606 4440 60.98 ON MT. ELLIOT 9/18/86 9:00 396 5670 35.21 OFF 10:00 306 4002 58.91 OFF 11:00 348 4032 60.72 OFF 12:00 444 3852 58.83 OFF 2:00 576 4374 58.23 OFF 3:00 732 5406 53.02 OFF 4:00 438 4368 58.81 OFF 74 TABLE 10 - CONTINUE JOHN R. 10/15/86 3:30 --- 5072 25.30 ON 4:00 --- 4912 25.06 ON 4:30 --- 5036 25.21 ON JOHN R. 10/16/86 3:30 --- 5264 25.62 OFF 4:00 --- 5112 32.53 OFF 4:30 --- 4632 25.51 OFF TRUMBULL 10/28/86 3:30 547 5509 35.60 ON 4:30 663 5294 36.38 ON 5:30 624 5169 40.60 ON TRUMBULL 10/29/86 3:30 600 5500 38.10 OFF 4:30 653 5649 38.38 OFF 5:30 624 5306 37.08 OFF GRATIOT 11/28/86 3:30 240 6204 58.27 ON 4:00 290 6236 58.98 ON 4:30 320 6256 46.95 ON GRATIOT 11/29/86 3:30 200 6064 63.19 OFF 4:00 270 6580 62.77 OFF 4:30 250 6508 58.64 OFF MT. ELLIOT 4/16/87 7:30 720 6552 37.84 OFF 8:00 492 6528 39.39 OFF 8:30 492 6840 48.01 OFF 9:00 324 5412 63.57 OFF 9:30 384 4512 65.54 OFF 10:00 456 4428 63.75 OFF 10:30 408 4704 64.06 OFF 11:00 408 4104 64.05 OFF 11:30 372 3972 66.79 OFF 12:00 384 4188 64.38 OFF 1:00 252 4260 66.70 OFF 1:30 288 4536 66.20 OFF 2:00 384 4776 63.65 OFF 2:30 1488 5712 47.36 OFF * — This is the volume on the freeway after the on-ramp, so it includes the on-ramp volume. ** - This is the observed speed in the merging area. between the ramp gore (i.e., the start of the acceleration lane) and filmed location varies for each sampled location. The distance 75 SPEED LINWOOD RAMP 60 x O 9" o a) 10 O O i 0 40 x 'o 20 0 4000 5000 6000 M FngB-a SPEED MT.ELLIOT RAMP go a: x, ( X g x X 0 40 O 20 X - metering on O- metering on O 4000 5000 6000 VOLUME “9.84) Figure 13. Speed-Volume Plots for Sampled Locautions 76 FIGURE 13 . CONTINUE SPEED. ' VAN OYKE RAMP 60 O O X x x ‘ o 40 O 20 O 3000 4000 5000 VOLUME Fly 43.: 938i) ’ JOHN R RAMP 60 40 O O 1 XX 0 20 0 4000 5000 6000 M FngLd 77 IUEMRE 13-‘mnuuuus SPEED GRA'I IOT RAMP 20 5000 6000 7000 M; F I g . I34 9%20 TRUMBULL RAMP SIX) 601) WIN) ggygg Fig.8-! CHAPTER 5 CALIBRATION AND VALIDATION Calibrating any model requires the analyst to make several runs for one set of data (i.e., one time sample at one location), and for each single run a chosen parameter(s) will be given a new value(s) until the difference between the MOE from the model and the observed MOE from the field data becomes statistically insignificant. The MOE that was compared in this study was the average speed. 5.1 THE CALIBRATION DATA In this study, the model was calibrated using the volume, vehicle mix, and speed data collected at the MT. Elliot location on the west- bound Ford Freeway on April 16, 1987. The sub-network that was used to calibrate the model, Figure 14, contains the stretch of westbound Ford Freeway that starts just upstream of the Mt. Elliot entrance ramp and ends after the merging area that follows the ramp. The merging area is the area of interest in this study, and it is the area where the field data were collected. Figure 15 shows the volume-speed curve for this data (based on 14 time points). This curve is very similar in shape to the 78 79 xuofioz-€m SSE .0: .3 SEE xv‘EMMmNR Brae. Qua 80 03:» .0: you «>08 09?».88... .2 .53.... 2a.: oc::o> ooon oomo oooo oomm ooom oom¢ 000* F P _ p h _ F \\ \ \ \ \ \ «060d \ aflwfl a... o \ \\ cw \ ‘ oomn Tow [om Ton filo¢ Tom Too (lld') paeds / Ton 81 classic volume-speed curve in the 1965 Highway Capacity Manual which was superimposed over the same figure. The main differences between the two curves are that in the Detroit data the traffic stream maintains peak speed over a wider range of volumes, and higher peak speeds in Detroit data. This indicated that the data are similar to the general nationwide traffic behavior, but have some special characteristics (specifically driver behavior in Detroit seems to be more aggressive than "average" behavior). This means that some of the default parameter values for INTRAS (which was calibrated to fit the general traffic behavior) need to be changed so the model can replicate the Detroit data. After coding the sub-network and loading it into the model, the model was run using the observed volume and vehicle mix data. 5.2 CHOOSING THE APPROPRIATE VARIABLES The first step in calibrating INTRAS was to test the different calibration variables available and choose the appropriate values that will cause the model to simulate the Detroit data with acceptable accuracy. This was done by testing one variable at a time and comparing the effect of each variable on the performance of the model (i.e., the resulting average speed at each time point). Testing the variables was done by changing their embedded values in the model. This process was done externally (i.e., without the need to recompile the model every time a change is made), since INTRAS allows a change of the embedded values of the calibration variables, through special input cards. 82 Some of the calibration variables were not tested because they deal with the movement of vehicles on the surface links (which was not the main concern here). The first variable to be tested was the acceptable lag in the target lane which determines if vehicles can change lanes. Different values were applied to that lag and tested. The embedded value for this variable is 3.1 seconds, and 8 new values were tested (using card type 35). These values were 2.5, 2.7, 2.9, 3.3, 3.5, 3.7, 3.9, and 4.1 seconds. A total of 126 computer runs were executed (i.e., 9 for each of the 14 time points), but the response of the model to the new values was not significant over the default value in simulating the observed speeds, so the default value was kept in the model. Amount of time needed to complete a lane-change maneuver was the second variable to be tested. The embedded value for this variable is 3 seconds. Since the field data indicated an aggressive driver's behavior, the new values tested (using card type 49) were 2 seconds and 1 second. A total of 42 computer runs were executed (i.e., 3 for each time point), but the response of the model to the new values was not significant over the default value in simulating the observed speeds, so the default value was kept in the model. "As each vehicle enters a link, it is assigned a free-flow speed. This is obtained by multipLying a percentage by the free-flow speed specified for that link. This percentage is obtained from a decile distribution." (Wicks and Andrews, 1980). This decile distribution was the third variable tested (using card type 40). The default assigned percentages (which should always add up to 1000) are: 83 I l 2. .3. A 2 .6. l 2 2 .LQ % values 75 81 91 94 97 100 107 111 117 127 where I is the driver type index. INTRAS defines 10 types of drivers on the road ranging from very aggressive driver (type 10) to timid driver (type 1). Since Detroit data are to the aggressive side, five percentage distributions that gave higher percentage values to the more aggressive driver types were tested. The following is an example of one of the distributions tested: I 1 .2. 2 a 2 2 l 2 2 .12 % values 74 77 86 90 92 103 110 116 122 130 A total of 70 computer runs were executed (i.e., S for each time point), but the response of the model to the new values was not significant over the default value in simulating the observed speeds, so the default value was kept in the model. Card type 40 was also used to test the fourth calibration variable which is the percentage of mean speed by driver type I on freeway links. The default assigned percentages (which should always add up to 1000) are: I .1. 2 2 9. 2 Q l 2 2 m % values 82 91 94 97 99 101 103 106 109 118 As in the case of the last variable, five percentage distributions that gave higher percentage values to the more aggressive driver types were tested. The following is an example of one of the distributions tested: 1 l .2. 2 2 2 .6. l 2 2 IO. % values 78 87 9O 93 95 105 107 110 113 122 84 A total of 70 computer runs were executed (i.e., 5 for each time point), but the response of the model to the new values was not significant over the default value in simulating the observed speeds, so the default value was kept in the model. The fifth and sixth variables tested variables (i.e., the sensitivity factors and the free-flow speed) were found to be the best variables to cause the model to simulate the Detroit data with acceptable accuracy, and they will be discussed in more detail in the next section. 5.3 THE CAR FOLLOWING MODEL The main formula in the INTRAS model that was focused on during the calibration is the "car following model", (see section 3.6.1). This model calculates and defines the acceleration of the following car depending on the relative locations and speeds of the two cars (i.e., the leading car and the following car), and type of driver of each car. This formula also contains an array of car-following parameters that relates to the "driver's sensitivity". The input values for this array can be changed externally by changing the values assigned to type of driver in card 43. The values that can be assigned to the parameters (i.e., sensitivity factor k in the equation in section 3.6.1) through this card range from O to 99. The smaller the value of the parameter, the more aggressive the drivers are assumed to be. The default values of the sensitivity factors (SF) are as follows: I 1 2. 2 a 2 .6. l 2 2 m SF 15 l4 13 12 ll 10 9 8 7 6 85 Ten sets with ten different values for the sensitivity factors were tested. The first set (S#l) had the default values shown earlier, the second set (S#2) had very low values (i.e., from SF- 9 for I- 1 down to SF- 0 for I- 10), which should reflect a very aggressive behavior, the third set (S#3) had values on the higher side (i.e., from SF- 30 for I- 1 down to SF- 21 for I- 10), and the rest of the sets had intermediate values between those two extremes. Table 11 shows the values that were used in each set. As stated earlier, the lower the values of the parameters the more aggressive the drivers are assumed to be. So, six of the ten sets were built with values lower than the embedded values to cover all the possible values in the lower side. The remaining three sets were built with values higher than the embedded values to observe the performance of the model on that side (which was not expected to give good results for Detroit data). The free-flow speed on the freeway links can have any value up to 99 mph, but the maximum speed on the ramp links is 67 mph. The value of the free-flow speed is assigned to each link via card type 02. Since the field data collected in Detroit indicated high speeds during the off-peak hours, the speed values that were tested on the different links of the network were on the high side. Five different sets of free-flow speed values were tested: 1 2 3 4 5 Freeway links: 65 7O 75 75 8O Ramp links: 50 55 55 65 65 Surface links: 45 45 50 50 55 86 TABLE 11. SENSITIVITY EACTOR.VALUES Set I 1 2 3 4 5 6 7 8 9 10 # S#l 15 14 13 12 ll 10 9 8 7 6 S#2 9 8 7 6 5 4 3 2 l O S#3 30 29 28 27 26 25 24 23 22 21 S#4 10 9 8 7 6 5 4 3 2 l S#5 ll 10 9 8 7 6 5 4 3 2 S#6 12 ll 10 9 8 7 6 5 4 3 S#7 13 12 ll 10 9 8 7 6 5 4 S#8 l4 13 12 11 10 9 8 7 6 5 S#9 l9 18 17 16 15 14 13 12 ll 10 S#10 21 20 19 18 l7 16 15 l4 13 12 87 The procedure that was used to select the best combination of SF and free-flow values is described in the next section. 5.4 CALIBRATION PROCEDURE 5.4.1 The Computer Runs 50 runs were executed for each time point to cover all the possible combinations of ten sets of SF and five sets of free-flow speeds. Since there are 14 time point in the calibration data, A total of 700 computer runs were executed in the process of selecting the best combination of SF and free-flow speeds. This was done by using the same sub-network for Mt. Elliot but the hourly volume and the vehicle mix were adjusted for each data point. Paired comparisons were performed between the observed data and each of the fifty sets, and the combination that gave the smallest average difference was composed of S#5 and speed set 3. Table 12 shows the results of 140 runs, or ten sets, with the average difference and the standard deviation at the bottom of each column. All the shown sets were executed with speed set 3, but each set has different SF values. 5.4.2 The Significance Level Test The rest of the calibration procedure will be focused on the set that were found to have the most favorable effect on the results (i.e., S#5, and speed set 3). Figure 16 shows the model speeds of this set and the observed speeds. 88 TABLE 12. OBSERVED AND MODEL SPEEDS FOR MT. ELLIOT T.P. O.S. S#l S#2 S#3 S#4 S#5 S#6 S#7 S#8 S#9 S#lO l 37.84 45.7 46.0 40.1 45.8 43.3 44.8 44.3 45.6 45. 42.6 2 39.39 40.4 35.3 34.7 38.8 43.5 40.0 41.7 40.9 38. 37.5 3 48.01 41.6 45.5 36.0 47.0 48.0 46.5 45.3 45.2 40. 37.0 4 63.57 50.1 51.7 40.3 49.3 55.6 56.2 54.0 53.5 36. 42.3 5 65.54 53.0 50.9 35.3 51.7 54.2 57.0 53.8 53.8 48. 36.7 6 63.75 52.9 52.7 32.2 53.2 55.7 55.2 54.2 56.7 51. 35.2 7 64.06 53.8 49.8 33.5 50.8 58.4 57.6 56.4 56.2 47. 34.2 8 64.05 52.9 53.7 48.7 53.8 60.9 60.5 54.3 59.3 46. 49.5 9 66.79 57.7 57.0 53.9 55.9 65.0 62.3 59.6 58.0 54. 54.3 10 64.38 57.0 55.1 49.9 55.3 61.0 60.2 56.7 56.1 54. 52.5 11 66.70 52.9 54.7 48.5 56.6 59.3 60.9 54.5 55.0 53. 50.8 12 66.20 56.2 55.2 50.1 55.5 62.8 59.7 58.7 60.2 53. 51.5 13 63.65 55.3 53.6 37.5 51.9 63.4 62.0 59.9 58.3 52. 40.5 14 47.36 46.3 46.8 44.6 47.3 51.3 48.9 48.8 48.3 45.] 45.9 Ave. D - 7.53 8.09 16.86 7.74 2.78 3.54 5.65 5.30 10.8 15.06 S.D- - 6.16 6.26 10.67 6.61 5.07 4.38 5.64 5.54 8.56 10.67 T.P.: Time Point, 0.8.: Observed Speed, Ave. D: Average Difference, S.D.: Standard Deviation 89 Since the sample size is small, the model output speeds were tested against the observed speeds using a paired comparison and a t- test. The pairs were in the form Di - Si - Ci, and the null hypotheses was: Ho : 6 - 0 and, H1 : 6 # 0 where: 6 - E(Di) - E(Si-Ci). The main assumption involved is that the paired differences Di's constitute a random sample from a normal population N(6,a ). All the differences are shown in Table 13. D - E Di/n and Sd - [E(Di-D) ]/(n-1) Which gave the following results for this sub-network: D - i 2.782 mph and Sd - 5.072 A 95 % confidence interval (CI) for 6 is given by the equation: 0 i (ca/2 * Sd)/ j—E Which gives: CI - (-O.13 L +5.69) Where: t is based on d.f.= 13. The t table gives to - 2.145 a/2 /2 9O uoaaau .0: you «cocoa Hana: ecu uo>uoaao .SH «Hausa we: ¢ _. n p N F F _. or o o h — F t _ b k L f E L L F 0 ocean 602030 3 ocean _oooE o I o _. 1 oN T on I . 10¢ O O I 1 on O O O . . . Tom I O I I I I 2 P: a a a M Fob paeds 91 TABLE 13. PAIRED COMPARISON OF SELECTED PARAMETERS TIME OBSERVED SPEED MODEL SPEED(S#5) DIFFERENCE Si - mph 01 - mph. 01 - Si-Ci 7 30 37 84 43 3 -5 46 8 00 39 39 43 5 -4 11 8:30 48.01 48.0 0.01 9 30 63 57 55 6 7 97 9 30 65 54 54.2 11 34 10 00 63.75 55.7 8.05 10 30 64.06 58.4 5.66 11:00 64.05 60.9 3.15 11 30 66.79 65.0 1.79 12:00 64.38 61.0 3.38 1 00 66 70 59 3 7 40 1 3o 66 20 62 8 3 40 2 00 63 65 63 4 o 25 2 3o 47 36 51 3 -3 94 D - 2.782 Sd - 5.072 92 A test of Ho : 6 - 0 is based on the statistic test: t - D / [ Sd/JE'] , d.f. - n-l That gave: t - 2.052 which is smaller than to - 2.145 /2 The calibration was considered successful when the null hypotheses passed the t-test, this was when the calculated value of t became insignificant (i.e., smaller than the tabulated t value). And at least half of the individual points passed the CI test, as shown in Figure 17. Which means that the model, with the new parameter values, is ready to be used to replicate the traffic behavior in Detroit with an acceptable margin of error. The calibration process was most effective in a speed range of 48-68 mph. A sample of the model outputs (i.e., the results of the simulated data at 7:30 a.m.) for the calibration sub-network at Mt. Elliot can 'be found in appendix B. 5. 5 THE VALIDATION The intent of validation is to run the model (in this case INTRAS) with different data than the data used for calibration, but without changing the final values of the calibrated parameters on card ‘43‘- The validation data can be from the same location (i.e., Mt. Elliot), or other locations on the Ford Freeway. 93 anon. HIPHQEH saucepan—woo Gnu. .hH Guam: m2: 818 . 8a mafia" o BONBHdeIO 94 The validation of the model consists of five parts as follows: a. The same location (i.e., Mt. Elliot) but with different data from a different date (i.e., 9/18/86) with the ramp metering off; b. The same location with different data from a different date (i.e., 9/17/86) with the ramp metering on; c. A new location (i.e., Cratiot) on the east direction of the Ford Freeway, with ramp metering off; d. A new location (i.e., Trumbull) on the west direction of the Ford Freeway, with ramp metering off; and e. A new location (i.e., Van Dyke) on the west direction of the Ford Freeway, with ramp metering on. Several data points for each of the five parts above were loaded to the model each as a sub-network. The model was run for each sub- network and the results of those runs are shown in Table 14. The model output speeds in each of the five parts compared very well to the field as can be seen also in Table 14. In almost all cases the difference between the observed speed and the model speed passed the CI test. The validation process gave good results when the speeds were in the range of 38-64 mph. Which is close to the calibration range. 95 TABLE 14. THE VALIDATION RESULTS 1955119! 25:; TI OBSERVED SPEEDS MODEL SPEEDS METERS 211.2LLIQI 9/18/86 9:00 35.21 40.00 OFF 10:00 58.91 61.90 OFF 3:00 53.02 $3.70 OFF MT, ELLIQI 9/17/86 9:00 60.03 46.20 ON 10 00 63.41 64.91 ON 3:00 56.12 55.90 ON 0351101 11/19/86 3:30 63.19 61.10 OFF 4:00 62.77 60.20 OFF 4:30 58.64 57.70 OFF TRUMBULL 10/29/86 3:30 38.10 39.30 OFF 4:30 38.38 39.70 OFF 2AN_DXKE 9/17/86 8:30 59.05 62.4 ON 9:30 60.54 64.8 ON CHAPTER 6 APPLYING THE CONTROL STRATEGIES Following the calibration and the validation of the model, the model is ready for simulating the whole network under study (i.e., the Ford Freeway inside Detroit city limits) in one run. Therefore, the whole network was coded in two files (i.e., each direction on a file) and made ready to run after solving some problems with the INTRAS user's manual. (see Appendix A for the coded network) 6.1 PROBLEMS WITH INTRAS USER'S MANUAL The attempt to run the entire one direction network was not easy because some INTRAS features did not operate in the manner described in INTRAS User's Manual, Wicks and Andrews, (1980). Three major problems in the user's manual were found during this research. First, the model, although it would accept a left-hand off— ramp, did not accept a left-hand on-ramp while the user's manual stated that "auxiliary lanes may occur on either the left or right hand side of the roadway." (Wicks and Andrews, 1980). The on-ramp from southbound Lodge Freeway on to the Ford Freeway is the only left- hand on-ramp and it had to be coded as a right-hand on-ramp to be accepted by the model. 96 97 Second, the user's manual gives specification to where the early warning, for exiting vehicles, should be located. "It (means the warning sign) must be positioned downstream of the previous off-ramp and upstream of the freeway link connecting directly to the specific off-ramp." (Wicks and Andrews, 1980). This was found not always accurate, at least not a must, in the sense that the model gave better results when some of the early warning signs where located upstream of the previous off-ramp. Third, the user's manual has a section titled Size Modification Procedures, this is to help the user change the capacities of model variables in case this is needed. In this project, there was a need to increase the maximum number of nodes, and the steps that were given in the user's manual for making that change was followed (i.e., changing the value of NTOTN in BLOCK DATA INTVAR, and changing the sizes of SIGI array in COMMON /A3/, SIG array in COMMON /A6/, and SNODE array in COMMON /A7/) but the model did not respond accurately. After checking the variable and array lists in the model four more arrays that needed to be changed, were found. Three of those arrays (IORG, IRV, and IREN) are located in COMMON /ONVEH/, and the fourth array (NACT) is located in COMMON /ACTlO/. 6.2 EVALUATING THE PRESENT CONTROL STRATEGY The control method used to run INTRAS is called "Clock Time Metering". To simulate clock-time control of on-ramps, one fixed metering rate (vehicles per minute) is specified at each node. A count down clock is assigned to each associated on-ramp and the signal 98 is set to "green" until a vehicle is discharged, and is then set to "red" (Wicks and Andrews, 1980). The evaluation procedure was conducted on the East Bound Ford Freeway for one peak hour as follows: 6.2.1 The Basic Run The first run of the network was done with the ramp metering off. It was considered the basis of comparing the do-nothing strategy (ST#1) with the present control strategy and the suggested control strategies. This was done to define the benefits of the ramp metéring strategy in terms of the MOEs of concern [i.e., average speed on the freeway, average speed of the whole system (including ramps and surface links), total delay, delay on the ramps and surface links, total vehicle-time, total vehicle-miles, and moving/total time]. 6.2.2 Applying the Present Control Strategy The second run was designed to test the operating plan currently used in this corridor (ST#2). The present metering rate is 15 vehicles per minute (i.e., 1 veh./4 sec). This rate was simulated on each ramp on the east direction of the freeway and the model was run for that direction. 6.2.3 Discussion of Results The results for the peak hour for both runs are presented in Table 15, where the significant benefits of the control strategy ST#2 can be clearly noticed. The increase of the average speed of the corridor is 8 % , the reduction in total delay is over 17 %, and about 99 TABLE 15. COMPARING MEASURES OF EFFECTIVENESS FOR ONE PEAK.HOUR (NO-METERING VS. PRESENT STRATEGY, ALL.NETWORK) M.O.E. NO-METERING PRESENT METERING Difference % ST#1 STRATEGY— ST#2 Vehicle-miles 73938.41 73852.29 - 86.12 -0. Vehicle-minutes 105051.02 97168.53 -7882.49 -7. Moving/Total trip 0.577 0.622 +0.045 +7. time Travel Time(min)/ 1.42 1.32 -0.10 -7. Veh-mile Speed mph 42.23 45.60 +3.37 +8. Total Delay 44399.56 36717.66 -7681.90 -17. (Veh-min) Delay Time(min)/ 0.60 0.50 - 0.10 -16. veh-mile 100 7,900 vehicle-minutes were saved in one hour. This demonstrated clearly the effect of ramp metering in increasing the efficiency of flow on the freeway. 6.3 TESTING NEW STRATEGIES AT LOCAL.ON RAMPS After determining the benefits of the present strategy, other metering strategies were tested to determine the metering rate that will maximize the benefits (i.e., increase the speeds and reduce delays). 6.3.1 Applying the Strategies The first step was to apply uniform metering rates to all the local on-ramps. The rates that were tested were: 5, 6, 7, and 8 second headway on all ramps. The first new strategy ST#3 (5 sec. headway) showed a minimal change in results from the present strategy on the freeway (i.e., ST#2) on the freeway and the freeway corridor overall, but the average speed on the ramps and surface links dropped . The second new strategy ST#4 (6 sec. headway) showed better results on both the freeway and the freeway corridor. The third new strategy ST#5 (7 sec. headway) showed further improvement of speeds on the freeway but the average speed on the freeway corridor was reduced as a result of the long queues on some of the heavy volume ramps. The fourth new strategy ST#6 (8 sec. headway) crashed in the computer because the length of some of the ramp queues exceeded the length of those ramps and the simulation was aborted. The results of all runs are shown in Table 16, 17, and 18. TABLE 16. COMPARING MOEs OF DIFFERENT METERING STRATEGIES AT 101 LOCAL ON-RAMPS ONLY (ALLLEEIEQRK) Strategy Veh-Minute Total Delay Change in Speed Change in # min min Delay mph Speed ST#1 No-Metering 105051 44399 0.0% 42.23 0.0% ST#2 4 sec. headway 97168 36717 -17.3% 45.60 8.0% ST#3 5 sec. headway 97206 37303 -16.0% 45.19 7.0% ST#4 6 sec. headway 93524 32921 -25.9% 47.62 12.8% ST#5 7 sec. headway 99795 39447 -ll.2% 44.99 6.5% ST#6 8 sec. headway crashed ST#7 6 sec. on ramps 91958 31782 -28.4% 47.98 13.6% w/volume >400 vph 102 TABLE 17. COMPARING MOEs OF DIFFERENT METERING STRATEGIES AT LOCAL.ON-RAMPS ONLY (FREEWAY ONLY) Strategy Veh-Minute Volume Density Speed Change # min veh/ln/hr veh/ln-mile mph in speed ST#1 No-Metering 101410.86 1571 37.2 42.20 0.0% ST#2 4 sec. headway 93466.49 1569 34.3 45.70 8.3% ST#3 5 sec. headway 92750.24 1555 34.0 45.70 8.3% ST#4 6 sec. headway 88798.58 1578 32.6 48.40 14.7% ST#5 7 sec. headway 85702.91 1579 31.5 50.20 18.9% ST#6 8 sec. headway crashed ST#7 6 sec. on ramps 87214.52 1562 32.0 48.80 15.6% w/volume > 400 vph 103 TABLE 18. COMPARING MOEs OF DIFFERENT METERING STRATEGIES AT LOCAL DIV-RAMPS ONLY (W) Strategy Veh-Minute Moving/Total Speed Change # min Time m in S eed ST#1 No-Metering 3648.05 0.82 43.00 0.0% ST#2 4 sec. headway 3702.04 0.80 42.10 -2.0% ST#3 5 sec. headway 4455.93 0.66 34.90 -18.8% ST#4 6 sec. headway 4726.18 0.62 32.90 -23.5% ST#5 7 sec. headway 14086.00 0.25 13.30 -69.0% ST#7 6 sec. on ramps 4743.78 0.62 32.80 -23.7% w/volume > 400 vph 104 To fine tune the model results, modifications were next made to the best defined strategy (i.e., ST#4). By observing the speed on each link separately and the change of speed between successive links, in the next run ST#7 the ramps that have less demand than 400 hourly volume (i.e., 6 on-ramps on the west direction) were not metered because the change in speed between the links before and the links after the those ramps was not significant with ramp metering than without ramp metering. The model was run and the metering rate for the 9 metered ramps was set to 6 seconds. The results of ST#7 showed further improvement on both the freeway and the freeway corridor as also shown in Table 16. Figure 18 shows curves of change in average speed among the different strategies on three levels: the freeway corridor, the freeway only, and the ramp and surface links only. 6.3.2 Discussion of Results Both Table 16 and Figure 18 show the last strategy ST#7 as the best strategy to be used in case of the Ford Freeway in Detroit. The results indicate that this strategy will maximize the benefits of the system for the strategies tested. For example the increase of the average speed over ST#1 is about 14 %, the reduction in total delay is over 28 %, and the anticipated saving in time is about 13,000 vehicle-minutes per peak hour. Also shown in Figure 18 the change in speed for both the freeway alone and the ramp-surface nodes alone. SPEED Onph) 60 50— 40— 30-4 20-+ 10— L 105 1- ST#4 2- ST#7 \ \ \ \ \ \ \ \ \ \ \ \ a—a SURFACE ' H FREEWAY ONLY o—o ALL NETWORK l l 4T I l ST#1 ST#2 ST#3 ST#4 ST#5 STRATEGY Figure 18. Comparing Average Speeds of Different Metering Strategies at Local On-Ranps Only 106 The constant improvement in speed on the freeway alone when the headway gets longer is clear, and also the sharp decrease in the speed of the ramp-surface roads. For example, the speed on the freeway increased from 48.4 mph (77.44 km/h) in ST#4 to 50.2 mph (80.32 km/h) in ST#5. But at the same time the speed on the ramp-surface roads decreased from 32.9 mph (52.64 km/h) in ST#4 to 13.3 mph (21.28 km/h) in ST#5, and that caused the decrease in the average speed of the whole corridor when the headway changed from 6 seconds to 7 seconds. 6.4 TESTING NEW STRATEGIES AT LOCAL.AND FREEWAY ON-RAMPS The next step after defining the best metering strategy to be applied at the local (non-freeway) on-ramps was to define the benefits of metering the on-ramp part of the freeway-to-freeway interchanges that connect the Ford Freeway with three freeways (i.e., the Jeffries, the Lodge, and the Chrysler) in the city of Detroit. While testing the control strategies at freeway on-ramps, the metering rates of local on-ramps were kept the same as the rates that gave the best benefits in case of metering local on-ramps only (i.e., in ST#7). 6.4.1 Applying the Strategies w/ Existing Geometry Three new runs were made with three different metering rates applied to the freeway on-ramps. The rates were 6 (ST#8), 5 (ST#9), and 4 second headway (ST#10) respectively. The first two runs ST#8 and ST#9 were crashed in the computer because the length of the queues on some of the freeway on-ramps were longer than the link length and 107 the simulation was aborted. The results of the third run (i.e., ST#10) are presented on Tables 19, 20, and 21. The critical ramp that caused the first two runs to crash is the on-ramp from the North Bound Lodge Freeway because of both the high traffic volume and the short storage space. To solve this problem without changing the existing geometry the model was run with all the controlled on-ramps with 6 second headway except North Bound Lodge on-ramp with 4 second headway (ST#ll). This was done to allow more vehicles to enter the freeway and reduce the storage space needed. The results of this strategy ST#11 are also presented in Tables 19, 20, and 21. 6.4.2 Applying the Strategies w/ MOdified Geometry The results of ST#11 did not reflect any improvement over the results of ST#10, so the next step was to keep the same rates as in ST#10 and modify the geometry of the North Bound Lodge on-ramp. This was done by increasing the length of the surface link before the metering signal on that ramp to accommodate more vehicles. This modification represents in the real world either increasing the length of that interchange leg or adding a second lane to the interchange leg. The results of this strategy are also presented in Tables 19, 20, and 21. 108 TABLE 19. Comparing MOEs of Freeway-to-Freeway Control Strategies (All Network) Strategy Veh- Total Delay Change in Speed Change in # MinutL mLin M mph Sat-2.31 ST#1 No-Metering 105051 44399 0.0% 42.23 0.0% ST#7 6 sec. on ramps 91958 31782 -28.4% 47.98 13.6% w/volume >400 vph ST#8, and ST#9 crashed ST#lO ST#7+4 sec. on 94130 33301 -25.0% 47.40 12.2% all freeway ramps. ST#11 St#7+6 sec. on 98641 38902 -12.4% 44.48 5.3% all freeway ramps except N.B. Lodge w/ 4 sec. ST#12 St#10+Extra storage 90163 29426 -33.7% 49.37 16.9% on N.B. Lodge- 109 TABLE 20. Comparing MOEs of Freeway-to-Freeway Control Strategies (Freeway Only) Strategy Veh- Volume Density Speed Change # Minute;_,veh(ln[hr vethp-mile mph 1p Speed ST#l No-Metering 101410 1571 37.2 42.20 0.0% ST#7 6 sec. on ramps 87214 1562 32.0 48.80 15.6% w/volume >400 vph ST#8, and ST#9 crashed ST#lO ST#7+4 sec. on 87653 1581 32.2 49.10 16.3% all freeway ramps. ST#ll ST#7+6 sec. on 89976 1554 33.0 47.10 11.6% all freeway ramps except N.B. Lodge w/ 4 sec. ST#lZ ST#10+Extra storage 83525 1575 30.7 51.40 21.8% 411.122.1243. 110 TABLE 21. Comparing MOEs of Freeway-to-Freeway Control Strategies (Ramp and.Surface Links) Strategy Veh-Minute Moving/Total Speed Changes min Timg mph in Speed ST#l No-Metering 3648.05 0.82 43.00 0.0% ST#7 6 sec. on ramps 4743.78 0.62 32.80 -23.7% w/volume >400 vph ST#8, and ST#9 ST#lO ST#7+4 sec. on 6476.82 0.46 23.90 -44.4% all freeway ramps. ST#ll ST#7+6 sec. on 8665.41 0.34 17.60 -59.1% all freeway ramps except N.B. Lodge w/ 4 sec. ST#IZ ST#10+Extra storage 6638.37 0.46 24.10 -43.9% on N.B. Lodge. 111 6.4.3 Discussion of Results Figure 19 shows curves of change in average speed among the different metering strategies on three levels: the freeway corridor (all network), the freeway only, and the ramp and surface links only. Tables 15, 16, and 17, and Figure 19 all show that the best strategy that should be implemented is the last one, ST#12. This is because the simulated MOEs for this strategy indicates that it will give the best results on the freeway and on the freeway corridor as one network in terms of increasing the average speed and reducing the vehicle- minutes spent in the system. The negative points about this strategy are the longer waiting time on the surface streets, and the need to modify the number of lanes on the North Bound Lodge Freeway on-ramp that enter the East Bound Ford Freeway. The longer waiting time is anticipated because of the high volumes traveling between the freeways that will affect this factor, but the overall benefits of ST#12 more than compensate for that. There is also the possibility that traffic will change routes to avoid the long queues which will reduce considerably the waiting time at the ramps. This last possibility can not be tested with the model because it is unpredictable, but it can be observed in the field. For example, recently when the N.B. Lodge Freeway in Detroit was closed for repavement the expectations were that there will be a huge increase in delay on all the alternative routes. But the observed case was much different than that and no noticeable increase in delay were reported. 112 60 - O ’3 0.1 8‘ LL] v 0. 2 U) 50 - / 4.0 .. 30 - 20 - H SURFACE +—+ FREEWAY H ALL NETWORK 1 o T T r I I ST#l ST#7 ST#lO ST#11 ST#12 STRATEGY Figure 19. Comparing Average Speeds of Different Metering Strategies at Local and Freeway Interchanges On-Rsmps 113 The results of the simulation runs show a 9.2 mph, or 22%, increase in the average speed on the main freeway from 42.2 mph without ramp-metering ST#l to 51.4 mph when applying ST#lZ. The reduction in vehicle-minute on the freeway (by applying ST#lZ) was 17886 vehicle-minutes (about 300 vehicle-hour), or a 17% reduction of the vehicle-minute in ST#l. The average speed on the freeway corridor also increased by 7.14 mph, or 17%, from 42.23 mph in ST#l to 49.37 mph in ST#12. The average speed on the surface roads dropped from 43 mph to 24 mph, but that reflected an increase of only 2990 vehicle- minute (about 50 vehicle-hour) on the surface roads. Which means that the final results reflect an overall 14888 vehicle-minute (about 250 vehicle-hour) reduction in time spent in the system. The feasibility issue of modifying the lineage on some of the freeway on-ramps depends on the specific design of each on-ramp and the possibility of increasing the length of that ramp, or adding another storage lane before the metering signal. In the this study, since the focus was on the data from one afternoon peak hour on the East Bound Ford Freeway, there was a need to mOdify the geometry of the North Bound Lodge Freeway on-ramp because of the large volumes on that ramp. The modification was done by adding a second lane on the surface road behind the ramp metering signal to accommodate more vehicles which are waiting to enter the freeway. The existing geometry of this ramp indicates that it is possible to add another storage lane. As a matter of fact a second lane already exists on almost the entire length of that ramp but it is marked with yellow stripes to keep vehicles out of that space. This lane can be easily used without the need of any change in the geometry, just by removing 114 the yellow stripes. So it is feasible to implement this strategy for the East Bound Ford Freeway afternoon peak-hours without any need for geometry modifications. This might not be the case for other situations like the traffic on the West Bound Ford Freeway, or the morning peak-hours, or a different freeway in Detroit. For each case, different procedure or different type of modifications might be needed. CHAPTER 7 SUMMARY, AND CONCLUSIONS 7.1 SUMMARY 1. The benefits of the ramp metering strategy currently used in Detroit (i.e., 4 second headway, ST#Z) are significant. More than 7,850 vehicle-minutes per peak hour are being saved, about an 8 % increase in the average speed in the corridor is noticed, and the ratio of moving time to total trip time has been increased by about 8 % (from 0.58 in ST#l to 0.62 in ST#Z). The best metering strategy for the local on-ramps only (i.e., 6 second headway only on the ramps with peak hour volume over 400 vehicles, ST#7) is expected to significantly increase the benefits of the control system. Savings of more than 13,000 vehicle-minutes per peak hour are anticipated, an increase in the average speed of about 14 % is also expected, and the ratio of moving time to total trip time is expected to reach 0.65 The optimal control strategy that was found to maximize the benefits of the ramp metering system (i.e., ST#lZ) did include the following elements: 115 116 a. No-metering on the local on-ramps that have an hourly volume less than 400 vehicles. b. 6 second metering headway (i.e., 10 vehicles per minute) was applied to the remaining local on-ramps. c. 4 second metering headway (i.e., 15 vehicle per minute) was applied to the on-ramp leg of the freeway-to- freeway interchanges. d. An additional storage lane was added to the on-ramp connecting the North Bound Lodge Freeway to the East Bound Ford Freeway. The results of the optimal control strategy ST#12 show a 9.2 mph, or 22%, increase in the average speed on the main freeway from 42.2 mph without ramp-metering ST#l to 51.4 mph when applying ST#12. The reduction in vehicle-minute on the freeway was 17886 vehicle-minute (about 300 vehicle-hour), or 17%. The average speed on the freeway corridor also increased by 7.14 mph, or 17%, from 42.23 mph to 49.37 mph. The average speed on the surface roads dropped from 43 mph to 24 mph, but that reflected an increase of only 2990 vehicle-minute (about 50 vehicle-hour) on the surface roads. Which means that the final results reflect an overall 14888 vehicle-minute (about 250 vehicle-hour) reduction in time spent in the system. 117 7.2 CONCLUSIONS 1. INTRAS simulation model can be used effectively in simulating both present urban freeway operations and any new strategies to be implemented on those freeways, but the INTRAS User's Manual needs improvement. INTRAS simulation model was calibrated and validated successfully for the City of Detroit. On the other hand, the current INTRAS model is unstructured, which makes it very difficult to change the internal logic during calibration, and one should be careful when doing that. INTRAS simulation model is most sensitive to changes in driver sensitivity, or driver type, It is also sensitive to changes in both the control strategies and the traffic characteristics (i.e., volume, vehicle mix, volume/capacity ratio, and desired speed). INTRAS is relatively expensive to Operate as are most of the mainframe simulation models. However, it is the only feasible technique that can be used to test the new strategies from both economical and practical points of view. Metering the freeway-to-freeway interchanges can be very effective in increasing the benefits of a ramp metering system, especially when there are more than one interchange in a limited space like the case in the City of Detroit. APPENDICES 118 APPENDIX A NETWORK DATA AND LINK-NODE DIAGRAM 119 A.1 LINK-MODE DIAGRAM This diagram (i.e., Figure 20) shows the total network coded for the Ford Freeway (both eastbound and westbound included). Only the eastbound direction (i.e, the lower part in Figure 20) was used in this project to test the control strategies on its whole length, which is about 15 miles. Some parts of the westbound direction were used during the calibration and validation process (i.e., the parts around the sampled locations on the westbound direction, like Mt Elliot, Van Dyke, John R., and Trumbull) along with the sampled locations on the eastbound (i.e., Linwood, and Cratiot). A.2 EEIEQRK DATA Eleven INTRAS card types were used to code and run the eastbound direction data. Those cards were: Run Control Card - Type 99 Simulation Title Card - Type 00 Network Name Card - Type 01 Link Geometry cards - Type 02 Freeway Link Operation Cards - Type 04 Ramp Link Operation Cards - Type 05 Surface Link Operation Cards - Type 06 Link Turning Movement Cards - Type 08 Sign and Signal Control Cards - Type 10 Volume Cards - Type 20 Simulation Control Card - type 60 The cards which include the data that were used to run the network with no-metering are listed later in this appendix. 170 I‘ll 9 knuchm E” .4. I. _ I 'Fl 3. W G 9 e 9 H O hunchm 1h” 0 O G MORNING O—vOznflfl: XHOBOI 05H. «.023 u): 1! 4. mu I’l G v f 3 8552.. ‘. III 0 00 ® .oN 0H5»: o>zo._ '\ 3.): mud 121 >(3wmcu mono; .0.2 1 0’. d R. GO GO zufiofil “Mag 5.! ..J. (v 06 55mm”. ... Mm; mus 56:8 Ev... oooszj dfl Iguana“. wwatmia 11 .1 II. . . . ... t .m.z 3553. dd. 92:3 5a: m z m H H H .‘s. at 9 U E 0.. n . . 4, . 6. O m. 9 G e 9 Q 9 m. O r . I .23qu .1 368 a 6833. 3:. 5.. s z [.2335 u. 3.1.4.0 as I .H .02 l. 9 I 0 v .000 0900 eta... @0000 c U 0 .... MM 0 9 Q. ,t p... . H H H H H o 03.: «.44.. 9.8.3 523: J. a 8.-. mm 89: uzxo .2. “al.—u $.31 41% it 25.3.50 .ou Daub 122 ><3wu¢u a Heinz. . 8.1.6.5.. H. *O z > H a . 0.1 . H g Zwaam» JUMWDC t H L .... 3. 1 mm. r t 1.. av AH mu mu Aw mu mm nw mu mWY Ill an”: 83.6 ...: ~55 23qu #945 F1 3:380 .8 3:3...— 123 ma 92.210 1 R 9 f. a... m. Q. is: 0.50 .8 253m 9” We a uzzou .02 a - t . . ‘ [Will cmZZOudd Continue Figure 20 . 26'— 124 EAST END 125 A.2 Network Data mam-emu emo— m W was": amt - N. IM‘ 126 an n run-cum»: sum-WW 127 128 129 130 131 132 133 134 135 136 137 138 ......3 ....2. ... 139 .__._§_._._._.. . 140 ...: ... s «E: n.3,. 141 142 143 144 145 3.. al.-3449 f( It: ...... vac—p.340 ¢( 9:4“ .1... 02:25.0 6'. VIC. 4... ‘37: IE 146 h... .1935 .04 mt: ...... $2.233 :4 Ht: 2.. 0243.. $4 wits “Q 0 o “as: I‘ M w M 9'“ Mow 1m“ 147 MT. «13.300 ¢( '8: V... $18.33 *4 .163 I... 024300 it 83.6 B.-. 1.1-:13 «C m M M 148 “so. 1!... .16050 IC '23.» b...- fiz‘g $‘ 2‘0 “.0. “2‘3 *( ”9“. a... 170.706 I C m w M m 149 hf. It... .15069 '6 I84» {I V... «2‘3 ¢( Ht(3mw3u each u. . C.Nmno on .05 O n.— O O_n O on O 0 mm .05 mo ow 5N h.— O.ano on .05 wauu ~02eao pmos Dozen—o seuupmzao «woo: zo.p¢:..muo zest muse mouse to was» \ussoes to panacea zaux . aims mxz.a sacs . o o o . o o \o «epxoo. o \o o a. o o o o. o o o 0 ago n .an .on . q . o o o . o o \c an \00. o \o 0 up o o o o o o a can as. n .6“ ..n . a . c o q . con" 0 \0 on \oo. o \o c- a» o o o o o o o o as». a .p« .4“ . a . o o o . o o \o nu \oo. o \o o as o o o o o o o o o n .0“ ..02. . 22:02:23. 23 988... 5.... ..u a 9:. so... as: :3 3:8 33m 4 a a :3 4 a a :3 2...... ~25 x23 5 can ¢.¢a .aum unapa>¢=o mace: zo._¢z.pmuo Dos. execum pmc.t to azaa ..u \uxsso> to .zuocua Duet nuz(3uuflu . am>¢u~2—n:m who .I.m\) prIGum 26-h180—410 h.\¢.\v .1. U! want pO—aau .b! nu >¢Iww¢u 0805 a) an2¢om 166 Oil---" ..-"--Il' .Nhu .. .0 .Nhn .. .0 Omwcacn Oh zo~h<438~m awkwaalcu mzuh Jana a<>uwh2—m:w m—Ip 02—330 Occu a wa>h 26 Gang—Owam m<) w=4<> bom\bmw0 pua\hmwo h0¢\hmwo h0t\pmua buathuo wantb gnu» . d<>¢flb2~02m u mcwhw8<¢< w0b 0024—45 boa >(400 >daw0 41.0. h\2 w2—b wluh U2~h 202 2.2 mud—2 .200 .200 .00 x2.. 4 32~J m40>0 hqm w0w00 wdu2u2w> >02 adp0h 0wmam -20) -2W) 0>< 0000 Im) 0>< thUEwn \2u212w> w40—1u>\m0200wm m0.pm.h(bm 22.4 u0-.(40200 168 .0.. I 04.2|Iu>\.2—2.02.. >(am0 .60.N I wa—232w>\a2—2.w2—. 40>(u. .0..«vn I .2.2n2w>.><4w0 41.0. .00.Nn I .232.0uuam .00. I .2w.200 .200000 .«..n. I .20.200 00<¢u>< .0.n.0I w2.. 0.0. 44.0.\02~>02 .00.00. I mu.02.2uwa0~20> .90.00« I mwau21040~20> m22.4 00¢u¢0m 02¢ 02(0 02.00402. m0..m~.<.m 8003.02 0.«« 0... .00.. 00.. 00.« 0n.0 «0.... qw.«m« 0.00. 0'. «cc 00¢.0. 02. m00<00><. 0.0. «.mq .000. 00.0 00.. 00.0 0.0 ..0 ..0. 0..0«. vn.«0 0.0« 0« 0.. 000 .0« .0« . a 0.00 ..n. ..v.. .... .m.« «0.0 0.0. ... «.v« 0«.00. «a... 0.00 0' 0«« 00' .0« ..« . n «.0. c..0. .0v0. 00.« ...n ««.0 0.«0 a... 0.00 «0.'0c .0.««. 0.00 no .0. .«0 ..« .0« . « ¢:\00.2 00.2-2.\20> u:\2.\:0> ...00 ...0. .\2 02.. 02.. 02.. 2.2 000.2 .200 .200 0220 .00 22.. . 00000 ...0200 02:.0> 0..2-20> .0000 0>02 ...0. -20> -20> 0>< 20:0 02¢. 20> \2.2-20> 0.0.20>\0020000 m0_.m—.<.m 82—4 >(3wwuu 0 ma . 02.. .< .00.:0 w>—.<40200 . 169 O'NNNNV'I’OQOQ 00000. 00. .0 ... . . 00 .0 00 000 0. 0 000 000 000 000000 « 0. a 0 v 0 0 . 0 0 . 0 0 0 0 n 0. « .. . a 0 «a. 0.«000 n 0 0.00 0« .0. . 0.« 0 ... . 000 0 0.« . 0« 0 m« . .« 0 0« 0. ...0« 0 .« . .0. 0 0« 00. 00 00 00. 00 .0 00. ...0.0 00. 0.0000 00. .0 00 00.. 00 .0. 0 .0 ... 00 00 0 a 000 . 00 .0 0000 .0 00 0 . 00 .0. . 00. 00.0000. 00 00 00.0000 00 .0 00 0: ..00090 .0 ... 0| .00.001 ...0.0 ... .0v. 0.0000 Q- 00.0000. .0 00 .0 00.0 00 .0. 050 .2.0\3 .202000 20..<0.4(> 00\0.\0 «:0 00.0 .2 .0.440 ..2 an >. \02040) 00 .200000 2102 2102 012.4 0010000 001 0 . ......0.0. 0 .0. 0.000. . 2100 0214 x2.4 0 20 0 . 0 0 0 . 0 \0 0« \00. 0 \0 4- .0 0.« . ..« 20..40.0..200. 024. :24: .0 0 04: 02.. .420402 2.000:.2. .20.: 0:2. .00. 004:0 000:0 24:0 024. 00: 000x20 0:0.4>:00 .00. 00200.0 240:.0220 00002 20..42..000 2000 00:0 0:000 00 0:». \020.0> 00 .200:0: 2402 2402 nu 022.. :24: 7 1 . 0 o 0 . 0 0 \0 «0.x00. 0 \0 0 0. 0 0 0 0 0 0 0 0 000 a .0« . 0 0 0 . 0 0 \0 0« \00. 0 \0 0 0. 0 0 0 0 0 0 0 00« 000 n .0« . 0 0 v . 00«« .0 \0 0« \00. 0 \0 v- 0. 0 0 0 0 0 0 0 0 0.0. n ..« . 0 0 0 . 0 0 \0 .« \00. 0 \0 0 0. 0 0 0 0 0 0 0 0 0 n .0« 20..40....z00. 024. 020000 .0:.0 .0 : 04: .20.: 0:2. .00. 004:0 000:0 0 0 4 20. 0 0 4 20. 24:0 024. 00: :.4: .000 0:0.4>:00 00002 20..42..000 20.0 020000 .0:.0 00 024. ..: x02000> 00 .200:0: 00:0 0024. .:4...x:4 2402 012.4 .100000 . .4>:0.2.0:m 000 .2.:\2 .202000 20..40..4> 00\0.\0 .0.440 ..2 I: .130000 0.0 04.0 .2 0000 00 .02100 ..... . x2.4 .0" . v ..u . n .0« 0 « ..0.... x2.4 4 4 4 171 0 0 0.00 0 0 0 0 0 0 0 ..0 0 0 0 0 0 0 0 0 u 0 «a. 0 «a. 0 20.0.0.00 0000 0 0 0.0 0 0 0 0 0 0 0 0.0: 0 0 0 0 0 0 0 0 0 0 «a. 0 «a. 0 20.0.0n0. 0000 C 0 5.00 0 0 0 0 0 0 0 0.0 0 0 0 0 0 0 0 0 0 0 «a. 0 «N. 0 Ann .0... 02<0 00. 0 p.00 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 u. h 0 0 000 0 .00 an .00 .on 0 >300 090 0 «.00 0 0 0 0 0 0 0 «.00 0 N 0 0 0 0 h N. 0. 0 «00 0 000 00 gen .5“ 0 >300 00. 0 v.00 0 0 0 0 0 0 0 0.'. 0 0 0 0 0 0 0 v. 0. 0 000 0 009 on Asa .00 0 >300 0 0 0.0 0 0 0 0 0 0 0 0.0 0 0 0 0 0 0 p N N 0 005 0 005 0 .0“ ..Ch. >300 0220 0000 00000 20—pdN—4022<20 040 A00>40 .20) 0 0 h 0 0 v n a 0 .b0 002p #004 000 .000 x2_4 00>» 004 4.02 0) 00 000202 b200000 h2020>02 200k 20> x204 024. 0.0 0>4 .20::00 0.0 0:000 >300 024. 2. . 02.. 0040000 0 0. 0. 02—p bc 00~h0~h\.2~8~u8—h >\A2—lvul~b au><¢h .Oo.han I Azuluzw>v>< .mnn.OI wink a—uh 44b0h\02~>08 .nu.nun I mwhnlnllmau—Iw> .OO.umD I mud—Inw40~1w> me—a moduuam 024 1848 @2—03402— muuhm—bdpw XROfihuz 0.00 0.«v .mnm. 05.0 nm.. no.0 90.01. 90.nvn u.'¢ #6 can w44h0h 02¢ mmG¢¢w>< h.nm $.00 .900. .0.0 N... 05.0 n.n «.0 p.mu .OO..¢. hm.0h¢ —.np nu wow .OG .06 .On . e n.—n «.«m .Onou 00.. «G.. mv.O «.0. 0.0 O... an.hhu Q..VQ. h.h« 00 0'6 000 .6“ .h« v n w on 0.«. ..nn. v0.0 n@.- OQ.O O.n. a.v. D.Oa «h.¢hn nh.ouu 5.50 on an. 00h AFN .ou ..N cixwa—S w4—8-24\Iw> ¢I\24\Iu> >(Juc achOh h\8 US—h wank miub Zn! mad—8 bzau hZOO 0210 #30 XZ~4 a OwwLm >p_m2wo m834°> ua—!-Iw> >(awo w>al a<~°p t1w> urn) G>< 83:0 u2<4 1w) \2—8n1w> . mac—Iw>\m0200ww wO-bmuhdhm Slug ><3ww¢u O O. D. uI—h h< hathSO w>ub443830 173 xxx . xxx xxx ‘ xxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxx xxxxxx xxxxx xxxxx xx x x xxxxxx xxxxx x x xxxxx xxxxxx x x x x x x x x x . x x x x x x x x x x x x x x x x x x x x x x xxxx xxxx x x x x xxxxx xxxx xxxx xxxxxx x xxxx x x x x x x x x x x x x x x x x xxxx xxxxxx xxxxx xxxxx x x xxxxx xxxxxx xxxx x x ,xxxxx xxxxxx xxxxx xxxxx x x x x x x x x xxxxxx xxxx xxxxx xxxxxx x x x x x x x x x x x x x x x x x x x x x x x xxxxxx x x xxxxxx x x x x .x x x x x xxxxxx x x x x x xx x x x xxxxxx x x x. x x xxxxx x x x x x x xx xx x x x x x x x x x x x x x xxxxx xx x x xx x x x x x xxxxx xxxx xx xxxxx xxxxx x x x x x x x x x x x x x x .x x . x x x x xxxxxx x .x x x x xxxx xxxx x x xxxxx x x x x x x x x x x xxxxx xxxxx xxxxxx xxxxxx xx xxxxx xxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx O. .. ...... O....... ......O... ...... .6 .0 . .. O Q .. . ... ... .......... .......... ... ... . .. . . .. . . .. . . .. . ... .I. .. ... ... A... . .0 . . .. . .... .... ... .....O.. 0.. .. .. .... .... .......... .......... ... ......O.. 0.0 ......fi... ......O... ........O. ......O... .......... .....it ... .0. ..I ...... .. .......... .......... .... .... 00¢... O. .. ... .0. .. .... .... . o. . . .. . O. .4. ... .0. ... ... ... . .. . . .. . . .. . . .. . ......GOOO 00...... .....0.. ........ . O. . . .. . O. .. ......OQO. ...... ...... ...... .. .. 01.01. ...... C. .. 40...... ....951. O. .. O. .0 ... Ob. .. .. . 1d .. .. .. .. .0 .. ... .. 0. .. .. .I. .. .. .. 1. .. G.‘ .6 .0.. 6. 9. .. ... to. .... .... i. .. .......G ... ... .0 O. ..9400 .. .. 174 C-"NHNH"NHNflflflflflflflflflflflflflflflflflflflflflflNHNflflflflflNNNNNNNHGNNHNNNNNNNNNC‘. 88.2. 00 CON. 6 0. .3 6. on In an 06 0' 0 09 0 6' 0 06 0. 06 v. «6 00 06 ha .6 06 «6 66 so u 0. 06 6. .0 0. a. .0 6- no .I .o .o c- .0 "0 HI houno 6. 6. 6. Oh. .I .o 6n~60606 6nn60.0 66h6000. 6656000. 6096000. 605609. 66960~n 6nn6000. 6666006 665600.- 6n~6000 66h606~ 6006 h. 66.6 00. 6606660. 6nhnnuw. 6n~6660. 6m~6000. 6nsnmmh 66h6000. 66660.. 66n6000. 6o~6066 66.60.0- 6656000. 66660n0. 6096000 606606? 666600! 6696096 66566566 6696006 66.6000. 6nh666. 60.6066. 66.6000 66.6066. 665609. 60n6006 66n6000. nahnon. 6696090 665666. 6u~6060o 66~600v 6656060 6656000. 6n~6006 6066066 66h6nh0. 6n~6006 60~60¢6. 66.6006. 66n6000. 6n~60v~ 605600.6 6636000. 66.60666 66h60vh 66960006 6696 on In 66 On an 96 6o 66 on 00 0' 6' h. 0' a. QC 6' 6' 'flfl'SOCFCC 00. .30 mtuuul .818-.00~-0.n— 660608 2000¢Z~ ¢2_80>) >d20 >¢3uu¢u . >QIUqu 0805 02:00 p640 .02dum bm—a d-«O 00.0 175 HNNNNNNHNNNNNNHHNNNNNNNNNNNNNNNNNNNNNNNNNNH“NNNNNNN'CNNRNNNNNNNNNNN 66 00. 60 00 0.660 60 00 6.660 .0 6- 6. 6- 6. 6. '0 6- v. 6. v. .I '0 6. 6. .0 Q- 6. 6. 'p Q. '0 Q. 'I '0 .00.006 .0'. .0660.6 6066006 606.066 600. 6006000 .06609v .006006 .06.000. .00. .006006 .00.00.. .00. .066006. .006000 .006060 .00600. .00.06. .0.. .066006 .00.006 .0.. .0660.6 .00.000. .06. .006006 .06.000 .00. .006000 60660.6 .006006 .06.006 .0v. .00.060 .0660V6 .066096 .00.066 .0.. .06600. .6660.6 .06.006 .6.. .066000 .666066 .00.006 .0.. .0'. .666066 .00.006 .0.. .0v. .066066 .06.066 .0v. .006066 .06.066 .0.. 60.6006 66.6060. 6066060 66.6000. 60.6060. 66.60.06 6666000. 606600.. no..nn x9...- .0..» on no. no.«a. «...»- .0... on... on no. oo.nnn an“... am no. -.«n« «an... ....u .0.». no... on on. can... ..«om. ca .0. ...n.. an...- o. no. no..nn .nn... I. ... .I.oa« cane.- an... ox... .. on. ...... .....- o..np. ax.n. .. ... ...... ax.o.o up.o. . ... ...o.. o...o. a... n x.. ...... oo.u0c 09.no- . mu. ...... ....o. ...no. a no. no.«o. «o.ao. u .0. ...oo. oo..o. .o no on am no... .0 no no no no .- .. oo 0. on 176 OINflflflNNNNNNHNNNNHNQN"NHNHNNNNNNHNNNNEN'"'Q.."""Q'QQQ-Q""Q'9' C30¢30<>0€30¢30<>0¢30<30¢>0<30C>0¢i0¢30¢50 066 006 006 006 00' 66. 66. 00. 066 "0'. 0¢>0¢D0<>0¢DC>0¢30¢30<30(30(30¢>0¢30<30¢30 '6 §§§§ §§ 006 066 006 00. 006 006 0'6 6. a] v. 6- 6. 6. Q. '0 6. '- 'I 6. 6- 6- 6. 6- 6- 6. 6- 6- 6. 6- 6- 6. 6- 6- 6. 6 00060 .0060 .0006 0660 .006006 .00.006 .0Q. .0060" .00600. .06.000 .0Q. .0660v0 .00600v .066006 .00.006 .6v. .666660 .666066 .00.0.6 .0.. .00600v .006006 .00.00v .0.. .006060 .666006 .00.000 .0.. .66600.. .066006 .06.066 .0.. .60600v .6660v6 .00.0.6 .09. .006060 .066066 .00.006 .6'. .666000 .666006 U -GVFI'IDIDCI‘IDO -(hFl'lDCDCI-II. gs 666666 6660'. .6660 60 066 0660.6 0.66v0 0.660 6.6.0 06 0.6 0.66.6 0.66.0 v.600 Q0 6.6 6.66.6 6.6600 ..666 00 0.6 0.6006 006060 0060' 6' 606 606006 006660 0060' CC '06 106606 606666 6066' 0' .06 .06006 006666 00.06 66 00. 00.60. 66..6. 00.66 66 60. 177 'r'!0"VQ00000560600IIIIOOI’OC’GQOOOQOOUUOUOUOUOCCCUOCOOODUOOOODOOIOIIDOC OGDCDOCDO ONO” 0006 006 0006 606 006 006 066 006 ------22222 22222222222222222222 "at a 66 .0660 66006 00. 00. 00. 00. 00. ...... ...... «an... «an... 0.«... ...... ...... ...... ...... ...... o.«eoa can... sauce. 10.... venue. non... .oaoo. can... ...... ...... .666060 60.60. 60..60 666660 666660 00 666 60 066 6.6.0 0.606 ..660 0060' 0' 106 6066' 0' .06 00.06 66 00. 00.66 66 00. 00.66 06 60. .0.06 00.06 06 00. 66 00. 60.v6 00.66 00.66 00 60 '0 60 60 .0 00 06 06 60 00 06 C>O¢3C>0¢30 .6.°6. 60.00. 60.00. 60.60. .0.00. 00. 00. 00. 00. 00. 00. 00. 00. 06.000 00.6 0 00.000 00.600 00.900 00.600 60.600 00..00 6 066 6 06. 0 006 0.0.. 0.60. 666 066 § 00. .0 §§§§§§§ 228228282828222 0' 6 ....-O’I-I--.III'OCCI--3222222822 66 66 66 66 . 60606 66006 -NHVOQUFOO U -flfl'8aUV-C 006 00.666 .0.066 .666.0 0660.0 066066 Q66000 v0.60. 60.0.0 06.06. 06.66. 66.6.0 06.66. 66.0.0 .6.06. 06.000 60.00. 00.000 00.000 00.00. v0.100 60.666 00.600 60.60. 60.600 .0.00. 00..00 .6600 0.660 06 0.6 '0 6.6 06 0.6 6' 606 6060' 00 666 06 '0. 0. 60. 0. .0. 00.6. 06.0. v. 06. 06.6. .. 06. 66.0. 0 .6. 00.6 6 60. 1 60. 6 60. 6 .0. 60 00 60 v0 60 60 .0 00 06 06 66 06 66 Q6 178 60 066 an o.“ 00 6.6 on 0.« .q .06 00 006 00 .06 .n no. 06 60. 06 00. 66 00. 00 666 06 00. 0. 60. 0. .0. 0. 06. 666666 0660.6 6.66.6 0.6006 606006 006606 .06006 00.60. 00.666 60.60. 00.666 00.666 666066 00.60. 60..66 .0.066 06.66. 06.06. 06.66. a 60.00 c n¢.6 60.6 .0.. on. 00. 6666'. cc. .66oo 6. no 00. c.6pv- o0. 6.660 6. on 6.6.0 0 60 co. n.6vv. 00. 0.606 6. 60 oo. oo. oo. 6.66v- oo. 00. ..660 v. o. 00. 066060 00. 0066' v 00 oo. ocusno o0. «060' a .6 00. 60666. 00. 6066' 0. 00 oo. 00. 00666. 00. 06.66 a '0 oo. .a..6- oo. 00. 00.66 0 00 00. 666060 00. v0.66 6 .0 oo. «a..no oo. .6.06 .6 60 oo..« 6 no 00. 666660 00. 666660 00. no.v6 no 6. w..66 o. 00 60.66 6 00 oo. vunOn- oo. 60.... 00. .66... oo. Onao.o oo. 0.... 6 on 66.0. 66 66 00. 66.6.0 00. 66.6. no 6 oo. n..o.¢ oo. 6..o. n .o 179 0 O 0 0 0 0 0 .-----‘--~------9---------2------------------’----. 00 0 2222 O00000000000000000O00000000000000000000000000000000000000000000000 180 66. 6. 66. 0. 0 0. O6. 000 6 00. 600000 00. 600000 0 .----------------------------------------------------—------------ N 2 O 000000000000000000000000000000000000000000000000000000000000000000 00. 60. 00. 60. 666 '66 00. 60. .66 O66 60. .0. On. 06. 06. 66. 06. a... 06. 66. 66. .6. O6. 00. 60. 00. 60. 00. 181 mo.nmv.om.wm.mo 000 h-ncncuvcwcumtncuwlnu>0'cU10. 82.. 0 0 0 0 0 C: (D 0 0 0 0 0 0 0 C) {D 0 0 0 0 0 C“ <3 0 0 0 0 0 0 O 600.. \z.x-zu> 00 06 60 00 00 66 60 00 00. 60. 06 06 00 60 06 00 00 00 .x! 0 0 0 <3 0 c: 0 c: 0 c: 0 <5 0 c: 0 c: 0 c5 0 0 0~ <3 0 c: 0 c: 0 c: 0 <3 0 q 03—. 0 0.9 9 9 9 9 9.0 . c ..0 9 9 . 9 9 9.9 9 0. 0 o 9 ..9 . 9 . 9 m . 0 9.9 o 9 9.. ... 9... o 9 9.9 9 9 9 0 9.9 0.. ... 9.9 9.9 ..9 9.9. o 99 9 9 .... 9... 9 . 9.9 0 9 0 9.. . . 0 9.9 . 9 0 0 0.. 0.. 0 0 9.9 u 9 9 9 9 0. . 9. .. 9.9 0 .. 0 9 9 o 9 .. 9.9 9 . 0 0.9 9.9 9 9.9 ..9 0 9.. 9 9 .9 9.9 ..o .0 9.9 9 9 0 9 9 0 . 0 9.9 0 9 0 . 9 c 9 03.. 03.. 03.. ....0 9:0: ...0. 0.0.:9>\90200.9 '9 .Nl’fl 0 69° 0! a up a .00 .00 .00 60 .0 06 60 .06 66 _66 .06 .60 .60 .00 ..0 60 0.. 06 .06 .0 00 00 00 .0 :68 06600 60.06 90.9. 60.66 .0 00 06.06 06.0. 00.99 99.999 99.9. 9..9. 9..9.. ...... ...90. .9.0. 9..9 09.9. .9 .9 .9 09 .6..- 09 99. 06... 60 0 9..0. .9.09 99.9.. 9..ou 60 6. 0. 9. .9.09 06 6. 2.: ozw> mu..m..0.m 82.. wU¢ 0000 26> .Cv ..06. .00..09 . .99 .09.. .00..09 . .99 .00.. ..0..99 . .09 “90.. ..0..09 . .00..09 . .09 .09.. .06 .60.. .90...9 . .09..99 . .06..66 . .00 .066. .09 .99.. .0. .60.. .0. ..9.. .00..6. . .09.... . ... .09.. .09..9. . ... ..9.. .66..O. . .9 ..... .00..9 . .n .90.. .v .90.. .6 .60.. .9 ..o.. x2.. 06 06 06 66 06 06 06 66 66 .6 06 6. 6. 6. 0. 0 UP 183 we..c:m no..¢:m outucaw uu¢.¢:u wudbflflm NUCbBOm no..a:u uu.u¢:n no..¢:a uu¢.c:m uudblflm .u..c:m uOanOm UOCtIOm uo«u¢:m uu¢.¢:u Uudulbm UUCbIrw Old. Aid. $84- ‘1‘. £843 $I¢¢ ‘3‘. “Id. 5". Alfie “It. ‘3‘. $3.8 681' 5318 O O O O O O C) <3 O O O O O O C3 ‘3 O O O O O O C, <3 O O C, <3 O O O O C n o O O V p O O O Ci <3 O O O H a O O O O O V <3 ‘I O O H O> o h ' O O O O O’ O C) O C) O <3 O <3 O <3 O <3 O <3 O <3 O <3 O <3 O <3 O <3 O <3 O l 11" .U 90 0O .0 HO o6coo'o'o'ooo'6006660666666666660666d HO .- .0 «O 9'. HO. v. HO Q. an 50 DNA an. an. In. no. N.. CH. 0'. kn. hp. an. bu. no. 0.. v0. 0.. DD. V. ha. 5 ‘ - 1. Oh. Oh. 9h. 09 «9‘ no as an no 3332383,: oooocadooooodddoooco occdjddc: C .9°.. coooc PI 0 d cu c a; 9 cu c on «i 6 «6 0 <5 6 O N O O u f fi O I. I 99¢ «an .I’ CCU...- “39".9 "..".°. an o n a an ca 9 o a .w. H ~ q o o o o q a _ o c a u» o OCOHQC n. O. O U! I! v Q Q O C e h n I Ob ' “UGO OEQOfiOD w? 0v 0' no «a O? ,nv 0Q .90 an 9' h! ,5. .NH Q6 .9. 9' .0. ha I. an «a In .0 on an .m '1 #6 Oh. hO. On 90‘ no. (I. On. a. h. h. HO. no. 5'. '0. ha 9. .na ION no. ,OQ- pan '- «a. O. ‘0' up. ..n on. w. .h BO. 09. «0. 90‘ 30. «C. ..u .6. On. In 9' N. 0. On mm «a .69 «a. h.- .Qfl .O. an .nn an ‘wn .- 0n no G. a. .66 66c; OOH d o d d o c 6 o N N <3 0 v' C O O <3 N O O N O O O " O N C O O <3 O f‘ O <3 0 I! n <3 O wwm OOQ OOO n0.- I9¢ Ova nan Onn on. GO VON vO' vOn .OO Ohfi own Ofl~ a.n COB hm? n00 GHQ new «a. .0. flaw lmq w.m ..Ofl.OOn. .00..00.. .mo..hn«— .nO..n0-. .m...nnn~ .....nnnv .nnn.kuv ...-.0... .n....nn- .....Onn. ..h..h~.- .w~..mn.— .09..nh.. ..h..Os.- .nw..0¢.- .OO..CU.— .nwp.n0.. ..O..O¢.- .00 .0nw? ..«n.nu . .nu .Onn. .o.«.«o _ .~.n..o . .on .o.- .v.~.oo . .vn .n.«. ...”.«n . .06 .O.n. .OO«.av . .hv .pOa. .rc«.0v . .vq .vOn. .mOn.nv . ha $0 an on an ~« on nu .« a. h. 0’ Q. «- O— 0' Q' HQ «Q Ov an In .9 On an vn an an .0 184 O O m mN.O Ov.. NQ.O 06v m0.mvwnhh‘n.wn m.Cw mm machop 024 mw0< mu nv.mm n.. a mum Annn.nunv .m wu m.. a 0m» AoNu.m_~. me mu mn.uw n._ . can .FOw.oo~. .v wu «0.00 .0.0- «\I 00030 ...0200 0 I. 0 «5 H .6. H 0. H .00 .H0 .0H H. OH 00 .00 .00 .00 .00 .00 ..O .00 .00 .00 .00 .«0 .OO .«0 O0 00 «0 .« 0 0 H 09 0 O0 .OH 00 .00 OH .0H «H .0H .0H HH .OH ..H .0H .0. ..H .HH .HH .HH .HH .0H «H OH .OH .0. .00«. .000. .0«O. .HOH. .OOH. .OOH. .H00. .HO0. .HHH. .0HH. .00H. .0«H. ..H0. .000. .00.. .00«. .«.0. .0HH. .HO0. .OHH. ..HH. .000. .HOH. .«OH. .OOH. .HHO. .«.O. 08:00) H0 0O. 0O. ««. H0. 00. «H. 0H. OH. OH. OH. 0«. 00 OH OH H« 0. 0« OH 0. OH «H 0. O. O. H. H0. .6 3.. 666 n ... ...H 6 ..H O 00.. «0.0 0.« H H « 0 O 0«.. 00.0 « O. «.0 0 0. H «0.0 0..0 H.0 ... H O O HO.. 00.0 0 0. H 0. «.OH O 0H.. 00.0 ..H 0 0 0.« O 0... 00.0 9 O 0 H. H OH O .... H«.O 0.. 0.0 0 0 O 00.. 0« 0 0.0 0 H « H O .... H«.0 . 0 0.H. 0 0. O 00.. 0«.O « H 0.« q O. O 00.. v«.0 H H ..0 H.0 0 00.0 «0 0 0.. 0 0 . 0 .0 00.. 0«.0 0,0 0 H. «.0. 0 0H,. .0.0 0 H «.H . O .0 «H.. «0.0 0 H 0.0 H . .0 «0.. .0.0 0 H 0.0. 0.«. .0 «0.. 00.0 «.0 0.« 0.H .0 00.. 0«.0 0 0 H.H H.v .0 00.0 «0.0 H.H 0.0 O H. .0 00.. «« 0 H . O 0 0 0 .0 .0.. 00.0 H 0 ..«. H .H .0 .0.. .0 O H.H 0 H. H.«. .0 0... 0« O H.H H.0 0.H. .0 00.. 0«.0 0.0 H.« 0 H .0 00.0 00 O 0 H 0 0. 0 HH .0 00.0 H0 0 0 . H 0 . .. .0 «0.0 00 0 O 0 0.H« 0 0H .0 H0.0 «0 O 0 . 0 0 0 « 0 00.0 «0 0 0 O 0 0H n OH 08.. 08.. w!.. ...-8...... :36 2,9. .16. \2.8-:u> 0.0.:u>\00200um C C «H. «O, H. «0. «0 OH O«. O. 00 0« «0. H« 888 HO. HH 00 00 .0 wHHH 000 000. ,«H0 «.0. .O«0 .OHO. .00 ..H 0O0. .00«. .000 .OHO .«OH. .OOH ..H H00 «H0 .00.. 000. 0O« O«0. OH0 HOH. 2.8 ...u) 3.3.»... .2: 2):: 8... .4 52:6 22.3356 0 «O. 0«. O« HH. H0 «H. 00 0«. OO. HH. «0 ,onvn 00H O«0. .H«.. .HHO .0.0. 0.0 .OOH .H00. 00« 000 00.. ...a .... .9... .«H.. .«.0 .«O. .00O. 00« HH.. «O0 OOO. mua.a .:.>. 0.00 00 0.0 H H .H .H 0.0 « ...H OH 0 0 0 0.«H 0. «.H 0. H 0 « 0.0H 0H O H. 0 ..O. O. O O 0 0 OH H. ..0 O 0.0. 0. ...H .H H 0 H H 0 0 H 0. OH H.« 0 0.0H 0. O O. .H 0.0. 0. 0.H O 0.0H «H 0 H. 0 H..H HH H « O 0 «H «H «200 .200 O>0 00:0 .0O0 O00 HHHH OHH. .«pv. .00. HHOH «0H .OH 0HH. 000 O00 00H O«0 «««. .«O HHH HOH 000 O«0. OOO 00H. «OH 000. «0« «H0. 0H0 000 0220 020. OOO0 H«O0 0.00 .OOQ 0H0. H«00 «000 0000 O000 0.«0 O0H0 ««0H HHO0 «HOV 0000 «O00 OOHH «0H0 HOH0 00.0 00.0 HO0H 00HH 00OH «:O 20) .OH ..H .OH .«H 2... .2. .on .om .0H .0H ..« .2 .HH .HH .HH ..H ..H .00 .OO .06 .OH .0. .O. .0. ... .«. .H. .0. .0. .0. .0. .0. .0. .n. .H. .N. a“. ... ... .0. .O. .0 .a .- .. J t .9 .O .0 .0 .OO .3 .. 3 .n 3 .a 2 .. 82-. 0H OH «H OH 0H 0H HH «H .H OH 0. 0. «. 0. 0. H. H. .. O. Ow.O u w;—Z:Iw>\AZ—I.wI—h ><4w0 .N0.. u w4~2IIw>\AZ~I.w2~h 4w><flh ..wm.mwflvvu Aznzqu>v>44wD AdhOb .nH.N0 n .IQ:.Owam .@00. u PZwFZOU hZNCCDU .O..Mh¢ u thPZDU m0<¢w>< .hhm.0u wink Q—Qh Adh0«\w2~>02 .5¢qow.fon mwhDZ~SIwJU—Iw> ..0.mnmmhu mud—25w40~1w> . ... L m.; waZ—a wU< .IH.\\ a.m. H.H« .m.0. OH.H HO.H OH.C m « «.H 0.6. 00.000 mm.HHH 0.0. 0. ««O. vam .«O .OO . 0O «.Om 0.00 ..mm. 00.0 0... «0.0 H m 0.«. O.wH H..H.OH OH.OO«. O.HH Hm 0H0 m«O0 AOO .00 . mm 0.00 ..HH .«OO. 00.0 OH.. «0.0 m 7 m m H H. H0.Hv.. OH.HOH ..H. « OOQ. OHHm .0O .0O . «O 0.06 m..H .m«m. 00.0 OH.. «0.0 m H H m «.H. am.00.. 00.000 0.0. «H .00. H.Hm .00 .HO . OO « mm H.0H .000. HH.O HO.. m«.C O ? m.0. o m. mc.«00. 00.0«H. m.HH .H 000 Ohvv .HO .Hw . 0O REFERENCES 188 @©@@@&& QQQOQG GO tiéfiaQQQuG @@¢@@@@Q00 99¢99G 0&6 0&0 0. W U: ‘9‘ 500 &o& Oc§oo¢¢ 956906 0‘ QéééoacfiGo ©$Gfl¢9¢0¢¢ 90009006 OGGQQGQGQQ GO Gtéégooo QGQQGQQOG 66% 000 00 GO 0G9 0&0 GQGQQGQG GOGQQt GGOQCGGQ OQGGOQQQGO Gt OGGGGGGQ OGGGGGQQ‘ 0G1 00a 0‘ GO 605 000 fiQQGGOQQ fle©4¢o QQOQQQQQQG QGQGQQGGGQ 66c Qtt QGO 6&0 906 000 006 0G6 69609900 @GQOQG QGQQQGQOGC OGGGOG¢¢OG 00¢ G60 00G 00% Gee QGG 9&9 0G6 GOOQOQ¢Q QQQQQQ Ga. mi...» m: Qt: ttfi 0:0 0 ©an de 4% @@©@¢19<@¢ thm¢é 6% Ga 966 @G@ GQé©e¢¢Q @éeéxi @@@@¢@ @@© oéc 9:0 :00 0% fit Q66¢¢£§¢¢9 @@Qé@w i% @0 999 9&0 @Qéfisrsé @cfidifl G0 90 00 at 90 00060060 QGOOGQQGG @Q 0% ©@ 90 Ga Q@ @e 00 ¢& 09 66 69 966000900 QQGOGGOQ QGG¢¢¢Q¢ GQGQQGoeé $0 69 0Q 06 66 Ge 90 06 on G0 ec 00 ©Qafl§¢0®fl @4966960 @@ 06 $4! ¢© Sifie ‘Q ©4©O 09 a; ééé Go @@ @@9 06 CG OQQQ @Q QQGQ @d QGQ @b 00 @b 96 9&0 90 @QQG Ga @699 00 90 006 00 Q@ 096 66 09 GQGG @m QGQQ 4© éGG mm Q0 GOOOOOOGGQ QGOGGQGGGG 06 06 000000 006000 00 00 6666660690 6660090060 Qcavfifiocfio oqafiwaéooo 60 GO 6GGGGO GQGQGO 00 66 G¢Qéétcw§6 @GGGQQQGGG 189 REFERENCES Aerde, M., Yager, S., Ugge, A., and Case, R., "A Review of Candidate Freeway/Arterial Corridor Traffic Models", 66th a Me 0 ra o t o esearch Bo rd, Washington,D.C., 1987. Buhr,.J., Meserole, T., and Drew, R., "A Digital Simulation Program of a Section of Freeway with Entrance and Exit Ramps", 331 Reg Reg 230, 1968, pp. 15-31 Buhr,.J.,‘Whitson, R., Brewer, R., and Drew, D., "Traffic Characteristics for Implementation and Calibration of Freeway Merging Control Systems", 531 Reg Reg 229, 1969, pp. 87-106 Drew, D. , "Gap Acceptance Characteristics for Ramp-Freeway Surveillance and Control", Hwy Res Rec 152, 1967, pp. 108-143 anda,IL, "An Analysis of Short-Term Implementation of Ramp Control on the Dan.Ryan Expressway", H21 Res Rec 279, 1969, p. 157 Gerlough, D., "Simulation of Traffic Flow", Highway Research 0 ecial Report 79, 1964, pp. 97-188 Gervais, E. , "Optimization of Freeway Traffic by Ramp Control", Hwy Reg Reg :9, 1964, pp. 104-118 Goldblatt, R”, ”Development and.Testing of INTRAS, A.Microscopic Simulation.Model", FHWA-RD-76-78, Vol, 2: User's Manual, Final Report. 1980. Goldblatt, R”, Hagerty, B., and Laban, T., "TRAFLO-M Macroscopic Traffic Simulation Model", Final Report, MDOT, EHEA-MI-RD-8S- 91, 1984. Garden, L., "A Concept of Deterrence Controls for Freeway to Freeway Interchanges", Traffic Engineering, Institute of Traffic Engineerg, V01 42, No 11, 1972, pp. 23-27 Highway Research Board, I'Highway Capacity Manual" , Special Report-87, Washington, D.C., 1965. Levy, 8., Glickstein, A., and Finley, L., "Application of Computer Simulation Techniques to Interchange Design PrOblems", Highway Research Board Bulletin, No, 291, 1961, pp. 139-162 May, A. , "Experimentation with Manual and Automatic Ramp Control", Hwy ReggRec 59, 1964, pp. 9-38 190 Munjal, P., Hsu, Y., and Lawrence, R., "Prediction and Stability of Freeway Gaps and On-Ramp Merging", any Res Rec 469, 1973, pp. 26-39 Newman, L., Dunnet, A. and Mies, G., "An Evaluation of Ramp Control on the Harbor Freeway in Los Angeles", Hyy Res Rec 222, 1969, pp. 44-55 Payne, H. , "FREFLO: A Macroscopic Simulation Model of Freeway Traffic", Irans, Reg Rec 222, 1979, pp. 68-76 Posner, E. , "A Model to Predict Benefits of Priority Lanes on Freeways", Iransnogtation §cienceJol. 10. No. 2, 1976, pp. 169-204 Ross, P., and Gibson, D., "Review of Road Traffic Network Simulation Models", Transportation Reseagch Record 644, 1977, pp. 36-41 Russel, G., “Ramp Control on Freeways in California", Hny Res Rec 212, 1969, pp. 153-154 Sakashita, M. , Lu, 6., and May, A., "Evaluation of Freeway Emergency Service Systems Using a Simulation Model" , Hwy Res M: 1971, pp‘ 26-45 Sinha, K. , and Dawson R. , "The Development and Validation of a Freeway Traffic Simulater", Hyry Res Rec 308, 1970, pp. 34-47 Tignor, S. , ”Operational Analysis of Freeway Moving-Merge Systems", Transportation Research Board, Trans ReLRec 533, 1975, pp. 1-21 Wattleworth, J., Pinnell, C., McCsland, W., and Drew, D., "Evaluation of Entrance Ramp Control on a Six-Mile Freeway Section', Hm Res Rec 152, 1967, pp. 22-76 Wattleworth, J., Wallace, C. , and Levine, M. , "Development and Evaluation of a Ramp Metering System on the Lodge Freeway", Iezng Transportation Institute, Research Report 488-3, 1967, pp. 1-46 Wicks, D. , and Lieberman, E. , ”Development and Testing of INTRAS, A Microscopic Simulation Model", FHWA-RD-76-78. Vol. 1 Program Design, Parameter Calibration, and Freeway Dynamics Components Development. Final Report. 1977. Wherry, L., A telephone interview with Mr. Larry Wherry, CALTRAN, City of San Diego, October 23, 1987. HICHI u\\Imflfl‘flmgmijlgiifllflwjifljiifil‘s 31