$3:
a .

. n M?! a
fin“. mm.

W“;

a m...
knumwwfi

. .v .1. .
1%.fim._é
,8. H :
F.u..rx.nu..,u.w9m.?

5.. .

. 3, .2. .1
a. 32,.
1...}.1‘

T.
.. ..

Ly;
a

1!

2.11.1
'v n I.» .

 

@321. .

 

 

 

 

13m
'4
Zoo:

{MM/Mo”

 

LIBRARY
M'Ch'gafi .State This is to certify that the
UHIVGI’SIty dissertation entitled

 

 

 

EVALUATION OF VARIABLE SPEED LIMITS IN WORK ZONES

presented by

Disapat Lavansiri

has been accepted towards fulfillment
of the requirements for the

PhD. degree in Civil Engineering

 

 

WW 4:, “flu/A
Major Professd’s Signature

/2 lfl/é 2M3

 

Date

MSU is an Affirmative Action/Equal Opportunity Institution

—.-o—---.-.-.-.----¢—a-.-n-o-a-o-o-o-Q-o-o-o-o—-— -

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

K.
APR 03,2005

 

9.51 914.0. 6;

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c:/CIRC/DateDue.p65-p.15

 

 

EVALUATION OF VARIABLE SPEED LIMITS IN WORK ZONES
By

Disapat Lavansiri

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

2003

ABSTRACT
EVALUATION OF VARIABLE SPEED LIMITS IN WORK ZONES
By

Disapat Lavansiri

The use of variable speed limits (VSL) permits the posted speeds to be
determined based on current conditions and to vary by time and location within work
zones. This research evaluated the effectiveness of the VSL system under various work
zone conditions based on data collected during a field test in work zones in Michigan.
The measures of effectiveness used in this evaluation included changes in average speed,
travel time through the work zone, 85‘h percentile speed, speed variance, and percentage
of higher speed vehicles. MOE data “before,” “dun'ng,” and “after” the VSL operation
were compared.

The results indicated that the VSL system worked well at locations in the work
zone where there were no ramps in the vicinity. Relatively large increases in the
displayed speeds (8-10 mph) at these locations resulted in small increases in the average
speeds (1-2 mph), which in turn resulted in the reductions in the travel times through the
work zone. The percentage of higher speed vehicles at most of the locations either stayed
the same or became lower “during” the VSL deployments, which indicates an
improvement in speed limit compliance. However, at most trailers the 85‘h percentile
speeds did not change in the same direction as the changes in the displayed speeds as

hypothesized. The impact of the system on speed variance was inconsistent.

Although the changes in the average speed and the travel time following the
deployment of VSL were small, they were in the right direction. The VSL did respond to
the day-to-day and trailer—to-trailer average Speed variations, and thus the credibility of

the posted speed limit was improved.

ACKNOWLEDGEMENTS

I would like to thank my advisor, Dr. William Taylor, for his invaluable advice
and support throughout my graduate student years at Michigan State University and for
giving me an opportunity to work on the variable speed limit project.

I would also like to thank Dr. Richard Lyles, my committee member, for giving
me an opportunity to work on the variable speed limit project and for his support and
direction throughout this project.

I would like to give thanks to Dr. James Hannan and Dr. Ghassan Abu-lebdeh, my
committee members, for their professional suggestions.

I would like to acknowledge the Federal Highway Administration and the
Michigan Department of Transportation for providing a financial support for this research.

Last but not least, I would like to Show my gratitude to my parents, Direk and

Supatra Lavansiri, who have always given me unconditional support and encouragement.

iv

TABLE OF CONTENTS

LIST OF TABLES ............................................................................. viii
LIST OF FIGURES ............................................................................ xi
CHAPTER 1 INTRODUCTION ............................................................ 1
1.1 Description ofthe problem 1
1.2 Objective and scope Of the research ................................................... 1
1.3 Research approach........................... 2
1.4 Measures of effectiveness ................................................................. 3
CHAPTER 2 LITERATURE REVIEW ..................................................... 5
2.1 Speed-accident relationship ............................................................ 5
2.2 Work zone speed limits setting and enforcement ................................... 10
2.2.1 Work zone speed limits setting ....................................................... 10
2.2.2 Work zone speed limit enforcement ................................................ 15
2.3 Previous application of VSL .......................................................... 17
2.3.1 Weather ............................................................................... 17
2.3.2 Grade .................................................................................. 21
2.3.3 Traffic flow ........................................................................... 23
2.3.4 Work zones ........................................................................... 25
2.3.5 Combination of two factors or more .............................................. 26
2.3.6 Conclusions .......................................................................... 30
CHAPTER 3 DESIGN OF THE STUDY .................................................. 33
3.1 VSL system details ..................................................................... 33
3.1.1 System components ................................................................. 33
3.1.2 System configuration ............................................................... 34
3.1.3 System operation .................................................................... 36
3.1.4 Operational parameters ............................................................. 36
3.2 Data collection methods ............................................................... 38
3.2.1 Remote traffic microwave sensor (RTMS) ...................................... 38
3.2.2 Pneumatic tube system ............................................................. 39
3.3 Study plan ................................................................................ 41
3.3.1 Pre-deployment test ................................................................. 41
3.3.2 VSL deployment in work zones ................................................... 42
3.4 Speed-limit-setting algorithm ......................................................... 48
3.4.1 First deployment ..................................................................... 49
3.4.2 Second deployment ................................................................. 51
3.4.3 Third deployment ................................................................... 52
3.4.4 Fourth deployment .................................................................. 53

CHAPTER 4 RESULTS OF DATA VERIFICATION AND PROBLEMS AND

LIMITATIONS FACED IN THE VSL DEPLOYMENT ................................ 56
4.1 Results of data verification ............................................................. 56
4.1.1 Remote Traffic Microwave Sensor (RTMS) .................................... 56
4.1.2 Pneumatic tube system ............................................................. 62
4.1.3 85th percentile Speed estimation .................................................. 70
4.2 Problems and limitations faced in the VSL deployment ........................... 72
4.2.1 Pre-deployment test ................................................................. 73
4.2.2 First deployment ..................................................................... 74
4.2.3 Second deployment ................................................................. 78
4.2.4 Third deployment ................................................................... 80
4.2.5 Fourth deployment .................................................................. 81
4.2.6 Conclusion ........................................................................... 82
CHAPTER 5 DATA ANALYSIS AND RESULTS ...................................... 83
5.1 Average speed ........................................................................... 85
5.1.1 First deployment ..................................................................... 85
5.1.2 Third deployment ................................................................... 93
5.1.3 Fourth deployment .................................................................. 97
5.1.4 Average speed analysis conclusion ................................................ 100
5.2 Travel time through the work zone ................................................... 101
5.2.1 First deployment ..................................................................... 102
5.2.2 Third deployment ................................................................... 103
5.2.3 Fourth deployment .................................................................. 104
5.2.4 Travel time analysis conclusion ................................................... 106
5.3 85th percentile speed .................................................................. 106
5.3.1 First deployment .................................................................... 106
5.3.2 Third deployment ................................................................... 108
5.3.3 Fourth deployment .................................................................. 110
5.3.4. 85th percentile speed analysis conclusion ....................................... 110
5.4 Speed variance .......................................................................... 111
5.4.1 First deployment .................................................................... 111
5.4.2 Third deployment ................................................................... 113
5.4.3 Fourth deployment .................................................................. 114
5.4.4 Speed variance analysis conclusion ............................................... 114
5.5 Percentage of higher speed vehicles .................................................. 115
5.5.1 First deployment ..................................................................... 116
5.5.2 Third deployment ................................................................... 119
5.5.3 Fourth deployment .................................................................. 125
5.5.4 Percentage of higher speed vehicles analysis conclusion ...................... 125
5.6 Effect of police presence ............................................................... 125
CHAPTER 6 CONCLUSIONS AND RECOMMENDATIONS ........................ 130
6.1 Conclusions .............................................................................. 130
6.2 Recommendations ...................................................................... 132

vi

BIBLIOGRAPHY .............................................................................. 135

vii

LIST OF TABLES

Table 2.1.I-Iighway visibility range criteria for VMS in Utah (Perrin et a1., 2000)... 18
Table 2.2. Speed-limit-setting algorithm for Washington State (Robinson, 2000). .. 19
Table 2.3. Speed-limit—setting algorithm for Finland (Robinson, 2000) ................ 20
Table 2.4. Speed-limit-setting algorithm for Breda-Netherlands (Robinson, 2000).. 21

Table 2.5. DTSWS advised speed and mean truck speeds with the system on and

off (Janson, 1999) .............................................................................. 22
Table 2.6. Revised DTSWS advised speed (Janson, 1999) .............................. 22
Table 2.7. Original Speed-limit-setting algorithm for UK. (Nuttall, 1995) ........... 24
Table 2.8. Environmental Constants for New Mexico (W ilkie, 1997) ................. 28
Table 2.9. Speed limit (km/h) setting algorithm for Germany (Warren, 2000) ....... 29
Table 3.1. Speed-limit-setting profiles ...................................................... 49
Table 3.2. Speed-limit-setting algorithm for the first deployment ...................... 50
Table 3.3. Speed-limit-setting algorithm for the second deployment .................. 51
Table 3.4. Speed-limit-setting algorithm for the third deployment ..................... 53
Table 3.5. Speed-limit-setting algorithm for the fourth deployment .................... 54
Table 4.1. Pneumatic tube average speed verification summary ........................ 66
Table 4.2. Pneumatic tube 30-minute volume verification summary .................. 67
Table 4.3. Problems and limitations faced in the VSL deployment ..................... 72
Table 4.4. Examples of incorrect occupancies ............................................. 75
Table 5.1. Available pneumatic tube data .................................................. 84
Table 5.2. Results of the F-tests on the average speeds from the first deployment... 89

viii

Table 5.3. Results of the F-tests on the speed reductions at the crossover from the
first deployment ................................................................................

Table 5.4. Results of the F-tests on the average speeds from the third deployment..
Table 5.5. Results of the F-tests on the average speeds from the fourth deployment

Table 5.6. Results of the F-tests on the travel times through the work zone from
the first deployment ...........................................................................

Table 5.7. Results Of the F—tests on the travel times through the work zone from
the third deployment ...........................................................................

Table 5.8. Results of the F-tests on the travel times through the work zone
from the fourth deployment ..................................................................

Table 5.9. Results of the Mann-Whitney tests on the 85th percentile speeds
“before” and “during” the first deployment ................................................

Table 5.10. Results of the Mann-Whitney tests on the 85th percentile speeds
“before” and “during” the third deployment ...............................................

Table 5.11. Results of the Mann-Whitney tests on the 85th percentile speeds
“during” and “after” the third deployment ..................................................

Table 5.12. Speed limits “during” the VSL deployments ................................

Table 5.13. Results of the “before” versus “during” percentage of vehicles with a
speed higher than 60 mph and 70 mph comparisons from the first deployment ......

Table 5.14. Results of the Mann-Whitney tests on the percentage of vehicles with
a speed higher than 60 mph and 70 mph “before” and “during” the first
deployment .....................................................................................

Table 5.15. Results of the “before” versus “during” percentage of vehicles with a
speed higher than 60 mph and 70 mph comparisons from the third deployment.

Table 5.16. Results of the “during” versus “after” percentage of vehicles with a
speed higher than 60 mph and 70 mph comparisons from the third deployment.

Table 5.17. Results of the Mann-Whitney tests on the percentage of vehicles with
a speed higher than 60 mph and 70 mph “before” and “during” the third
deployment .....................................................................................

Table 5.18. Results of the Mann-Whitney tests on the percentage of vehicles with
a speed higher than 60 mph and 70 mph “during” and “after” the third deployment

ix

93

96

99

102

104

105

107

109

109

115

118

119

122

122

123

123

Table 5.19. Times that the patrol vehicles were present in the fourth deployment. .. 126

LIST OF FIGURES

Figure 1.1. I-96 work zone site ............................................................... 3

Figure 2.1. Accident involvement rates with and without turning vehicles (West

and Dunn, 1971) ............................................................................... 7
Figure 2.2. Accident involvement rates by variation from mean speeds

(Committee for Guidance on Setting and Enforcing Speed Limits, 1998) ............ 8
Figure 2.3. Work zone speed limit procedure flowchart (GME, 1996) ................ 13
Figure 3.1. VSL master trailer ............................................................... 34
Figure 3.2. RTMS placement (IRD, 2002) ................................................. 35
Figure 3.3. RTMS microwave beam and its footprint (E18, 2000) ..................... 38
Figure 3.4. Pneumatic tube system............... ........................................... 40
Figure 3.5. Location of VSL trailers during the pre-deployment testing ............... 42

Figure 3.6. Example of the variations in day-to-day and trailer-to-trailer average

speeds from 5PM to 6PM “before” the first deployment ................................. 43
Figure 3.7. Map of the first deployment work zone ....................................... 44
Figure 3.8. Map of the second deployment work zone ................................... 45
Figure 3.9. Map of the third deployment work zone ...................................... 46
Figure 3.10. Map of the fourth deployment work zone ................................... 47
Figure 3.11. VSL site layout for the first deployment .................................... 50
Figure 3.12. VSL site layout for the second deployment ................................. 51
Figure 3.13. VSL site layout for the third deployment .................................... 52
Figure 3.14. VSL site layout for the fourth deployment ................................. 54
Figure 4.1. Volume comparisons at trailer 3 for the first RTMS data verification... 57
Figure 4.2. Speed comparisons at trailer 3 for the first RTMS data verification. 57

xi

Figure 4.3. Volume comparisons at trailer 3 for the second RTMS data
verification ......................................................................................

Figure 4.4. Speed comparisons at trailer 3 for the second RTMS data verification...

Figure 4.5. Volume comparisons at trailer 1 for the third RTMS data verification...
Figure 4.6. Speed comparisons at trailer 1 for the third RTMS data verification.
Figure 4.7. Volume comparisons at trailer 2 for the third RTMS data verification...
Figure 4.8. Speed comparisons at trailer 2 for the third RTMS data verification. . . ..
Figure 4.9. Volume comparisons at trailer 3 for the third RTMS data verification...
Figure 4.10. Speed comparisons at trailer 3 for the third RTMS data verification...

Figure 4.11. The difference between the RTMS and pneumatic tube average speed
“before” and “during” the first deployment ................................................

Figure 4.12. The difference between the RTMS and pneumatic tube average speed
“before” and “during” the third deployment ...............................................

Figure 4.13. Pneumatic tube versus RTMS average speed comparison at trailer 1
from 10AM to 11AM in the first deployment .............................................

Figure 4.14. Pneumatic tube versus RTMS average speed comparison at trailer 1
from 10PM to 11PM in the first deployment ..............................................

Figure 4.15. 85th percentile speed and average speed plus 5 mph comparison from
trailer 3 before the first deployment (30-speed bin pneumatic tube data) ..............

Figure 4.16. 85th percentile speed and average speed plus 5 mph comparison from
trailer 3 during the first deployment (l6-speed bin pneumatic tube data) .............

Figure 4.17. Examples of data that repeat themselves ....................................
Figure 4.18. Raised short-range modem antenna at trailer 6 .............................

Figure 5.1. Average speed profiles between 6:00AM and 8:30AM from the first
deployment .....................................................................................

Figure 5. 2. Average speed profiles between 10: 30AM and 12: 30PM from the first
deployment... .

xii

58

58

59

59

60

60

61

62

65

66

69

69

70

71

75

76

86

87

Figure 5.3. Average speed profiles between 4:00PM and 6:00PM from the first
deployment ..................................................................................... 87

Figure 5.4. Average speed profiles between 8:00PM and 10:00PM from the first
deployment ..................................................................................... 88

Figure 5.5. Average speed and average displayed speed profile comparison from
5:00PM to 6:00PM “during” the first deployment ........................................ 91

Figure 5.6. “Before” and “during” speed reduction at the crossover comparison
from the first deployment ..................................................................... 92

Figure 5.7. Average speed profiles between 10:30AM and 12:30PM from the third

deployment ..................................................................................... 94
Figure 5.8. Average speed profiles between 5:00PM and 6:00PM from the third
deployment ..................................................................................... 94
Figure 5.9. Average speed profiles between 8:00PM and 10:00PM from the third
deployment ..................................................................................... 95
Figure 5.10. Average speed and average displayed speed profile comparison from
2:00PM to 3:00PM “during” the third deployment ....................................... 97
Figure 5.11. Average speed profiles between 10:30AM and 12:30PM from the
fourth deployment ............................................................................. 98
Figure 5.12. Average speed profiles between 8:00PM and 10:00PM from the
fourth deployment ............................................................................. 99
Fi gure 5.13. Average speed and average displayed speed profile comparison from
8 : 00PM to 9:00PM “during” the fourth deployment ...................................... 100
F i gure 5.14. “Before” versus “during” speed variance comparison between
1 2:00AM and 12:00PM from trailer 1 (right lane) in the first deployment ............ 112
F i gure 5.15. “Before” versus “during” speed variance comparison between
1 2:00PM and 12:00AM from trailer 3 in the first deployment .......................... 112
1hi gure 5.16. “Before” versus “during” speed variance comparison between
3 : 00PM and 12:00AM from trailer 7 in the third deployment ........................... 113
Fi gure 5.17. “During” versus “after” speed variance comparison between 8:00PM
and 7:00AM from trailer 4 in the third deployment ....................................... 114
xiii

Figure 5.18. “Before” versus “during” percentage of vehicles with a speed higher
than 60 mph comparison between 12:00PM and 12:00AM from trailer 1 (right
lane) in the first deployment .................................................................. 117

Figure 5.19. “Before” versus “during” percentage of vehicles with a speed higher
than 70 mph comparison between 12:00PM and 12:00AM from trailer 1 (right
lane) in the first deployment .................................................................. 117

Figure 5.20. “Before” versus “during” percentage of vehicles with a speed higher
than 60 mph comparison between 12:00PM and 12:00AM from trailer 2 in the
third deployment ............................................................................... 120

Figure 5.21. “Before” versus “during” percentage of vehicles with a speed higher
than 70 mph comparison between 12:00PM and 12:00AM from trailer 2 in the
third deployment ............................................................................... 120

Figure 5.22. “During” versus “after” percentage of vehicles with a speed higher
than 60 mph comparison between 8:00PM and 12:00AM from trailer 7 in the third
deployment ..................................................................................... 121

Figure 5.23. “During” versus “after” percentage of vehicles with a speed higher
than 70 mph comparison between 8:00PM and 12:00AM from trailer 7 in the third
deployment ..................................................................................... 121

Figure 5.24. Effect of police presence on the average speed between 10:00AM and
2:00AM from trailer 1 in the fourth deployment .......................................... 127

Figure 5.25. Effect of police presence on the speed variance between 10:00AM
and 2:00AM from trailer 1 in the fourth deployment ..................................... 128

Figure 5.36. Effect of police presence on the percentage of vehicles with a speed

hi gher than 60 mph between 10:00AM and 2:00AM from trailer 1 in the fourth
deployment ..................................................................................... 128

xiv

Chapter 1

INTRODUCTION

1.1 Description of the problem

In work zones, static speed limits are sometimes not credible for the motorists
because they cannot reflect up-to—date roadway conditions, traffic flow conditions, and
construction activities. For example, there are many long work zones marked with low
speed limits where no construction activity occurs until miles after the start of the work
zone. The use of variable speed limits (VSL) permits the posted speeds to be determined
based on current conditions and to vary within the Site, which may result in higher speed
limit compliance and lower speed variance. Consequently, VSL may reduce congestion
and improve safety in work zones. However, the use of VSL in work zones has not been
thoroughly studied. In this study, the implementation of VSL in work zones was tested to
determine their effectiveness under various conditions.
1.2 Objective and scope of the research

A VSL system designed to improve traffic flow and safety in work zones was
evaluated in this research. The scope of the research includes:

a. identification of measures of effectiveness (MOEs) to quantify the impact of the

VSL system; and

b. evaluation of the impacts of the VSL system on these MOEs under various work
zone conditions.

1.3 Research approach
This research is based on data collected during a field test of variable speed limits
in work zones in Michigan. This project was sponsored by the Federal Highway
Administration (FHW A) and the Michigan Department of Transportation (MDOT). The
VSL system was tested under various traffic conditions, a variety of work activities, and
different types of traffic separations. The hourly traffic conditions ranged from relatively
low volume to congested conditions. The work activities and traffic separations ranged
from minor work (concrete patching) with exposed workers separated from adjacent
traffic only by barrels to intense work (total reconstruction) with workers separated from
adjacent traffic by a median. The four work zones used in this study were part of the one
large work zone, which was located on I-96 from US-127 to Wacousta Road as shown in
figure 1.1 (Images in this dissertation are presented in color.). This part of I-96 skirts the
southern and western edges of Lansing, Michigan. Different speed limit setting
algorithms were developed for each work zone. The MOE data were collected “before”,
“during”, and “after” the VSL deployment (although not in all three time periods in each
zone). The data “before” and “after” each VSL deployment were used as a base case. For
the base case, the speed limit was posted at a constant 50 mph. The MOEs in the
“before,” “during,” and “after” conditions are compared and discussed in the following

chapters.

 

 

 

 

 

 

 

 

 

/
/
//
A m
1" ’/ I 'JLL'J, 11.5 f’ ‘
' A," 1" Hubsofl Corp 0m grammar. ‘

 

 

 

 

Figure 1.1. [-96 work zone site
1.4 Measures of effectiveness
The variables measured or calculated to determine the effectiveness of the VSL
system are:

a. Average speed: The average speed at locations within the work zone where the
VSL signs were placed is used as one MOE. Changes in the average speed at
these locations are one measure of a change in the motorists’ behavior.

b. Travel time through the work zone: The travel time through the part of the work
zone where the VSL system was located is also used as an MOE. This travel time

is estimated from the average speeds. A decrease in travel time in the part of the

work zone where there was no construction activity is expected to be one of the
benefits from the VSL system.

85th percentile speed: The 85‘h percentile speed at locations within the work zone
where the VSL signs were placed is used as an MOE because it is the variable that
traffic engineers use to set the speed limit in speed zones, and it is also one of the
variables used to set the variable speed limit in this project.

. Speed variance: The speed variance at locations within the work zone where the
VSL signs were placed is used as an MOE. This is an indirect measure of safety
since the accident involvement rate has been hypothesized to increase with an

increase in speed variance.

. Percentage of higher speed vehicles: Since most of the speed limits in this study

ranged from 50 mph to 70 mph, percentage of vehicles with a speed higher than
60 mph and 70 mph at locations within the work zone where the VSL signs were
placed are used as an MOE. The percentage of higher speed vehicles is used as a

surrogate measure of speed limit compliance.

Chapter 2

LITERATURE REVIEW

The literature on speed characteristics and accident relationship will be reviewed.
Then, the mechanism for setting and enforcing speed limits in work zones, particularly in
VSL-controlled work zones, will be reviewed. Finally, previous applications of VSL will
be discussed.
2.1 Speed-accident relationship

Many studies have been conducted on the relationship between speed and
accident involvement. The most widely referenced study was conducted by David
Solomon in 1964. Solomon (1964) documented the speed and characteristics on a sample
of vehicles and drivers involved in accidents on two- and four-lane main rural highways
in 11 states. This study was based on the accident records of nearly 10,000 drivers, speed
Observations, and interviews with 290,000 drivers. Solomon found a U-shaped
relationship between crash involvement rates and travel speeds with the lowest crash
involvement rate at a speed slightly above the average speed. He indicated that “the
greater the variation in speed Of any vehicle from the average speed of all traffic, the
greater its chance of being involved in an accident” and that “the severity of accidents

increased as speed increased, especially at speeds exceeding 60 mph.” However, critics

of Solomon’s study question the dependence on police accident reports for the pre-crash
speeds of the accident-involved vehicles, since these can be unreliable.

West and Dunn (1971) conducted a study on accidents, speed deviation, and
speed limits on Indiana Highway 37. This study was based on the average speeds from
magnetic loop detectors and the estimated speeds of accident-involved vehicles from
professional accident investigators. The accuracy of the speed estimate in Solomon’s
study was addressed by recording the pre-crash speeds of the accident-involved vehicles
at the time of the crash. The results showed that the involvement rate was higher for the
vehicles deviating on the slow side than for the vehicles deviating on the fast side.
However, after the accidents involving turning vehicles were deleted, the involvement
rate was approximately the same for high and low speed deviations and the involvement
rate was lowest when the vehicles traveled close to the average speed. The U-shaped
relationship (figure 2.1) became flatter than Solomon’s because the elevated accident
involvement rates for the slower vehicles were reduced. Harkey, Robertson, and Davis
(1990) also found the U-shaped relationship between crash involvement rate and
deviation from the average speed with the lowest involvement rate at 7 mph above the
mean speed or at about the 90th percentile of the travel speeds observed. This study was
based on two-lane and multi-lane highways in North Carolina and Colorado. The
accidents considered in this study were limited to weekday, non-alcohol, and non-
intersection involved.

A more recent study by Aljanahi, Rhodes and Metcalfe (1999) investigated the
relationship between traffic speed characteristics and personal injury accidents under free

flow conditions. This study was based on accident records in Tyne and Wear County Of

the United Kingdom, and in Bahrain. Only non-intersection accidents not involving
pedestrians were considered. Aljanahi et a1. indicated that in Bahrain, there is
statistically significant evidence of an association between mean speed and accident rate,
while in Tyne and Wear the statistical relationship is weaker and points to a stronger

relationship between accident rate and the variability of traffic speed.

45.0 1
35.0
30.0
25.0 .
20.0
15.0 .
10.0
i
l

<-15.5 -15.5 to-5.5 -5.5 to+5.5 +5.5 to+15.5 >+15.5

 

 

crash involvement rate per
million vehicle miles

 

 

 

 

 

deviation from average speed (mph)

 

 

‘1
fr— -, W'tmm'ssvchs'e?1i".:_W;¢92Lt9rngv§_vsb!slssl ‘

Figure 2.1. Accident involvement rates with and without turning vehicles
(West and Dunn, 1971)

All of these studies were conducted on non-limited access highways. A study
similar to Solomon’s was done on interstate freeways in 20 states by Cirillo (1968).
Cirillo indicated that as the speed of a vehicle deviates from the mean speed of traffic,
either above or below the mean speed, the chance of the vehicle being involved in an
accident increases. The lowest accident involvement rate occurred at 12 mph above the
mean speed. The comparison of the accident involvement rate by deviation from mean

speed between Solomon’s and Cirillo’s is Shown in figure 2.2.

 

 

 

 

      
 
 
    

 

I I T r I F I
50,000 w Solomon Nighttime (rural)
Solomon Daytime (rural)

U, ......... Cirillo Freeways
O
5 10,000 , . . _. _.
a .
.2
E Frldes et a1.
§ (urban) !
s 5‘
Q ‘~ 5
a ’
E Fildes et a1”: 1"
l; (rural) f
O 100 n, .... .\.. ... . .....J’
> \‘
a \ z
a ‘\ ;
S a“? .f
O \J’

L k i I.

 

 

U U I T TI I
-40 -30 -20 -10 0 10 20 30 40
Deviation from Average Speed, mph

Figure 2.2. Accident involvement rates by variation from mean speeds
(Committee for Guidance on Setting and Enforcing Speed Limits, 1998)

Garber and Gadiraju (1989) performed a study on “factors affecting speed
variance and its influence on accidents” on 36 locations of interstate freeways, arterial
highways, and collector roads in Virginia from 1983 through 1986. Traffic data were
collected by a traffic recorder, while accident data were obtained from the Virginia
Department of Transportation and Virginia Department of Motor Vehicles. Garber and
Gadiraju concluded that “accident rates increase with increasing speed variance for all
classes of roads" and that “the accident rate on a highway does not necessarily increase
with an increase in average speed.” They also suggested that “speed variance will
approach minimum values if the posted speed limit is between 5 and 10 mph lower than

the design speed.” Lave (1985) investigated the relationship between speed and the

fatality rate based on the accident records in 1981 and 1982 on interstates, arterials and
collectors in all of the states, except Alaska and Hawaii. The results indicated that there is
no statistically discemable relationship between the fatality rate and average speed,
though there is a strong positive relationship between fatality rate and speed variance.

There are also a few studies where conflicting results were found. Fildes,
Rumbold, and Leening (1991) examined crash involvement rates as a function of speed
on urban arterials and two-lane rural roads in Australia. The results indicated that the
accident involvement rate has a positive linear relationship with travel speed instead of
the U-shaped relationship found in previous studies (figure 2.2). The accident
involvement rates were lowest at speeds below the average speed and highest at speeds
above the average speed. The problems with this study were that the results were based
on a small sample size and self-reported crash involvement and speed. Another
Australian study by Kloeden, McLean, and Moore (1997) partly supported the result from
Fildes et a1. Kloeden et al. found a positive relationship between the probability of
involvement in a casualty crash and speeds above, but not below the average speed. The
study focused on weekday, daylight accidents, excluding alcohol-related crashes in the
Adelaide metropolitan area. The Committee for Guidance on Setting and Enforcing
Speed Limits (1998) indicated that the absence of a significant association between speed
and crash involvement at speeds below the average traffic speed may well be the result of
the study design that excluded all but injury crashes, since crashes at lower speeds tend to
be less severe.

Although the results from all of the studies reviewed are not completely

consistent, most studies found that there is a U-shaped relationship between crash

V-i-e‘ 3""

involvement rates and travel speeds with the lowest crash involvement rate at speeds
slightly above the average speed and that the accident involvement rate increases with an
increase in speed variance, but does not necessarily increase with an increase in average
speed.

Because of the short duration of each VSL deployment in this study, it was not
possible to conduct a safety analysis.

2.2 Work zone speed limits setting and enforcement
2.2.1 Work zone speed limits setting

A major problem in setting work zone speed limits is the lack of uniform
guidelines for determining the appropriate speed limit in any situation. The Manual on
Uniform Traffic Control Devices (MUTCD, 2001) only presents a general guideline for
determining speed limits. The MUTCD states that “when a speed limit is to be posted, it
should be the 85th percentile speed of free flowing traffic, rounded up to the nearest 10
km/h (5 mph) increment. Other factors that may be considered when establishing speed
limits are the following: A. road characteristics, shoulder condition, grade, alignment, and
sight distance; B. the pace speed; C. roadside development and environment; D. parking
practices and pedestrian activity; and E. reported crash experience for at least a 12-month
period. ”

In an attempt to develop work zone guidelines, Graham-Migletz Enterprise, Inc.
(OMB, 1996) conducted a study for the National Cooperative Highway Research
Program (NCHRP) on “A Procedure for determining work zone speed limits.” The
research approach included a survey of state highway agencies to determine the existing

work zone speed limit policies, and an analysis of speed and accident data from 68 work

10

 

zone sites in seven states: California, Florida, Georgia, Iowa, Missouri, Montana, and
New York. The existing highway agency work zone speed limit policies and guidelines
were investigated by sending questionnaires to all 52 state highway agencies in the
United States. The results indicated that there were three general policies used by state
highway agencies: (1) eighteen states avoid speed limit reductions whenever possible
(Speed limits are set based on specific factors when they are necessary); (2) five states
use blanket speed limit reductions at all work zone sites; and (3) twenty-nine states
establish the need for a work zone speed limit based on specific factors. Among the seven
states where the analysis of speed and accident data were conducted, California, Florida,
and Iowa avoid speed limit reductions whenever possible, Georgia and Montana use
blanket speed limit reductions, and Missouri and New York establish the need for a work
zone speed limit based on specific factors. In the past, Michigan used a combination of
these approaches, with a 45-mph regulatory speed limit in work zones with lane closures,
and a flexible policy on speed limits where work was on or outside of the shoulder or in
work zones without lane closures.

The results from the speed data analysis conducted in the GME study indicated
that motorists reduced their speed in work zones, even in work zones with no speed limit
reduction. The average speeds in work zones with no speed limit reduction were
approximately 5 mph lower than they were upstream of the same work zone. The speed
limit compliance was greatest in work zones where the speed limit was not reduced, and
decreased when the speed limit was reduced more than 10 mph. The results from the
speed and accident data analysis showed that a speed limit reduction of 10 mph resulted

in the smallest increase in speed variance within the work zone relative to upstream of the

11

“...IIII‘IJ -'

Y
I

 

work zone and the smallest increase in the accident rate in the construction period relative

to pre-construction period. GME concluded that:

> “Work zone speed limit reductions should be avoided whenever possible, particularly
in work zones where all work activities are located on the shoulder or roadside areas
and when no work activities are underway.

> A IO-mph reduction below the normal speed limit is desirable when work takes place
on or near the travel way, particularly on rural freeways, or when personnel are
required to work for extended periods in an unprotected position within 10 feet of the
edge of the traveled way.

> Work zone speed limit reductions larger than 10 mph are undesirable and should be
avoided except where required by restricted geometric or other work zone features
that can not be modified.”

GME recommended a four-step work zone speed limit procedure: (1) determine
the existing speed limit; (2) determine the work zone condition that applies; (3) determine
which factors for the applicable work zone condition apply; and (4) select the work zone

speed limit. The work zone speed limit procedure flowchart is shown in figure 2.3.

12

: MW

 

 

Step 1

Determine the 5 Determine the

 

Step 2

 

 

existing speed work zone
limit condition that
a plies
Condition 1

 

 

 

Activities which are

 

Step 3

Determine which
factors for the
appropriate
conditions apply to
the specific site

 

 

 

 

Step 4
Select the work
zone speed limit

 

 

 

 

No speed limit
reduction

 

l9 more than 10 ft from the
edge of the traveled way
(Roadside activity)

 

 

 

Condition 2
Activities which encroach
upon the area closer than

ft to the edge of the
traveled way (Shoulder
activity)

 

 

Condition 3

Activities which encroach
upon the area from the
9 edge of the traveled way
to 2 ft from from the edge
of the traveled way

(lane encroachment)

 

 

Condition 4
Activities which require
_>‘ an intermittent or moving

  
   

-> 10 ft but not closer than 2 ‘9

 
   

  

. re any of the fact

for condition 2 present
0)

Yes

 
  
    

- re any of the fact
for condition 3 present

recommended except

 

in unusual situations

 

 

4‘

 

Speed limit may be
reduced by 10 mph

 

 

 

 

  

Speed limit may be
reduced by 10 mph

 

 

 

 

 

 

 

operation on the shoulder
(Moving activity on
shoulder)

 

 

 

Condition 5

Activities which encroach
upon the area between the
centerline and the edge of
the traveled way

(Lane closure)

 

 

 

Condition 6
Activities requiring a
acmporary detour to be
constructed

(Temporary detour)

 

 

 

  
   

 

Condition 7

Activities which encroach

upon the area on both sides
of the centerline of a .

9 .

roadway or lane line of a

multilane highway

(Centerline or laneline

 

 

  
   

 

encroachment)

 

 
   

-reany ofthefac v

for condition 5 present
?

 
  
    

- c any of the factr
for condition 6 present

 

No speed limit

reduction

recommended except
in unusual situations

 

 

4‘

 

  

 

Speed limit may be

reduced by 10 mph

 

 

 

  

Speed limit rmy be
reduced by 10 mph

 

 

 

 
   

: c any ofthe fact

for condition 7 present
?

No speed limit

‘ reduction

recommended except
in unusual situations

 

 

  

 

Speed limit may be
reduced by 10 mph

 

 

Figure 2.3. Work zone speed limit procedure flowchart (GME, 1996)

13

‘1 Flt-i “an. .5 a.“

Migletz et al. (1999) conducted a validation study of their 1996 work zone speed
limit procedure in 30 work zones in the same seven states. This study focused on work
zones where speed limits were either not reduced or where the speed limit was reduced
by 10 mph. The mean speeds, 85th percentile speeds, and speed variances in work zones
that implemented state highway agency policies were compared with work zones that
implemented the GME work zone speed limit procedure. The average reductions in mean
speed and 85‘h percentile speed from upstream to within the work zone in work zones
where the speed limit was based on the GME procedure were closer to the posted speed
limit reduction than work zones where the speed limit was based on highway agency
policies. Speed variance was higher in work zones than upstream of those same work
zones regardless of which procedure was used. In’work zones where the speed limit was
based on highway agency policies, the speed variances increased by 61 and 34 percent
for speed limit reductions of 0 and 10 mph respectively, while in work zones where the
speed limit was based on the GME procedure, the speed variances only increased by 16
and 10 percent for speed limit reductions of 0 and 10 mph respectively. The substantial
difference in speed variances between the work zones where the speed limit was based on
highway agency policies and the work zones where the speed limit was based on the
GME procedure was assumed to result from the more consistent speed-limit-setting
procedure using the work zone activities and locations. Migletz et a1. concluded that
“the application of the work zone speed limit procedure leads to more uniform speed and,
therefore, should result in safer and more efficient traffic operations.” However, it is not

clear why there would be a difference in the speed variance between locations where the

14

speed limit change was zero based solely on the procedure used to determine that no
change in the speed limit was required.
2.2.2 Work zone speed limit enforcement

Richards, Wunderlich, and Dudek (1985) evaluated several work zone speed
control techniques, including flagging, law enforcement, changeable message signs,
effective lane width reduction, rumble strips, and conventional regulatory and advisory
speed signing. The results indicated that flagging and law enforcement were effective
methods for controlling speeds in work zones. The best flagging treatment reduced
speeds an average of 19 percent, while the best law enforcement treatment reduced
speeds an average of 18 percent.

There were three types of law enforcement used in this study: a stationary patrol
car, a circulating patrol car, and a police traffic controller. The police traffic controller is
a uniformed officer standing on the side of road next to a speed sign and manually
motioning traffic to slow down. Among these three types of law enforcement, the
stationary patrol car and the police traffic controller were equally effective, while the
circulating patrol car was less effective. Although the police traffic controller was
effective, the authors concluded that it was not appropriate for a freeway due to the safety
concerns. The stationary patrol car performed slightly better with the light and radar on.

Ullman (1991) verified this finding, but he also found that a radar signal without
the presence of visible law enforcement personnel had only a small effect on average
speeds within and approaching work zones. However, he determined that a radar signal

had a slightly greater effect on vehicles approaching work zones at a speed greater than

15

65 mph, and on trucks because these two groups of vehicles tend to be equipped with a
radar detector.

The results from a study by Richard et a1. (1985) confirmed that enforcement is an
important factor in work zone speed limit control. A study of the effect of enforcement
strategy on traffic speeds by Hool et a1. (1983) confirmed some of the findings mentioned
before. Hool et al. reported that the marked stationary patrol vehicle was the most
effective speed control tactic and that the unmarked patrol vehicle, even when issuing
citations, had little effect on the speed.

The studies mentioned above concerned the enforcement of the speed limit in
work zones in general; the enforcement of variable speed limits will be described next.
One of the most successful VSL systems is located on the London orbital, M25. The
variable speed limits are mandatory and are enforced by photo radar. Harbord and Jones
(1996) indicated that if the motorists violate the speed limits, they will receive a ticket in
the mail with a photograph of their vehicle, which includes details of the time, date,
location, the vehicle speed, and the posted speed at the time of violation. Nuttall (1995)
reported that this enforcement resulted in a very high speed limit compliance rate, 98
percent, during the first three months of operation in 1995. After two years of operation,
Harbord (1998) indicated that there was no evidence that the level of compliance had
declined. However, he reported that a small number of drivers increase their speed
slightly between VSL sign gantries if the gantries are located significantly more than 1
km apart.

No study was found on the enforcement of variable speed limits in the United

States, but there is a study on judicial issues related to the enforcement of variable speed

16

limits by Hines (2002). This study examined the impact Of judicial decisions and judicial
enforcement on the likely success of enforcing a variable speed limit program. Hines
concluded that, “where there is a delegation of authority to an administrative agency with
appropriate limitations, and the agency acts within those limitations, speed limits that are
established may be enforced without fear that they will be subject to challenge.” She
also stated that legal issues that will arise from the enforcement of variable speed limits
should be no different from the legal issues that have been considered by courts in
adjudicating alleged violations of prima facie speed limits and other fixed maximum
Speed limits.
2.3 Previous application of VSL
The previous applications of VSL are discussed below in terms of the factors used
to set the speed limits.
2.3.1 Weather
> Arizona: The Fuzzy Variable Speed Limit Device (FVSLD) was installed on the I—40
corridor in rural northern Arizona. Placer (2001) indicated that the objective of the
FVSLD project was to develop a reliable system to detemrine a safe speed for the
current road and weather conditions and communicate this speed to drivers. The
speed limit was based on a fuzzy logic algorithm that approximates the human
reasoning process. According to Placer, this project was divided into 3 phases. In
Phase I, the fuzzy concept was validated and a software system that displayed prudent
highway speeds based on real-time weather and road surface conditions was
developed. Phase II was concerned with hardware and software upgrades and

equipment and sensor installations at the test site. Phase HI included data collection

17

regarding the road surface and weather information and confirmation of the status Of
enforcement and liability for Arizona. No results were published on this study.

Utah: Perrin et a1. (2000) reported on the use of the Adverse Visibility Information
System Evaluation (ADVISE) system on I-215 to reduce variability in speeds
between vehicles during reduced visibility events. The variable message signs (VMS)
are used to advise drivers of the appropriate speed for the current condition based on
the visibility information from visibility sensors, as shown in table 2.1. Penin et al.
indicated that the ADVISE system decreased the standard deviation of the speed by
22 percent and increased the overall mean speed by 15 percent on the average among
all of the visibility ranges. Penin et al. indicated that “this was accomplished by
advising the overly cautious drivers that the safe speed was faster than they initially
thought.”

Table 2.1. Highway visibility range criteria for VMS in Utah (Perrin et al., 2000)

 

 

 

 

m::::r:e::l;ty VMS message
> 250 No message
200 to 250 “Fog Ahead” alternating with “Poor Visibility”
150 to 200 “Max Speed 50” alternating with “Poor Visibility”
100 to 150 “Max Speed 40” alternating with “Poor Visibility”
60 to 100 “Max Speed 30” alternating with “Poor Visibility”
< 60 “Max Speed 25” alternating with “Poor Visibility”

 

 

Washington: A VSL system was installed on I-90 across Snoqualmie Pass. The
system consists of 13 light-emitting diode (LED) changeable message signs over 40
miles, but only 17 miles of the road are operated with a VSL during the winter

months. Robinson (2000) indicated that the objectives of the system are to improve

18

safety and to increase the availability of road condition and weather information to
motorists crossing the Snoqualmie Pass. The regulatory speed limits are set based on
roadway conditions as shown in table 2.2. The decision to reduce the speed limit is
based on feedback from multiple weather stations, snowplow operators, and the State
Patrol. A “Traction tire” (in table 2.2) is defined as a studded tire. A “traction tire”
may also be a tire that meets tests identified by the Rubber Manufacturers Association
(RMA) defining the tire as suitable for use in severe snow conditions. Ulfarsson et a1.
(2002) analyzed the effect of the VSL system on the relationship between mean
speeds and speed deviations. The relationships were examined both at a VSL site and
at a site close by. Ulfarsson et al. reported that the effect of VSL was to significantly
reduce mean speed and slightly increase the speed standard deviation, but the
magnitudes of the reduction in the mean speed and the increase in the speed standard
deviation were not reported. The results also suggested that drivers Show
compensatory behavior, drivers accelerated faster when exiting the VSL zone when
the speed limit was reduced than when it was not. Ulfarsson et al. indicated that this
compensatory behavior could increase accident frequency downstream of the work
zone, negating the safety benefits of lower mean speeds.

Table 2.2. Speed-limit-setting algorithm for Washington State (Robinson, 2000)

 

 

Roadway conditions Speed limit (mph)
Normal 65
Traction tires advised 55
Traction tires required 45
Chains required 35

 

 

 

 

19

> Australia: Twelve VSL signs were installed over 7 miles of the F6 Toll Way South of
Sydney. Robinson stated that the objective of this system was to reduce the number
of rear-end collisions in fog. The displayed speeds, which are based on the visibility
distance and the speed of the preceding vehicle, are advisory. No results were
published on this study.

> Finland: A VSL system was installed on E18 in Southern Finland. The system
consists of 67 VSL signs, 13 changeable message signs (CMS), and two unmanned
weather stations. Robinson stated that the objective of this system was to improve
safety without decreasing driver motivation to Obey speed limits. The regulatory
speed limits are based on roadway conditions as shown in table 2.3. Rama and
Luoma (1997) reported that “81 percent of the drivers said the posted speed limit was
appropriate and 95 percent of the drivers stated that the variable speed limits were
useful.” Rama (1998) also indicated that in winter a change of the posted speed limit
from 100 mm to 80 km/h decreased the mean speed of cars traveling in free-flow
traffic by 3.4 km/h, in addition to the average mean speed reduction of 6.3 km/h
caused by adverse weather and road surface conditions, and that the reduction of the
speed limit was accompanied by a decrease in the speed variance. However, the
values of mean speeds and speed variances before and after the speed limit reduction
were not reported.

Table 2.3. Speed-limit-setting algorithm for Finland (Robinson, 2000)

 

 

Roadway conditions Speed limit (km/h)
Good 120
Moderate 100
Poor 80

 

 

 

 

20

> Netherlands (Breda): A VSL system was installed on A-l6 near Breda to elicit safer
driving behavior during fog. The system consists of VSL signs spaced approximately
0.5 miles apart over 7.4 miles of road, 20 visibility sensors, and automatic incident
detection. The posted speeds are based on the visibility from sensors, as shown in
table 2.4. The posted speeds are also based on incidents. If an incident is detected, the
first and second signs upstream from the incident will display 50 and 70 mm
respectively. Robinson stated that this system resulted in a reduction in mean speed
during fog conditions of about 8 to 10 km/h. However, it is not clear whether the
observed reduction is relative to the condition without VSL or the condition without
fog.

Table 2.4. Speed-limit-setting algorithm for Breda-Netherlands (Robinson, 2000)

 

 

Visibility (feet) Speed limit (km/h)
> 140 100
140 to 70 80
< 70 60

 

 

 

 

2.3.2 Grade

> Colorado: Janson (1999) evaluated a Downhill Truck Speed Warning System
(DTSWS) in the Eisenhower Tunnel on I-70. In order to reduce runaway truck
accidents, the DTSWS advises vehicle-specific safe operating speeds for long
downgrades for trucks with greater than 40,000 lbs. gross vehicle weight. The
DTSWS consists of a weigh in motion (WIM) sensor, changeable message signs
(CMS), and inductive loop detectors. The displayed advisory speed for each truck is
calculated based on weight, speed, and axle configuration. The advised speeds and

the mean truck speeds with the system on and off are shown in table 2.5.

21

Table 2.5. DTSWS advised speed and mean truck speeds with the system
on and Off (J anson, 1999)

 

 

 

 

Weight range Advised With system on With system off
(1138.) Speed Mean speed # of trucks Mean speed # of trucks
(mph) (mph) (mph)

40,000 to 48,600 30 40.1 15 48.4 27
48,600 tO 51,200 25 30.5 43.5 3
51,200 to 55,500 20 37.6 4 34.9 3
55,500 to 63,500 15 32.8 41.5 5
63,500 tO 80,900 10 30.0 24 34.4 25
>80,900 5 16.9 1 25.6 1

 

 

 

 

 

 

Janson indicated that on the average the mean truck speed was reduced by 7.6

mph from 41.1 mph to 33.5 mph. However, he found that 2.4 mph of the 7.6 mph

change in speed was the result of an increased numbers of heavier trucks in the time

period when data were collected with the sign activated. He also indicated that truck

drivers surveyed responded very positively to the system and its potential to improve

safety. Based on the result of this study, Janson revised the advised speeds. The

revised advised speeds are shown in table 2.6.

Table 2.6. Revised DTSWS advised speed (J anson, 1999)

 

 

 

Weight range (le.) Advised speed (mph)
40,000 to 48,500 35
48,500 to 55,000 25
55,000 to 80,000 15
>80,000 10

 

 

 

22

 

> Oregon: The Oregon Department of Transportation installed a Downhill Speed
Information System (DSIS) on I-84 between Pendleton and La Grande to advise truck
drivers of the safe speed for traveling down Emigrant Hill. Montagne and Bell (2000)
stated that the purpose of this system is to reduce the frequency and severity of
downgrade truck accidents. The system consists of a WIM scale, CMS, and an
automatic vehicle identification (AVI) reader. The advisory Speed for each truck is
based on vehicle weight. Only properly weighed transponder-equipped trucks receive
an advisory speed. Bell (2001) indicated that the DSIS had not been deployed. No
more updated literature was found on the progress of this project.

2.3.3 Traffic flow

> Michigan: Wilkie (1997) documented the use of VSL in Michigan. Twenty-one VSL
signs were installed and operated over 3.2 miles of the John C. Lodge Freeway (M-
10) from 1962 through 1967. The objectives of this system were to warn motorists to
decelerate before reaching a congested area and to notify motorists leaving a
congested area to accelerate to help disperse the congestion. The variable speed
limits could range from 20 to 60 mph in increments of 5 mph. The speed limits were
set based on occupancy and volume data from the closed circuit television (CCTV)
images and real-time freeway speeds. The variable speed displays did not
Significantly increase or decrease vehicle speeds. This was presumed to be because
motorists did not understand variable speed displays with no accompanying message.

> United Kingdom: A VSL system was installed on the M-25 London orbital to smooth
traffic flows by reducing stop-start driving conditions. The system consists of VSL

display stations spaced approximately 0.6 miles apart over 14 miles, loop detectors at

23

0.3 mile intervals, and CCTV. Robinson (2000) indicated that the system is
intended to slow vehicles approaching a queue and that the system has additional
logic to prevent frequent fluctuations in speed limits. Speed limits were originally
changed according to the vehicle volume as shown in table 2.7. However, Harbord
(1998) indicated that this algorithm was replaced by a more sophisticated algorithm
after about the first month, but no document on the new algorithm has been found.

Table 2.7. Original Speed-limit—setting algorithm for UK. (Nuttall, 1995)

 

 

Volume (vphpl) Speed limit (mph)
< 1,650 70
1,650 to 2,050 60
> 2,050 50

 

 

 

 

As mentioned before, this system resulted in very high speed limit compliance
(98 percent) during the first three months of operation in 1995 (Nuttall, 1995). But
there were also a lot of complaints about very strict enforcement by speed cameras.
Harbord (1997) indicated that the VSL system also resulted in a 21 percent reduction
in injury accidents, more even lane usage, and more uniform headways. However, he
stated that the accident reduction percentage may have been influenced by roadwork
in the before period and that making an allowance for the effects of roadwork could
reduce the improvements in accidents to 11 percent. Harbord also performed a postal
survey of 1,600 drivers using the system. The results indicated that 60 percent of all
drivers reported that the system had resulted in an improvement, while only 10

percent believed it had made traffic congestion worse.

24

2.3.4 Work zones

> Arkansas: Tudor et al. (2003) reported the use of the Smart Work Zone Technology
system (ADAPT IRTM system) on 140 in Lonoke County to improve safety and
manage congestion in the work zone. The ADAPT IRTM system consisted of a central
system controller, two highway advisory radios (HARs), five traffic sensors, five
changeable message signs (CMSS), and two supplemental speed stations per lane
closure. The system was designed for the CMSs to display downstream traffic speed
information followed by delay information. If the speed differential between
successive traffic sensors exceeded 10 mph, the upstream CMS would display
“REDUCE SPEED TO XX MPH” and follow with the message “YY MINUTE
DELAY.” If the speed differential was below 10 mph, the CMS would only display
the delay information. The central system controller was programmed for a 10-minute
cycle time.

A problem encountered was that the public reported that the delay estimations
were not sufficiently accurate. Therefore, the delay messages were simplified to
display the generic messages, “EXPECT DELAYS” or “EXPECT LONG DELAYS.”
Tudor et a1. (2003) indicated that the ADAPT IRTM system resulted in a reduction in
the fatal crash rate of about one crash per 100 million vehicle miles traveled when
compared to two other work zones with similar traffic characteristics. However, this
conclusion was based on only two fatal crashes that occurred in the VSL site during
the study. Tudor et al. also reported that there was no significant reduction in the rear-

end crash rate.

25

> Minnesota: Robinson documented the use of VSL in work zones on urban freeways
in Minnesota. When construction workers are not present, the speed limit continues
to be 65 mph. When construction workers arrive, a designated worker changes the
regulatory speed limit to 45 mph. This system resulted in a significant reduction in
work zone speeds compared to a conventional static work zone speed limit Sign.
However, the magnitude of the speed reductions was not reported.

> Virginia: Park and Yadlapati (2003) proposed a VSL control logic that accounts for
both safety and mobility measures in work zones and tested it through a microscopic
simulation program (VISSIM). In this logic, a method of finding the optimum speed
with respect to a minimum safe distance equation (MSDE, the safety measure) and a
travel time (the mobility measure) was described. This study evaluated this control
logic under varying speed limit compliance rates (i.e., 70, 80, and 100 percent) and
two demand conditions (i.e., undersaturated and oversaturated conditions). For the
undersaturated condition, the proposed control logic only performed better than a
VSL system that simply posts the average speed of vehicles in work zones when the
compliance rate was 100 percent. The results were mixed for 70 and 80 percent
compliance rates. For the oversaturated conditions, the proposed control logic
produced lower travel times, but much smaller MSDE values.

2.3.5 Combination of 2 factors or more

> Nevada: Robinson reported the use of VSL on I-80 next to a coal-fired power
generation plant in a canyon with a river where fog is likely to occur. The system
consists of visibility detectors, speed loops, 3 Road Weather Information System

(RWIS) weather station, and advance warning signs. The regulatory speed limits are

26

set in increments of 10 mph based on the 85tln percentile speed, visibility, and
pavement condition. The signs display a speed limit and a warning message in
adverse weather conditions and remain blank on clear days. Robinson indicated that
the reliability of the visibility sensor limited the operation of this system.

New Jersey: Wilkie (1997) documented the use of a VSL on the New Jersey
Turnpike. In all, 120 VSL signs were installed over 148 miles to provide early
warning of slow traffic or hazardous road conditions to motorists. The system also
consists of loop detectors and weather sensing equipment, and advance warning
signs. Swindler et al. (2001) reported that speed limits are set based on the average
travel speed from the TRANSMIT system, which determines the speed by using the
EZ Pass electronic toll equipment with 2-mile detector spacing. The regulatory speed
limits are displayed automatically and can be overridden for lane closures and
construction zones. Speed limits can be reduced from the normal speed limit in five-
mph decrements, to 30 mph. The speed limit reductions are based on any of six
reasons: crashes, congestion, construction, ice, snow, and fog. Wilkie reported that
the State police officers feel that the signs are effective only when the reduced speed
ahead message is also illuminated.

New Mexico: The VSL system was installed and operated on eastbound I-40 in
Albuquerque from 1989 through 1997 to provide a US test bed for VSL equipment
and algorithms. Due to road widening, the system was dismantled in 1997. Harwick
(1989) reported that the system consisted of three roadside stations, which contained
a control cabinet, a VSL sign, and a hazard warning message board. Harwick

indicated that the objective of this system was to reduce accident frequency by

27

advising motorists of hazards through posted speeds that reflect driver behavior. The
regulatory speed limit was equal to the smoothed average speed plus an
environmental constant rounded to the nearest 5 mph increment (Wilkie, 1997). The
environmental constants are shown in table 2.8. The posted speeds ranged from 30 to
50 mph and were updated every minute. Wilkie reported that this system resulted in a
slight reduction in accidents and that the ability of the system to post speed limits that
reflected traffic conditions was hindered by high average speeds, which exceeded the
National Maximum Speed Limit of 55 mph.

Table 2.8. Environmental Constants for New Mexico (W ilkie, 1997)

 

 

Environmental condition Constant (mph)
Light +7.5
Dark +5.0
Light and Precipitating +2.5
Dark and Precipitating +0.0

 

 

 

 

France (Marseille): Swindler et al. documented the use of a VSL system on 5 miles of
the southbound lane of a major highway in Marseille. The posted speed limits are
based on prevailing speed, weather conditions, and presence of trucks. Speed and
volume data are collected by overhead radar and television surveillance. Speed limits
are set by measuring the mean speed of each 20-vehicle group and rounding up to a
multiple of 10 km/h up to 110 km/h. The speed limits of 80-90 mm, which give the
maximum served volume, were displayed when the volume reaches a specific level.
The system also calculates the mean speed difference between two detection stations

located 0.13 mile apart to warn the drivers of slow vehicles ahead. A 50 percent

28

reduction of crash rates from 1976 to 1992 was achieved by cutting the mean speed
by as little as 3.8 km/hr. Moreover, this system resulted in a decrease in speed limit
violations.

> Germany: Zackor (1979) reported on the use of VSL systems in Germany. VSL
systems were installed on A8 between Salzburg and Munich, A3 between Sieberg and
Cologne, and A5 near Karlsruhe. The VSL Sign spacing is 0.9 to 1.2 miles. The
objectives of this system were to stabilize traffic flow, reduce crash probability,
improve driver’s comfort, and reduce environmental impacts. The regulatory speed
limits are based on traffic flows, and environmental conditions, as shown in table 2.9.
Zackor indicated that the speed dispersion and the frequency of critically short
headways were reduced with the Operation of the VSL. Motorists responded better to
electronic signage than the fixed signage, because the electronic signs provided the
advisory information. Warren (2000) indicated that the use of variable speed limits
reduced the crash rate by 20 to 30 percent.

Table 2.9. Speed limit (km/h) setting algorithm for Germany (Warren, 2000)

 

 

Traffic Day Dark Day Dark Extreme
Flows Dry Dry Wet Wet
Low 130* 120 110 110 90..60
Mod 110 110 100 100 90.60
High 90 90 80 80 90.60
Unstable 70 70 70 70 60

 

 

 

 

 

 

 

 

* Unrestricted at very light traffic flow.

> Netherlands (between Amsterdam and Utrecht): Wilkie (1997) documented the use of
a VSL system on the A2 highway between Amsterdam and Utrecht. The objective of
the system was to create uniformity of speeds and volumes within and between lanes,

thereby reducing the risk of shock waves, crashes, and congestion. The system

29

consists of VSL signs spaced approximately 0.6 miles apart over 12.4 miles of road,
loop detectors every 0.3 miles, and automatic incident detection. The standard speed
limit is 120 km/h, while the variable posted speed is 50, 70, or 90 km/h depending on
l-minute average speed and volume across all lanes. . The regulatory speed limits are
posted in a red circle and enforced by photo radar. The advisory speed limits are
posted without a red circle. If an incident is detected, a speed limit of 50 km/h is
displayed. In general, the drivers complied with the speed signs. The severity of
shockwaves and speeds in all lanes were reduced, while the average occupancy
increased during the speed control. The average headway difference among lanes
became smaller and the variation in headways decreased.

> Sweden: A VSL system was installed on 20km of rural highway E22 in southern
Sweden. Swindler et al. reported that “the system aims to combine real time traffic
data and data from the road weather information system (RWIS) with a local climate
model predicting the road conditions on that special road section.” The speed limits
are displayed with different symbols to explain the reason for the speed reduction.
Swindler reported that this system resulted in a 10 percent reduction in speed and a 20
percent reduction in speed variance. However, he did not indicate the reduction in the
speed limit required to achieve this result.

2.3.6 Conclusions

The literature indicates that the use of VSL signs under specific weather related
conditions (fog, icy or snow covered roads) results in a reduction in the average speed.

However, the speed reduction is significantly less than the posted speed reduction. Most

30

systems use an increment of 20 km/h to warn of various weather conditions, and achieve
Speed reductions of less than 10 km/h.

The use Of VSL signs to advise truck drivers of the safe speed on grades showed
similar results. Truck speeds decreased with the use of the signs, but the average speed
exceeded the posted speed across various weight classes by an average of approximately
14 mph.

The literature also leads to the conclusion that the use of VSL signs without
notifying the driver of the reason for the change in the speed limit was not effective, but
this research was conducted nearly 40 years ago. More recently, a VSL system combined
with strict enforcement with photo radar has been demonstrated to be effective in
reducing speeds to the enforced speed (the posted speed plus the tolerance used for
enforcement).

The literature does not contain any guidance on the time and/or distance
frequency with which the posted speed limit can change and still be effective, nor on the
lag time between a change in the posted speed and the speed adjustment by the motorists.
All of the experiments have been with systems that do not change rapidly (weather,
grade, congestion) except the tests in Detroit (which found no response) and Albuquerque
(where the speed limit could not be displayed because it exceeded the NMSL).

The reports on the system used in London do not discuss the frequency with
which the VSL Sign changed, nor if the motorist speeds responded to each change, nor
the delay time in the response. These are all critical issues that need to be understood if

the system is to be effective in a work zone.

31

The minimum and maximum distance between VSL signs in also not known. The
system in London used an average spacing of 0.6 miles (1 km), but this distance may not
be appropriate in a work zone, where conditions could change over shorter distances.

Another unknown is whether motorists will increase speed above the speed they
would choose to drive with static speed limit signs. The ADVISE system in Utah
reported an increase in speed when the VSL signs are lighted under foggy conditions, but
this may be the result of motorists responding to the visibility of the signs rather than the
posted speed.

In this study, the system will be dynamic and will display both speed limit
increases and speed limit decreases over relatively short distances. The response will be
measured in short time increments to determine the ability of the system to encourage

drivers to alter their speed frequently in response to the posted speed limit.

32

Chapter 3

DESIGN OF THE STUDY

This study was designed to evaluate the effectiveness of the VSL system in work
zones by comparing field data collected “before,” “during,” and “after” VSL operation.
The VSL system details, data collection methods, study plan, and speed-limit-setting
algorithms are presented in this chapter.

3.1 VSL system details

The VSL system components, system configuration, system operation, and
operational parameters will be discussed in this section.
3.1.1 System components

The VSL system called “the IRD Speed Ranger Safety System,” which was
manufactured by International Road Dynamics Incorporated (IRD), was used in this
study. The VSL system as implemented in this study consists of one master trailer (trailer
1, shown in figure 3.1) and six slave trailers (trailers 2-7). The slave trailers contain
controller electronics, a short-range wireless modem (SRWM), remote traffic microwave
sensors (RTMS), simple system diagnostics, a user interface, a variable message Sign
(VMS), solar panels, and a battery pack. The master trailer contains everything that the
slave trailers contain plus a long-range wireless modem (LRWM) and a highway surface

condition monitor (HWCM). The number Of slave trailers used at each site varies based

33

on the length and geometry of the work zone section. The VSL trailers were classified
into three types: downstream, middle, and upstream. The master trailer is used as the

downstream trailer, while the slave trailers are used as the middle and upstream trailers.

   
 

Solar panels

Long range modem antenna

  
 
 
 
 
 
 
 
  

RT S
Highway surface condition monitor
Short range modem antenna

Variable message sign

 

Controller electronics
User interface

Short range modem
Long range modem
System diagnostics

 

 

 

Battery pack

Figure 3.1. VSL master trailer

3.1.2 System configuration

The VSL trailers need to collect traffic data by using the RTMS and communicate
those data with each other in order to set speed limits. The proper system configuration
will ensure that the VSL trailers communicate with each other effectively and the RTMS
collect traffic data accurately. To set the proper system configuration, the following

factors need to be considered:

34

> Line of sight: A clear line of sight between short-range modern antennas on adjacent
trailers is required for the system to communicate. If the adjacent antenna cannot be
seen, either the trailer must be moved or the antenna must be raised. Each trailer is
programmed with the short-range modem IDS of the adjacent trailers only, thus the
communication must pass through adjacent trailers and trailers cannot be skipped.

> Site layout: Because of the line of sight requirement, the horizontal and vertical
alignment of each site has an effect on the spacing between each trailer. The spacing
between each trailer has to be shorter in an area with a lot of curves, hills, and other
obstructions that can block the line of sight. In a flat area, the VSL trailers can
communicate with one another at a distance up to 1.5 miles.

> Distance from the road: The distance of the trailers from the road is governed by the
RTMS operation. It is recommended that the RTMS be placed at least 5 feet (1.52
meters) above the shoulder and no more than 10 feet (3.05 meters) from the edge of a

driving lane. Figure 3.2 shows the RTMS placement.

DETECTION ZONE

  

 

 

5 (1.52) - 17 (5.18) ....

 

8 (2.44) """"'

 

 

 

 

 

 

IE].

I I
[SHOULDER LANE 1 ' LANE 2
10 (3.05) 120.66) 12 (3.66)

 

Figure 3.2. RTMS placement (IRD, 2002)

35

3.1.3 System operation

The speed limit on each VSL trailer is either set at a constant value or is based on
the average speed plus five mph, which was found to be a good approximation of the 85th
percentile speed (see section 4.1.4 of this report). The speed limit can be based on the
speed measured at that trailer or the speed measured at the next downstream trailer.
Periodically, the master trailer will transmit a speed limit to the closest upstream trailer
based on traffic data from the RTMS for one or more of the trailers and a surface
condition report from the HWCM. In turn, each of the middle trailers will transmit a
speed limit to the closest trailer upstream as they receive an updated speed limit from the
closest downstream trailer. When the most upstream trailer receives an updated speed
limit, it will start a reverse transmission process by transmitting its traffic data to the
closest trailer downstream. Each of the middle trailers will collect the data transmitted
from the closest upstream trailer, append its traffic data and transmit all the data to the
closest downstream trailer. When the data reaches the master trailer, it appends its own
traffic data, uses all the data to set the speed limit at each trailer for the next period, and
stores the data so that they can be transferred off site through the long-range wireless
modem. When the speed limits are updated, the updated speed limits on all of the trailers
can be automatically sent to police officers through pagers for enforcement purposes
(IRD, 2002).
3.1.4 Operational parameters

In order to operate the VSL system effectively, the following parameters need to

be determined and set. These parameters need to be set on each trailer individually.

36

‘> Maximum positive differential: A maximum positive differential is the maximum
positive speed limit difference allowed between the speed limit on that trailer and the
speed limit on the closest downstream trailer. Since this value has no effect on traffic
disruption, it can be set as high as possible. A maximum positive differential of 30
mph, which is the maximum value allowed by the system, was used in this study.

> Maximum negative differential: A maximum negative differential is the maximum
negative speed limit difference allowed between the speed limit on that trailer and the
speed limit on the closest downstream trailer. The maximum negative differential of
10 mph was used in this study for a smooth transition in speed zones. A 10 mph
negative differential is the maximum value specified in the Michigan State Police
speed zoning guidelines.

> Maximum speed change: A maximum speed change is the maximum speed limit
difference between two subsequent time periods. Since this value has no effect on
traffic disruption, it can be set as high as possible. A maximum speed change of 25
mph, which is the maximum value allowed by the system, was used in this study.

> Trailer update period: The trailer update period indicates how often speed limits are
updated. The system allows update periods of 5 to 15 minutes. The trailer update
period was set to five minutes in this study, but in reality the speed limits were
updated about every six minutes because it takes about one minute for the system to
complete a communication cycle.

> Road Factor: A road factor defines a percentage of the calculated speed limits to
display, based on the four following road conditions: dry, wet, snow, and ice. Because

this study was conducted in the summer, snow and ice were not present. The dry and

37

wet road factors were set to 100 percent. In essence, this factor was not utilized in this
study.
3.2 Data collection methods

There were three methods of traffic data collection proposed for use in this study:
an Autoscope system, RTMS, and pneumatic tubes. However, the results from the initial
test indicated that the Autoscope system did not provide accurate speed and volume
measurements, particularly when shadows were present. This resulted in the decision not
to use the Autoscope system for the data collection in this study.

3.2.1 Remote traffic microwave sensor (RTMS)

RTMS provides presence detection of vehicles in multiple zones. Its ranging
capability is achieved by frequency modulated continuous wave (FMCW) operation. The
sensor transmits a microwave beam and receives energy reflected by both moving and
stationary objects in its path. At any given time there is a difference between the
frequencies of transmitted and received target signals, this difference in frequencies is
proportional to the distance between RTMS and the target. The RTMS detects and
measures this difference and computes the distance to the target. The RTMS detects
presence of objects in seven feet wide radial range slices in the path of the microwave

beam as shown in figure 3.3.

In... --'
a . ' ”we

\ 2' .5
s I t'

\. ‘0' f
\' ., .f..-
\I ‘.

MI" “so! ecu.
“—

e;Microwave beam

2m range slice
3.. 3! '3

I r' .‘ .'
\ .- .... 0‘ t. D,
D O I O

 

_ J .:.i ~_ <———- Beam footprint

 

Figure 3.3. RTMS microwave beam and its footprint (E18, 2000)

38

In this study, the RTMS was mounted in a forward-looking configuration with
detection zones aligned along the direction of travel. Each detection zone consists of
three consecutive range slices essentially forming a “speed trap”. An average speed is
determined from the transit time through the detection zones. The system can be used to
detect vehicles approaching or after they have passed the transmitter. The RTMS was set
up to detect vehicles after they have passed the transmitter in this study (Electronic
Integrated Systems Inc. (EIS), 2000).

The RTMS was used to collect traffic data used for evaluation during the VSL
operation and the data from RTMS was also used for setting the speed limit. The RTMS
recorded the following traffic data at each VSL trailer: a 6-minute average speed, a 6-
minute traffic volume, and a 6-minute lane occupancy. The average speed and volume
were verified during the pre-deployment by a comparison with data from a radar gun and
a manual traffic count, respectively. The average speed and volume were also verified
during the VSL deployment by a comparison with data from pneumatic tube detectors.
3.2.2 Pneumatic tube system

The pneumatic tube based system consists of two rubber tubes that are placed
about six feet apart across one or more lanes of traffic. One end of the tubes is sealed and
the other is attached to a mechanism (detector) that is actuated by air pressure. An
average speed is determined from the transit time between the two pneumatic tubes. The
pneumatic tube detector can also classify vehicle types into the 13 FHWA vehicle-
classification categories by determining the number of axles and their spacing. A

recording mechanism includes a clock and a device for recording the time, the count, and

39

the average speed over a specified time (15 minutes was used in this study). (French and
Solomon, 1986)

Pneumatic tube detectors installed by MDOT were used to collect the traffic data
“before,” “during,” and “after” the VSL operation. During the VSL deployment, they
were installed at the same locations as the VSL trailers as shown in figure 3.4. The
pneumatic tube data provided speed distributions for each vehicle type in lS-minute
increments. The individual vehicle speeds were recorded in either a 16 or 30 speed bin
format. The lS-minute average speed, 85'h percentile speed, and speed variance were
estimated by assuming that all of vehicles in each bin traveled at the middle speed of that
bin. Thus, the precision of these values depend on the number of speed bins. The average
speed and volume were compared with the RTMS data for verification. The advantage of
the pneumatic tube detector is that the output can be used to estimate the 85th percentile
speed and speed variance, while the output from the RTMS cannot. The disadvantages of
the pneumatic tube detector are that it can not cover more than two lanes of traffic, and

tubes are prone to ripping loose or tearing when a vehicle slides its wheels when crossing.

 

Figure 3.4. Pneumatic tube system

40

3.3 Study plan

This study was divided into 2 parts: the pre-deployment test and the VSL
deployments in the work zones. The purpose of the pre-deployment test was to ensure
that the VSL system worked properly, while the purpose of the VSL deployments in the
work zones was to evaluate the impacts of the VSL system under various work zone
conditions.

3.3.1 Pre-deployment test

The pre-deployment test was conducted to ensure that the VSL system worked
properly before the field deployment in the work zones. The functional characteristics of
the system were tested by verifying that the VSL trailers were able to communicate with
each other, with the pagers, and with a remote computer. The abilities of the system to
collect traffic data and road surface conditions and to use those data (along with a
specific algorithm) to set speed limits were evaluated. The volume collected by RTMS
was compared with the volume from a manual count, while the speed collected with
RTMS was compared with the speed data collected by a radar gun.

The VSL system of one master and three slave trailers was deployed on the
southbound direction of Okemos Road, south of [-96 from 04/11/02 to 05/11/02. Okemos
Road is a two-way, two-lane asphalt road with 4 to 6 foot shoulders and a posted speed
limit of 55 mph. Since the VSL system was being calibrated, the VMS was turned off
during the pre-deployment testing. Each VSL trailer was located about 0.5 miles apart as

shown in figure 3.5.

41

\'l

MAPeVisr‘E l

r.
N
I"
d
3%,

Sandnm _Rd

pa wept-69H
'pu sowaxo
l
l
l
9091039 I.

WhithyFk‘l ,-

 

 

Hc It Rd Holt Rd

 

.’ W WWH
l
pg bolting
DU suetuuns

 

,Lamms ,, . Wingfiug E] VSL trailer

ozooe moon-steam. Inc; ”hm... T;hn61mb6
Figure 3.5. Location of VSL trailers during the pre-deployment testing

3.3.2 VSL deployment in work zones

In the work zone deployments, the VSL system first operated with static black on
white speed limit signs (50 mph) covering the VMS on each trailer while the speed-limit—
setting algorithm was checked and the “before” data were collected. Thus, “before” data
was collected when the system was present but the VMS was covered with a static fixed
speed limit sign. When the speed-limit-setting algorithm was operational and enough
“before” data had been collected, the static speed limit signs were removed from the
VMS on each trailer. The VSL system was then put in full operation (the VMS was

uncovered) and the “during” data were collected for about one week in each deployment.

42

During the time period that the “before” and “during” data were collected by the VSL
system, the pneumatic tube detectors were also used to collect data for comparison
purposes.

After the VSL system was removed fi'om the work zones and the regular static
speed limit signs were replaced, the pneumatic tube detectors collected the “afier” data
for about one week. This study covered four deployments in four different work zone
configurations. These four work zones were good candidates for the VSL deployments
because there were variations in day-to-day and trailer-to-trailer average speeds in these
work zones. The VSL can reflect these variations in the displayed speeds, while the static
speed limit cannot. Example of the variations in day-to-day and trailer-to-trailer average
speeds is shown in figure 3.6. All seven VSL trailers were used in the first three

deployments, while only six trailers were used in the fourth deployment.

 

65

average speed (mph)
Ur
o

 

0.0 0.5 1.0 1.5 2.0 2.5 3.0

T7 T6 T5 _ T4 _ T3 T2 T1
distance relative to trarler 7 (miles)

i—-—Wed., 05/22/02 -°:Thu, 05/23/02 —~—Fri., 05/31/02 13‘ deployment
j+Sat., 06/01/02 +Sun., 06/02/02 —~—Mon., 06/03/02 VSL display: off
U—Tue, 06/04/02

 

 

 

 

 

 

 

L ,7 1,7, 7 i 7

Figure 3.6. Example of the variations in day-to-day and trailer-to-trailer average speeds
from 5PM to 6PM “before” the first deployment

43

> First deployment: The VSL system was installed on a 2.7-mile section of eastbound
I-96 between the overpass ramp from I-69 to 0.7 miles east of Creyts Road as shown
in figure 3.7. There was a total reconstruction of the eastbound lanes of 1-96. The
eastbound traffic was shifted to one of the normal westbound lanes through a traffic
crossover. Two-way traffic (one lane in each direction) was maintained on the
westbound lanes with a temporary concrete barrier wall separating the two lanes. The
eastbound traffic was shifted back to the 2 eastbound lanes through the traffic

crossover, which was located just west of trailer 1.

OS

Mum -...1,-= i
1-

“ISM-III!) ‘1

  

 

 

 

 
  

 

 

n:
31 4‘3. 31 :-
Egan. ‘-_
~" 9.; 1‘5; Bishopry E
l . g E

I .

. V, ‘ ‘ I Dr .
. ._/‘ . r o I VSL trarler
02002 lineman Inc; 02002 Nav'natbn Toehnobobs “hi."

Figure 3.7. Map of the first deployment work zone

 

> Second deployment: The VSL system was installed on a 2.6-mile section of
eastbound I-96 from 0.7 miles west of Wacousta Road to the on-ramp from exit 90 as
shown in figure 3.8. There was concrete patching on the left lane, and only the right
lane was open. Barrels were used to separate the two lanes. The construction

activities were conducted between 6AM and 8PM.

MUST"?

 

 

Econ--
noon
:2 a:
.. -.-
z I
3 I;
. § 2,
31 .8
. . WSIOIW. ..
W
Grand my” H
H

  
 
   
    
    

Rolling] HIII Ln

Wacousta Rd

Bauer Rd

0‘” Tran ipOrtatibn

  
       
   

W Eaton Clinton

7 Eato n

Figure 3.8. Map of the second deployment work zone

“noun-4i T»
02002 floooodoom. "10462012 Wu Toohmbobe

> Third deployment: The VSL system was installed on a 2.2-mile section of
eastbound I-96 between the on-rarnp from I-69 and 0.7 miles east of Creyts Road as
shown in figure 3.9. There was a total reconstruction of the westbound lanes of 1-96.

The westbound traffic was shified to one of the normal eastbound lanes through a

45

traffic crossover, which was located just west of trailer 1. Two-way traffic (one lane
in each direction) was maintained on the eastbound lanes through the deployment

with a concrete barrier wall separating the two lanes.

      
    

infuwiv'.‘ » 7

L. L!!! "9'1. . _

pa 1" to s

2.1-1115011-0‘ ark

 

‘09 "urns

-
92002 undocummrnc.;o2ooa Maren room-oboe- ' , . If ‘
Figure 3.9. Map of the third deployment work zone

> Fourth deployment: The VSL system was installed on a 2.6—mile section of
westbound 1-96 between the on-rarnp from exit 93a and Grand River Avenue as
shown in figure 3.10. There was concrete patching on the right and middle lanes.
Only the left lane was open. Barrels were used to separate the lanes. The construction
activities were conducted between 6AM and 10PM. The construction zone ended and

the westbound traffic became two lanes right before trailer 6.

46

 

 

 

 

 

MHan
‘7 u.
(I
i
E
m“
t/W no.
(3"
. . Clinton
E‘ Eaton
‘6
.2 .
II -.__.-,... . . ~
2 g .3 . i I L , “Nil 29
. .. a x . .
3 -.
I? \ g OUR» - -.::»_y;til.j,,ls~ ,
1121-. ,7
. ,8 n » .. ..: . ~- :9
l . 9 1 3.. _J \{QI‘
Willow Hwy WW J .
_—' wave a I . .'
. ; Yaw, . r: e r
I“; -. 37‘5““, F’Jgf g. :1
~ 2K." > J 6 ‘
g 5 a“ . ik‘x L..—\.z -4}
6 A 9:» Q 3,. g ,_
' ~‘7 g 3 E—Lfibé. *3." r E >‘r OI
. , ‘P r ,7 .
. I .“- m ; 3.361; I VSLtrailer
I ‘2; an. Patrolcar
F Khulna“ HW V 3 W ~
02002 Woormtdoomlne 92002 Wu Toehnobabs :1 1

Figure 3.10. Map of the fourth deployment work zone

The effect of police enforcement was also tested in this deployment.
According to the literature review, the presence of a stationary patrol car is one of the
most effective methods of speed limit enforcement. Thus, a stationary patrol car was
programmed to be present at three locations in the work zone during the different
time periods: 1) Grand River Avenue; 2) I-69 overpass; and 3) Eaton Highway (see
figure 3.10). The police ofiicers were notified of the updated speed limit through a
pager. Traffic citations were issued only in extreme cases. The stationary patrol cars
were scheduled to be present one day “during” the VSL operation and one day “after”

the VSL operation for comparison purposes.

47

3.4 Speed-limit-setting algorithm

International Road Dynamics Inc. (IRD) designed the VSL system to vary the
speed limit based on observed speeds, occupancy, and weather conditions. When the
occupancy reaches user-specified low or high occupancy thresholds, the speed limit is set
to the maximum and minimum (pre-set) speeds, respectively. Since this study focused on
setting speed limits based mainly on the 85th percentile speed, not the occupancy, the low
and high occupancy thresholds were set at values of 0 and 90 percent, respectively, which
effectively “turned off” occupancy as a variable. Thus, when the occupancy was in the
range of 0 to 90 percent, the speed limit was set based on the 85th percentile speed, which
was estimated to be the average speed (v) plus 5 mph. The 85th percentile speed had to be
estimated because the RTMS does not collect individual vehicle or “binned” speeds,
which would be required to calculate the 85th percentile speed. This 85th percentile speed
estimation was verified as discussed in the next chapter. Overall, the minimum and
maximum speed limits used by the VSL system in this study were 40 mph and 70 mph,
respectively although the maximum speed allowed at any trailer varied by deployment.
Other factors used to determine the speed limit included geometric constraints and
construction activity.

There were five different profiles used to set the speed limit as shown in table 3.1.
Each speed profile indicates the maximum speed limit, the minimum speed limit, and the
recommended speed limits for each observed average speed range. All of the speed
profiles were restricted to using the same eight speed range format. The size of each
range is 5 mph, except the opening range for the maximum and minimum speed. The

speed-limit—setting algorithm in each deployment will be explained next.

48

Table 3.1. Speed-limit-setting profiles

 

 

PROFILE l 2 3 4 5

L0. THRESHOLD - 0% 50* 6O 70 70 60

HO. THRESHOLD — 90% 40 40 40 60 40
M11) OCCUPANCY

v < 40 40 40 40 60 60

40 s v < 43 45 45 45 60 60

43 s v < 48 50 50 50 60 60

48 S v < 53 50 55 55 60 60

53 s v < 58 50 60 60 60 60

58 s v < 63 50 60 65 65 60

63 s v < 68 50 60 70 70 60

v 2 68 50 60 70 70 60

 

 

 

 

 

 

 

 

v: average speed
* The numbers shown In the table are displayed speed limits.

3.4.1 First deployment

Since there were no construction activities near the traveled lane, the speed limits
were mainly based on geometric constraints. The algorithm was set to increase the speed
limits at the locations where there were no geometric constraints and decrease the speed
limits at the locations where there were geometric constraints. The maximum speed limit
MDOT allowed in this construction zone was 60 mph because the lane was narrower than
normal.

The site layout and the speed-limit-setting algorithm for the first deployment are
shown in figure 3.11 and table 3.2, respectively. The starting point in the VSL-controlled
area was well after the start of the work zone, and the speed limit (using static signs) was
50 mph for traffic approaching trailer 7 (the first trailer encountered by the motorists

traversing the site). The maximum speed limit allowed on trailer 7 was 50 mph because

49

it was located in advance of an on-ramp from I-69. The maximum speed limit allowed on

trailer 2 was also 50 mph because it was located at a traffic crossover. The speed limits at

trailers 2 and 7 were based on the speeds recorded at these respective trailers.

 

 

 

 

 

 

 

 

 

 

Wall
-—9 —\
// LEI
-—->
1-69
Figure 3.11. VSL site layout for the first deployment
Table 3.2. Speed-limit-setting algorithm for the first deployment
. . VSL based on avg.
Trailer Displayed Speed speed at trailer

1 Based on profile 3. l

2 Based on profile 1. 2

3 Based on profile 2. 3

4 Based on profile 2. 3

5 Based on profile 2. 4

6 Based on profile 2. 5

7 Based on profile 1. 7

 

 

 

 

 

The speed limit on trailer 3 was also based on the speed at site 3 because the

speed limit might have been too low if it was based on the downstream speed at trailer 2.

The speed limits at trailers 4, 5, and 6 were set based on the speeds at the next

downstream trailer.

The work zone ended right before trailer 1, so the maximum speed

limit allowed at trailer 1 was 70 mph, which was the same as in normal non-construction

conditions.

50

3.4.2 Second deployment

Since the construction activities were only carried on between 6AM and 8PM

during the weekdays, this deployment used two different speed-limit-setting algorithms:

one during the daytime and one for nights and weekends.

Because the construction

workers worked on a concrete patching operation close to the traveled lane and only

barrels separated them from adjacent traffic, the maximum speed limit allowed in the

work zone during the daytime was 50 mph. The maximum speed limit allowed at night

and on the weekends was 70 mph, which was the same as normal condition because there

was no construction activity at these times. The site layout and the speed-limit-setting

algorithm for the second deployment are shown in figure 3.12 and table 3.3, respectively.

 

 

 

 

 

70 *
mph III E63“ El"
<——-
..: --'J , . ,‘ . ‘3...
.._ .J-.. lane closure J- ”fr-5F

 

*Signs 6 and 7 were located where the SO-mph and 60-mph static signs used to be, respectively.

Figure 3.12. VSL site layout for the second deployment

Table 3.3. Speed-limit-setting algorithm for the second deployment

 

 

 

 

 

 

Trailer 6AM - 8PM weekdays 8PM gaggkzfikdays Vigfigsfftfglgjg
1 Based on profile 1. Based on profile 3. 1
2 Based on profile 1. Based on profile 3. l
3 Based on profile 1. Based on profile 3. 2
4 Based on profile 1. Based on profile 3. 3
5 Based on profile 1. Based on profile 3. 4
6 50 mph Based on profile 3. 5
7 60 mph Based on profile 4. 6

 

51

 

Since the speed limit prior to the lane closure was 70 mph and the speed
difference between any Sign and the previous upstream Sign could never be less than —10
mph, the speed limits on trailers 6 and 7 were fixed at 50 and 60 mph, respectively during
the day and the speed limit on trailer 7 ranged between 60 and 70 mph during the night
and on weekends. The speed limits at trailers 2 to 7 were set based on the speeds at the
next downstream trailer. At the end of the construction zone, the existing static speed
limit sign was used to post the normal speed limit of 70 mph.

3.4.3 Third deployment

Since there were no construction activities near the traveled lane, the speed limits
were mainly based on geometric constraints. The site layout and the speed-limit-setting
algorithm for the third deployment are shown in figure 3.13 and table 3.4, respectively.
The starting point of the VSL-controlled area was already in the work zone, and the speed

limit was set at 50 mph for vehicles approaching the VSL system.

 

 

 

 

 

 

SB
131801ng 111mm

 

 

 

 

Figure 3.13. VSL site layout for the third deployment
Since there were many on and off ramps in the vicinity of trailers 4 to 7, the speed
limits were governed by the location of the ramps. The speed limits on trailers 5 and 7
were programmed to be fixed at 60 mph unless the occupancy was higher than 90
percent. Due to a communication problem with a short-range modem cable, the speed

limit on trailer 7 was fixed at 60 mph, regardless of the occupancy. The speed limit on

52

trailer 5 was also fixed at 60 mph because the speed limit could not be set based on the
speeds at that trailer or the downstream trailer, which was located in a merging area.

Table 3.4. Speed-limit-setting algorithm for the third deployment

 

 

Trailer Displayed Speed Vitilgzszetd $3133
1 70 mph N/A
2 Based on profile 3. l
3 Based on profile 3. 2
4 Based on profile 1. 4
5 Based on profile 5. 5
6 Based on profile 1. 6
7 60 mph N/A

 

 

 

 

 

The maximum speed limit allowed at trailers 4 and 6 was 50 mph and the speed
limit was set based on the speed recorded at these respective trailers because there were
ramps immediately downstream from each trailer. If the speed limit on trailers 4 and 6
was set based on their downstream trailers, the speed limits could have been too high.
The maximum speed limit allowed at trailers 2 and 3 was 70 mph because there were no
ramps in the vicinity of these two trailers and the eastbound traffic did not need to cross
over at the end of the work zone. The speed limits at trailers 2 and 3 were set based on
the speeds at the downstream trailers. The work zone ended right before trailer 1, so the
speed limit on trailer 1 was set at 70 mph, which was the same as in the normal condition.
3.4.4 Fourth deployment

Since the construction activities were only carried on between 6AM and 10PM,
this deployment used two different speed-limit-setting algorithms, one for day and one
for night. Because the construction workers worked on a concrete patching operation

close to the traveled lane and only barrels separated them from the traffic, the maximum

53

speed limit allowed in the work zone during the daytime was 50 mph. The maximum
speed limit allowed at night was 70 mph, which was the same as normal conditions
because there was no construction activity at night. The site layout and the speed-limit-

setting algorithm for the fourth deployment are shown in figure 3.14 and table 3.5,

    

respectively.
0......
from exit 93

      

 

El m

Figure 3.14. VSL site layout for the fourth deployment

Table 3.5. Speed-limit-setting algorithm for the fourth deployment

 

 

 

_ VSL based on avg.
Trailer 6AM - 10PM 10PM — 6AM speed at trailer
1 Based on profile 1. Based on profile 3. 1
2 Based on profile 1. Based on profile 3. 1
3 Based on profile 1. Based on profile 3. 2
4 Based on profile 1. Based on profile 3. 3
5 Based on profile 1. Based on profile 3. 4
6 70 mph (fixed) 70 mph (fixed) N/A

 

 

 

 

 

The starting point of the VSL controlled area was already in the work zone with a
speed limit of 50 mph prior to that point. The speed limits at trailers 2 to 5 were set
based on the speeds at the downstream trailers, while the speed limit at trailer 1 was set
based on the speed at that site because trailer 1 was not set up to communicate with the
downstream trailer (trailer 6). At the end of construction zone, VSL trailer 6 was used to
display the normal fixed speed limit of 70 mph. Since trailer 6 was the only trailer that

used a different speed-limit-setting profile than others, trailer 6 was set up to operate

54

independently. This setting allowed the speed—limit-setting profile at trailers 1 to 5 to be

changed remotely though the long-range modem.

55

Chapter 4

RESULTS OF DATA VERIFICATION AND PROBLEMS AND
LIMITATIONS FACED IN THE VSL DEPLOYMENT

4.1 Results of data verification

The speed and volume measured by the RTMS system and the pneumatic tube
system were verified to ensure that the data from these two systems are accurate enough
for the VSL evaluation. The use of the average speed plus 5 mph as a surrogate measure
of the 85th percentile speed is also verified in this section.

4.1.1 Remote Traffic Microwave Sensor (RTMS)

The average speed and volume measured by the RTMS system in the pre-
deployment test on Okemos Road were compared with an average speed measured with a
radar gun and a count of traffic. To be consistent with the RTMS data, an aggregation
period of 6 minutes was used to determine the average speed and volume.
> The first RTMS data verification was performed at trailer 3 during the pre-

deployment. The RTMS was set up to detect vehicles approaching the transmitter.
The results of volume and speed comparisons are shown in figures 4.1 and 4.2,
respectively. The volume comparisons showed a significant undercount by the RTMS
system. The speeds measured by the RTMS were scattered about the radar speed

measurements, averaging about 2 mph lower than the speed measured by the radar

56

gun. Because of the significant undercount of traffic volume, some of the factors that

affected the RTMS accuracy were adjusted in the next verification test.

l

I

 

50

40
30 ~

20

 

 

 

Volume (vehicles/6 minutes)

 

0 . i l . Tit : . - -. T i I Whirl . l _.
308 320 332 344 356 408 420 432 444 456
L_._§La'3_Tf_m_(P_M)_ L

l—o—Count—r—RTMS} if

-__ ___..___~-_____L-__.___. _‘-" ‘—" _._ ‘W‘m‘T *' _______._J

Figure 4.1. Volume comparisons at trailer 3. for the first RTMS data verification

 

1
l
l
i

Speed (mph)

 

 

 

 

308 320 3:32 3:44 356 498 4:20 432 4:44 456
___ StartTime(PM) __
g—o—RadarGun—I—RTMSII

 

 

 

 

 

Figure 4.2. Speed comparisons at trailer 3 for the first RTMS data verification
> The second RTMS data verification was also performed at trailer 3 during the pre-
deployrnent. The RTMS was still set up to detect vehicles approaching the
transmitter, but the position of the RTMS relative to the traffic lane, the angle at

which the sensor looks down onto the traffic, and the sensitivity of the system were

57

adjusted to improve the traffic count. The results of volume and speed comparisons
are shown in figures 4.3 and 4.4. The difference in the volume estimated by the
RTMS system and the counts was greater than in the first test, and the RTMS
reported a constant average speed of 44 mph for each interval. Since the accuracy of
the speed and volume measurements was even worse than the first verification, the
RTMS was rotated 180 degrees to sense vehicles after they had passed the trailer in

the next verification.

 

 

 

 

 

l

47,35
13
,E30
fg25
1,220

.2
1615*
:3,
ID
.>0

 

 

Figure 4.3. Volume comparisons at trailer 3 for the second RTMS data verification

 

i
56 l
i 54 ~ i
52 ~n~ .
48 L L ...LLL ._-LL. __L .. .. __-.LLL. __ _- L L. _L...L. L, .

l
l
l
l
l
l

 

 

Speed (mph)

i 44 .- . ~. -L_LL . . . J WJJ JJJ, P

42 L... ...L - ..LL.. . L L L LLL.-L.L...

 

 

 

l 40 , w . . ,

g 353 339 345 351 357 403 409 405 421 427
1 Start Time (PM)

i —+—Ra¢u(hm—a—irnusj

l

 

l
I

 

Figure 4.4. Speed comparisons at trailer 3 for the second RTMS data verification

58

> The third RTMS data verification was performed at trailers 1, 2, and 3. To improve
the accuracy of the speed and volume measurements, the RTMS software was
upgraded and the RTMS set up was changed to detect vehicles after they passed the
transmitter. The RTMS setup at all three trailers were the same. The results of volume

and speed comparisons are shown in figures 4.5, 4.6, 4.7, 4.8, 4.9, and 4.10.

1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

 

M I
O

Volume (vehicles/6 minutes)

8

1
1
30

20

 

 

 

 

 

1 9-21 9-36 951 1006 1021 10-36 1051 11:06
1 Start Time (AM) 1
1 1+ Court 4 RTMS1 1

 

 

 

 

 

 

 

 

;, 1
1 56 1
1 55 4 -—~ 1
1 -= 1
1 g- 53 ~ 1
. D. 51 1 .. . 41

1 (I) ‘ "

1 50 _ .--- 44 . 4 .4 1
1 49 » 4 44 ~4 4 44 4 44 1
‘ 48 y 1 1 , , 1
1 921 9-36 951 10.06 1021 10-36 1051 1106 1
1 StartTime(AM) 1
1 1+Radaer—r-RTMSJ1 1

 

Figure 4.6. Speed comparisons at trailer 1 for the third RTMS data verification

59

 

 

€920 '
18

123

1.5

1815* ‘
i~o

‘E

‘E10-*~'~ ' _ ““'*“ * ‘ “‘ i”"‘
1.:

l°>’

‘gsr 4 44—7 *5 *
1.

1>0

 

 

 

 

1 1131 1137 11:43 11:49 1155 12.01

 

A
NJ:-

 

A4}
0

Speed (mph) .

DJ
00
1
1
1
1
1
1
1
1

 

 

U)
ON

 

 

1131 1137 11:43 11:49 1155 12:01
1 StartTime(AM)
1+Radaer —I—RTMS1

1
1
Figure 4.8. Speed comparisons at trailer 2 for the third RTMS data verification
The results from the third RTMS data verification showed that the volume
detection accuracy improved considerably at each of the trailers tested. Although the
volume measured by the RTMS was different by as many as 10 vehicles/6 minutes
from the volume measured by the manual traffic count, the difference was clearly

smaller than before. There was no improvement in the speed detection from the first

verification. The speeds measured by the RTMS still averaged about 2 mph lower

60

than the speed measured by the radar gun, but the range was from +1 mph to —2 mph
at trailer 1; from +5 mph to —7 mph at trailer 2; and from +1 mph to ——2 mph at trailer
3. Based on these results, the RTMS was set up to detect vehicles after they have
passed the transmitter during the VSL deployment in the work zones.

Although the volume measured by the RTMS was still as high as 20 percent
different from the volume measured by the manual traffic count, the accuracy was
acceptable because the traffic volume was not used as an MOE in this study. The
traffic volume was only used to provide a general idea about the level of congestion.
The 2-mph accuracy of the speed measurement was also acceptable for the VSL
evaluation. Moreover, the VSL evaluation focused more on the relative value of
speed in “before” and “during” conditions than the absolute value of speed in each

condition.

1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
_1

A
o 1

 

NLRB.)
MOM

 

 

3 i7:
1
1
1
1
1
1
1
1

1
1
1
1
1
1
1
1
1

Volume (vehicles/6 minutes)
N
O

 

 

O

 

1 1 v 4

1 10:57 11:12 1127 11:42 1157 12:12 1227 12:42!
1 StartTime(AM) 1
1

 

my

1
1
1

S+Count

l_._.__.._.______ __..

 

Figure 4.9. Volume comparisons at trailer 3 for the third RTMS data verification

61

Speed (mph)
-5 h L)! U! U1 U1 U!
0‘ 00 O N J:- O‘ 00

 

A
4;

1057 11:12 1127 11:42 1157 12:12 1227 12:42 1
Start Time (AM) 1
.+ WIGWfRTMSJ
Figure 4.10. Speed comparisons at trailer 3 for the third RTMS data verification
4.1.2 Pneumatic tube system
The average speed and volume measured by the pneumatic tubes were compared
with the RTMS data for verification in each of the VSL deployments in the work zones.
Only the data in the “before” and “during” conditions were available for comparison, the
RTMS data in the “after” condition was not available because the VSL trailers were
removed form the work zones “after” the deployments. The RTMS system was set to
IBpOrt volume data in 6-minute increments, while the pneumatic tube system reported
Volume in 15-minute increments. Thus the volume in each 30-minute period was
Coaapated. At most of the trailers, the 30-minute volume from the RTMS data was
higher than the pneumatic tube data about half of the time and lower the other half of the
time because the reporting period from the RTMS data was not precisely 6 minutes.

Some aggregate times for five RTMS volume periods were a little bit less than 30

minute S, while others were a little bit more than 30 minutes.

62

C]

> First deployment

0 Trailer 1: The average speed from the RTMS data was about 4 mph lower than the
pneumatic tube data in the “before” condition, while the average speed from the
RTMS data was about 0.5 mph higher than the pneumatic tube data in the “during”
condition. The 30-minute volumes in the right lane from the RTMS data were
approximately 100 percent higher than the pneumatic tube data most of the time in
both the “before” and “during” conditions. The high RTMS volumes might occur
because the RTMS counted some vehicles in the left lane.

Cl Trailer 2: The average speed from the RTMS data was about 5 mph lower than the
pneumatic tube data in the “before” condition, while the average speed from the
RTMS data was about 1 mph higher than the pneumatic tube data in the “during”
condition. The 30-minute volumes from the RTMS data were about 5 percent higher
than the pneumatic tube data about half of the time and 5 percent lower the other half
of the time in both the “before” and “during” conditions.

Trailer 3: As was the case in trailers 1 and 2, the average speed from the RTMS data
was about 5 mph lower than the pneumatic tube data in the “before” condition. The
average speed from the RTMS data was about 0.5 mph lower than the pneumatic tube
data in the “during” condition. The 30-minute volumes from the RTMS data were
about 15 percent higher than the pneumatic tube data about half of the time and 15
percent lower the other half of the time in both the “before” and “during” conditions.
Trailer 4: Only the “during” data was available for comparison at his trailer. The

average speed from the RTMS data was about 2 mph lower than the pneumatic tube

63

data. The 30-minute volumes from the RTMS data were about 7 percent lower than
the pneumatic tube data.

:1 Trailer 5: Only the “during” data was available for comparison at this trailer. In the
PM, the average speed from the RTMS data was almost the same as the pneumatic
tube data, but in the AM the average speed from the RTMS data was about 20 mph
lower than the pneumatic tube data. There might be an error in the pneumatic tube
data collection because the average speed in the AM was much higher than in the PM,
while the RTMS speeds were nearly equal. The 30-minute volumes from the RTMS
data were about 10 percent higher than the pneumatic tube data about half of the time
and 10 percent lower the other half of the time.

D Trailer 6: The pneumatic tube data at this trailer were not available.

B Trailer 7: The average speed from the RTMS data was about 5 mph lower than the
pneumatic tube data in the “before” condition, while the average speed from the
RTMS data was about 4 mph lower than the pneumatic tube data in the “during”
condition. The 30-minute volumes from the RTMS data were about 10 percent higher
than the pneumatic tube data about half of the time and 10 percent lower the other

half of the time in both the “before” and “during” conditions.

0 The “before” versus “during” average speed difference (between the RTMS and
pneumatic tube) comparison is shown in figure 4.11. It can be noticed that the speed
measurement of either the RTMS system or the pneumatic tube system was
inconsistent because the average speed differences between the RTMS and pneumatic

t“be—s in the “before” condition were not equal to the “during” conditions.

64

O

tube speed (mph)
:‘o

 

RTMS speed - pneumatic

 

 

BREEZERQ

 

Figure 4.11. The difference between the RTMS and pneumatic tube average speed
“before” and “during” the first deployment

> Second deployment
There was a problem with the VSL system during the second deployment. Thus the

data were not collected. This problem will be explained in the next section.

> Third and fourth deployments
Only the “during” data were available for comparison in the fourth deployment. The
“before” versus “during” average speed diflerence (between the RTMS and
pneumatic tube) comparison from the third deployment is shown in figure 4.12. The

results from the third and fourth deployments, along with the first deployment are

included in tables 4.1 and 4.2.

65

tube speed (mph)

 

RTMS speed - pneumatic

1 2 3 4 5 6 7
Trailer

I before I during

Figure 4.12. The difference between the RTMS and pneumatic tube average speed
“before” and “during” the third deployment

Table 4.1. Pneumatic tube average speed verification summary

Results lst deployment 3rd deployment

TMS was 1
ms was 0.5 than
ms was same as
ms was 0.5 lower than tube

ms was 1 lower than tube
was 2 lower than
was 2 lower than pneumatic
3PM-8PM (RTMS was 4 mph higher)
was 4 lower than tube

were not

 

The direction in the parenthesis next to the trailer number indicates the hne used in the speed comparison.
The trailers with no parenthesis indicate that there was only one lane at those trailers.

66

Table 4.2. Pneumatic tube 30-minute volume verification summary

 

 

 

 

 

 

 

 

 

Results lst deployment 3rd deployment (16:12)].
before during before during during
RTMS was 100 % higher than pneumatic tube 1(right) 1(right)
RTMS was 30 % higher than pneumatic tube 5
RTMS was 10 % higher than pneumatic tube 7 7
RTMS was almost the same as pneumatic tube, but 2, 3, 7 2, 3, 5, 2, 3, 4, 1(right), l, 5
varied by time slice because of the interval time 7 6 2, 3, 4,
variation 6
RTMS was 7 % lower than pneumatic tube 4
RTMS was 10 % lower than pneumatic tube 3,
6 (left)
The data were not available 4, 5, 6 6 1(right) 5 2, 4

 

 

 

 

 

 

 

 

The numbers in the table indicates the trailer numbers.
The direction in the parenthesis next to the trailer number indicates the lane used in the volume comparison.

The trailers with no parenthesis indicate that there was only one lane at those trailers.
From tables 4.1 and 4.2, it is clear that the average speed differences between the
RTMS and pneumatic tubes in the “before” condition were not equal to the “during”
conditions, particularly in the first deployment. This indicated that the speed
measurement of either the RTMS system or the pneumatic tube system was inconsistent.
However, the 30-minute volume differences in both the “before” and “during” conditions
were nearly equal. The volumes measured by the pneumatic tube were less than 10
percent different than the volumes measured by the RTMS at most of the locations. The
accuracy of pneumatic tube volume measurement was good enough to provide a general
idea about the level of congestion.

To identify the data collection method that provided the inconsistent average
speeds, the average speeds during one hour on all of the available dates from both the
RTMS and the pneumatic tubes were plotted over time. The average speeds during off-
Peak hOurs from 10AM to 11 AM and from 10PM to 11PM were used because the

\“fluence of traffic on the average speeds was lower during the off-peak hour. The data

67

from trailer 1 in the first deployment was used for this comparison because the difference
between the average speed differences between the RTMS and pneumatic tubes was very
high at this trailer. “Before” the deployment, the RTMS average speeds were about 4
mph lower than the pneumatic tube average speeds, while “during” the deployment, the
RTMS average speeds were about 0.5 mph higher than the pneumatic tube average
speeds. The pneumatic tube versus RTMS average speed comparisons at trailer 1 are
shown in figures 4.13 and 4.14.

From figures 4.13 and 4.14, it can be seen that the average speeds measured by
the RTMS were consistent, while the average speeds measured by the pneumatic tubes
were inconsistent. The pneumatic tube average speeds “before” the experiment
(OS/23/02-05/25/02) were about 4 mph higher: than the pneumatic tube average speeds
“during” the experiment (06/06/02-06/09/02). Since the pneumatic tubes were frequently
ripped up, the errors in the pneumatic tube average speeds may come from the difference
in the distance between the two tubes each time the pneumatic tubes were replaced. This
difference in the distance between two tubes affected only the speed measurement. Thus
the pneumatic tubes still provided consistent traffic volume measurements. Because of
the inconsistent speed measurement, the pneumatic tube data were not used in the
average speed analysis. However, the inconsistent speed measurement does not effect the
speed variance because the speed variance only depended on the relative values of
speeds, not the absolute values. Therefore the pneumatic tube data were still used in the

speed variance analysis.

68

 

i 8

$83

average speed (mph)

VI
0

54

52

Thu Fri, Sat, Fri,
05/23 05f24 05/25 05/31

 

SaL
06/01

bRTMS + new £5.21

 

Sun, Mon, Tue, Wed, Thu, Fri, Sm, Sun Mon,
06/02 06/03 06/04 06l05 0606 06107 06/08 06109 06/ 10
date

 

 

Figure 4.13. Pneumatic tube versus RTMS average speed comparison at trailer 1
from 10AM to 11AM in the first deployment

 

$SS28

average speed (mph)

&

 

Thu, Fri.
05/23 05/24 05f25

Sat, Fri,

05/31

Sat,
06/01

Sun, Mon, Tue, Wed, Thu, Fri, Sat, Sun, Mon,
06102 06/03 06/04 06l05 06/06 06/07 06/08 06/09 06I10
date

1§MWL5£1

 

 

 

Figure 4.14. Pneumatic tube versus RTMS average speed comparison at trailer 1
from 10PM to 11PM in the first deployment

69

4.1.3 85'” percentile speed estimation

The use of the average speed plus 5 mph as a surrogate measure of the 85th
percentile speed was verified. The 85th percentile speed and average speed plus 5 mph
were plotted over time for comparison. Even though it was determined that the
pneumatic tube data did not provide an accurate estimate of the mean speed, it was used
in this verification because the purpose of this verification is to determine if the value of
the 85th percentile speed can be estimated from the average speed. The 85th percentile
speed and the average speed plus 5 mph were compared using all of the available
pnetunatic tube data. Examples of the 85th percentile speed and average speed plus 5

mph comparisons are shown in figures 4.15 and 4.16.

 

 

 

 

 

 

 

 

 

lst deployment
treatment. before

1f§5§iie§nii 16 speed 4.: averagEEfieellrsingfl

 

 

170

1

165

1B

1;

1855444 4 444444— 4444 44 4
18‘

150~4444444444 444
1

145-

1 888888888888
1 starttime

1

1

1

1

1

 

 

 

Figure 4.15. 85th percentile speed and average speed plus 5 mph comparison
from trailer 3 before the first deployment (3 O-speed bin pneumatic tube data)

70

_r V- ,A,,,,,, .7 h , .71

speed (mph)

 

o o o
o 9 o
o ... N

l
1 start time
1

.7 ~77 777* 7~777777,1
. LT§53h26rser1tilssgeej :axeggzspgeq 153921 mszdmeiflyzng

O o O O
9 9. .o. o 9 9 .o.
m V’- 00 ON

10:00
ll 00

 

Figure 4.16. 85th percentile speed and average speed plus 5 mph comparison
from trailer 3 during the first deployment ~(16-speed bin pneumatic tube data)

The results indicated that almost all of the differences between the 85th percentile
speed and the average speed plus 5 mph were less than 1 mph when the 30-speed bin
pneumatic tube data were used in the comparisons and that almost all of the differences
between the 85th percentile speed and the average speed plus 5 mph were less than 2.5
mph when the l6-speed bin pneumatic tube data were used in the comparisons. The
number of speed bins had an effect on the value of the 85th percentile speed relative to the
average speed plus 5 mph in this verification with the more accurate results coming from
the smaller speed bin. Thus, the average speed determined from the RTMS data plus 5
mph may provide a better estimate of the 85Ill percentile speed than the results in this
verification indicated since the average speed determined from the RTMS data does not

have to be estimated fi'om “binned” data.

71

4.2 Problems and limitations faced in the VSL deployment

The problems and limitations faced in the VSL deployment are summarized in
table 4.3. The details of each problem and limitation are discussed later by each of the
VSL deployments. Number one to number seven problems and limitations in table 4.3
were first found in either the pre-deployment or the first deployment and existed in all of
the deployments, but they will be discussed only in the deployment where they were
discovered. A number in a parenthesis behind the bullet in the problem discussion
section indicates the problem number from table 4.3.

Table 4.3. Problems and limitations faced in the VSL deployment

 

 

 

Problems and limitations in 161310;“? 4
1. inconsistent length of the speed limit update period it x x x x
2. l-minute communication time between the master trailer and pagers x x x x x
3. data storage problem x x x x x
4. trailer setting on slave trailers cannot be changed remotely x x x x
5. occasionally the remote computer could not connect to the VSL x x x ' x

system

6. inaccurate occupancy measurement x x x x
7. repetitive VSL data set x x x x
8. short-range modern broke down x x
9. problem with the fixture of the static speed limit signs over VMS x
10. problem with speed-limit setting algorithm x x x x
11. speed-limit round-off error x x x x
12. improper speed limit (e.g., 54, 64 mph) x x x
13. pneumatic tubes were ripped up frequently x x x
14. RTMS broke down x
15. battery on a VSL trailer failed x x

 

 

 

 

 

 

 

 

x indicates the deployment that had that problem

72

4.2.1 Pre-deployment test

> (1) The VSL trailers were able to communicate with each other, with the pagers, and
with a remote computer. The pre-deployment testing confirmed that the VSL trailers
communicated with one another at a distance of 0.5, l, and 1.5 miles. A problem was
that it took about one minute for the VSL trailers to communicate with each other to
provide the data to the master trailer to update the speed limits. Thus, when the
programmed speed limit update period was set at 5 minutes, the actual update time
was 6 minutes instead. Moreover, the length of the speed limit update period was
also inconsistent. The length of the speed limit update period was occasionally shorter
or longer than 6 minutes. The length of the speed limit update period ranged from 2 to
15 minutes. IRD should solve this problem in the next version of the VSL system.

> (2) It took about one minute for the master trailer to transmit the updated speed limits
to the pagers. During this one minute, the speed limit that the police would have
enforced was different from the speed limit on the VSL trailer. This delay will have to
be addressed in any enforcement program.

> (3) Data storage was also a problem. The downloaded data indicated some unusual
results including incorrect data and missing data in some time periods. These errors
happened randomly. This problem was prevented by downloading and saving data
frequently. The data were downloaded approximately every three hours, except
between 12AM and 7AM. The data were only downloaded once between 12AM and

7AM.

73

> There was no problem with the speed limit setting algorithm. The VSL system set the
speed limit properly including during the rain, when the speed limit was set to be
lowered by 20 percent to test the weather sensor.

> Since the problems and the limitations that were discovered in the pre-deployment
test had no major impact on the operation of VSL system, the VSL system “passed”
the pre-deployment test.

4.2.2 First deployment

> (4) When the operational parameters and speed limit setting profiles on the slave
trailers were different from the master trailer, the operational parameters and speed
limit setting profiles on the slave trailers (i.e., trailers 2 to 6) could not be changed by
the remote computer. The changes had to be made at each trailer by connecting the
computer directly to the trailer. Only the trailer setting on the master trailer could be
changed remotely.

> (5) At random times, the remote computer could not connect to the VSL system
through the long-range wireless modem because the modem signal was not strong
enough. When this problem happened, the connection had to be made directly at the
location of the master trailer.

> (6) The occupancy measured by the RTMS is not functioning properly. Examples are
shown in table 4.4. For instance, in the first and the second examples, the average
speeds and the occupancies were the same in both the first and the second
observations, but the volumes in the first observations were much higher than the

second observations. However, the occupancies were only used to trigger the VSL

74

system to display the minimum or the maximum speed limit and the inaccurate
occupancy measurement did not affect this process.

Table 4.4. Examples of incorrect occupancies

volume occupancy 5 volume occupancy

l

 

> (7) The data set from the same VSL trailer occasionally repeats itself in two or more
consecutive periods. Examples are shown in figure 4.17. The data from trailers 2, 4,
5, 6, and 7 in these two time periods were exactly the same. The data from trailer 3 in
these two time periods were almost the same. Only one value was different. IRD

should solve this problem in the next version of the VSL system.

 

IRD Speed Ranger VSL SRWM#=lllllll Traffic Log
Date,Time,Trailer#,Angpd,DpySpd,Occ,Vol,VB1,VB2,VB3,VB4,VBS,VB6,VB7,VB8

19:46:18, W00,,2,,004,008,003,002 004,003,068

 

A shaded line indicates that the data repeats itself.

 

 

 

Figure 4.17. Examples of data that repeat themselves

75

> (8) The short-range modems broke down for a few days during the first deployment
testing period (while the static speed limit signs covered the VMS) because a bridge
in the vicinity of trailers 5, 6, and 7 blocked the line of sight. During this period, the
VSL system could only record the data from the master trailer because all of the VSL
trailers could not communicate with each other. This problem was solved by using
poles to raise the short-range modem antennas at trailer 5, 6, and 7 as shown in figure

4.18.

 

Figure 4.18. Raised short-range modem antenna at trailer 6

> (9) The static speed limit signs, which were used to cover the VMS, fell down
because the Velcro that was used to hold them was not strong enough. This problem
was solved by using plastic straps to hold the static speed limit signs.

> (10) During the first deployment test, while the static speed limit signs covered the
VMS, the displayed speed on trailer 2 was 60 mph most of the time instead of ranging

between 40 and 50 mph, as it was set. The assumption was that this problem

76

occurred because the maximum positive differential was set at a default value of 10
mph and the displayed speed on trailer 1 was overridden to display 70 mph. (The
speed limit on trailer 1 was overridden to display 70 mph at the beginning because the
70 mph speed limit sign was not available.) After the maximum positive differential
on all of the trailers was changed to 25 mph, which was the maximum value allowed
by the system at that time, the problem still existed. Moreover, when the short-range
modems broke, the displayed speed on trailer 1 was equal to the maximum positive
differential of 25 mph all the time because trailer 1 could not communicate with
trailer 2. [RD tried to fix these two problems by updating the VSL system software.
With the new software, the VSL system could be set to display preset default speeds
when there is a communication problem and the maximum positive differential can be
set at 30 mph. However, the problem with the displayed speed on trailer 2 still
existed. It appears that the VSL system did not use the new maximum positive
differential of 30 mph in setting the speed limits. The maximum difference between
the displayed speed on trailers l and 2 still would not exceed 10 mph. Finally, this
problem was avoided by programming an override on the speed limit on trailer 2 to
display 50 mph all the time. The override did not have any negative impacts because
the average speed at trailer 2 was more than 43 mph all the time and according to the
algorithm, the displayed speed on trailer 2 would have been constant at 50 mph
anyway.

(11) The displayed speeds on trailers other than trailer 2 were wrong occasionally,
when the average speeds were close to the borderline between two speed ranges (i.e.,

a round-off error). For example, when the average speed was 53 mph at trailer 3, the

77

VMS on trailer 3 displayed the wrong speed limit of 55 mph instead of the correct
speed limit of 60 mph. IRD explained that a round-off error occurred because when
the data to set the speed limit was sent to the master trailer, the average speed was
lower than the borderline speed (e.g., 53 mph), but a few faster vehicles passed the
VSL trailer during the speed limit setting process and increased the average speed to
exceed the borderline speed. A round-off error can also occur the other way around.
In this case, when the speed limit was set the average speed exceeded the borderline
speed, but a few slower vehicles passed the VSL trailer during the speed limit setting
process and decreased the average speed. Since this problem happens when the
average speed is changed by a few vehicles, it usually happens when there is low
traffic volume.

> (12) The displayed speed in the VSL data log sometimes (rarely) showed improper
numbers, such as 54 and 64 mph. However, these improper speeds were seen only in
the data log and were not actually seen on the VMS.

> (13) The pneumatic tubes that were used to collect traffic data along with the RTMS
were ripped up frequently. Thus, the data collected by the pneumatic tubes is not
complete. This problem was partially alleviated by checking the tubes regularly and
notifying MDOT immediately to fix the tubes when a problem was found.

4.2.3 Second deployment

> (10, 11) During the second deployment test, the VMS at trailers 1 to 5 were covered
with the 50-mph static speed limit signs and the VMS at trailer 6 and 7 were
overridden to display 50 and 60 mph, respectively. The displayed speeds during this

period exhibited some unexplained errors and some round-off errors. Furthermore,

78

the volume data were irregular. The traffic volumes should be either the same or
decrease trailer-to-trailer from 7 to 1 because there are only off-ramps in this area.
But the traffic volume at trailer 1 in several time periods was higher than the volume
at other trailers. Although the problems with the displayed speeds and volumes
existed, they were not serious. Thus, the decision was made to uncover the displays
and relax the override on trailers 6 and 7. That day was on Saturday so the algorithm
for “no workers present” was used.

(6) At about the same time that the override on trailer 7 was relaxed, the occupancy
value at trailer 1 increased to above 80 percent and stayed there for about one and a
half hours. This high occupancy resulted in the speed displayed on trailers l and 2 to
be 40 mph (i.e., the minimum speed limit allowed at these 2 trailers) because the high
occupancy threshold was set at 70 percent at that time. The 80 percent occupancy
was unrealistic because the average speeds were still in the 505, the volumes were
around 100 vehicles per 6 minutes, and no heavy traffic volume was observed in the
field. This problem was avoided in the next deployment by increasing the high
occupancy threshold to 90 percent. Thus, the minimum speed limit will not be
displayed unless the occupancy is at least 90 percent.

(10) When the displayed speed on trailer 2 was 40 mph, the displayed speed on trailer
3 was programmed to be constrained to be no higher than 50 mph because the
maximum negative differential was set at 10 mph. However, the VMS on trailer 3
occasionally displayed 55 mph or more.

(10) About 2 hours after the VMS was uncovered, the VMS at trailer 7 displayed 40

or 45 mph when the minimum speed limit allowed at trailer 7 was set at 60 mph.

79

> Since there were so many unexplained problems and this construction zone was
scheduled to last only 3 more days, the decision was made to abort this deployment.
The assumption was that these problems might have resulted from the way the speed
profiles were set up. Trailers 1 to 6 were set up to use the speed profile from the
master trailer, while trailer 7 was set up to use its own speed profile. In the first
deployment where there were no unexplained problems about the speed-limit setting,
all of the trailers were set up to use their own speed profile. Thus, in the future
deployments, all of the trailers should be set up to either use their own speed profile
or use the speed profile from the master trailer.

4.2.4 Third deployment

> (8) During the third deployment test, the cable for the short-range modem antenna at
trailer 7 was broken. Thus all of the trailers could not communicate with each other.
This problem was solved by setting trailer 7 to operate independently and display a
static speed limit of 60 mph. The rest of the trailers operated as a system using their
own speed profile set up at each trailer.

> (10, 11) There were some errors noted in the displayed speed, particularly at trailers 2
and 3. Although most of the errors were round—off errors, there were also some errors
that might come from the fact that the system did not take the preset value of the
maximum positive differential into consideration when the speed limits were set.
Even though the maximum positive differential between trailers 2 and 3 was set at 30
mph, the maximum positive differential would not exceed 10 mph during the first

three days that the displays were revealed to the public. After that, the maximum

80

positive differential did exceed 10 mph without any adjustments to the VSL system
and only an occasional round-off error occurred.

> (12) The improper displayed speed of 54 mph also found in the data log a few times
during this deployment.

> (14) The RTMS at trailer 5 broke during the second day after the displays were
revealed to the public. The traffic data at trailer 5 were not collected and the VMS at
trailer 5 automatically displayed a default speed limit of 50 mph. After that, the
displayed speed on trailer 5 was overridden to display 60 mph for the remainder of
the deployment. However, trailer 5 was still able to communicate with trailers 4 and 6
and pass the data back and forth.

> (15) The battery at trailer 4 failed about 14 hours before the end of the third
deployment. Only the data from trailers 1 to 3 were available after that.

> (13) The pneumatic tubes that were used to collect traffic data were ripped up
frequently. Thus the data collected by the pneumatic tubes is not complete.

4.2.5 Fourth deployment

> (15) About 4 hours after the VSL system was installed on I-96, the traffic data from
trailers 4 and 5 were missing and the displayed speed on trailer 3 was constant at 50
mph. This problem might have been caused by the loss of power at trailer 4, the same
as at the end of the third deployment. [RD was able to fix this problem a few days
later. 1

> (10, 11) The round-off errors occurred less frequently during this deployment. There

were also a few displayed speed errors that cannot be explained. For example, in one

81

6-minute period, trailers 1 to 5 all displayed 40 mph. Although the average speed
during that period was low, it was not low enough to display a speed limit of 40 mph.
> (12) The improper displayed speeds of 44 and 54 mph were also found occasionally
in the data log during this deployment.
> (13) The pneumatic tubes that were used to collect traffic data were ripped up
frequently. Thus the data collected by the pneumatic tubes is not complete.
4.2.6 Conclusion
Even though there were many problems and limitations faced during the VSL
deployments, most of the problems and limitations could be overcome, avoided, or
alleviated and the VSL deployments were evaluated, except in the second deployment
where the unexplained speed-limit-setting problems led to the decision to abort the
deployment. Most of the problems and limitations resulted from the limitations and
unreliability of the current VSL system. IRD should be able to overcome these problems

and limitations in the next version of the VSL system.

82

Chapter 5

DATA ANALYSIS AND RESULTS

The discussion in this chapter is organized by MOE within each of the VSL
deployments. The effect of police presence, which was only tested in the fourth
deployment, is discussed at the end of this chapter. The data were separated between
weekend and weekday data for the analyses, with the weekday defined as Monday at
6:00AM to Friday at 6:00PM and the weekend defined as Friday at 6:00PM to Monday at
6:00AM.

Since the RTMS data provide only a 6-minute average speed, not individual
vehicle speeds, the data from the display trailers can only be used to determine the
average speed and the travel time through the work zone. The RTMS data are only
available “before” and “during” each VSL deployment since the trailers were removed
when the construction activity changed.

As was mentioned in the previous chapter, the pneumatic tube speed data were
not reliable so they were not used in the average speed analysis, nor in the travel time
analysis. In spite of the inconsistent speed measurements, the pneumatic tube data were
used in the examination of the 85th percentile speed, the speed variance, and the
percentage of higher speed vehicles because they are the only disaggregate data available.

Moreover, the analyses focused on the changes in the MOE, which should be reasonably

83

accurate as long as the pneumatic tube data were collected using the same installation of
the tubes. Since the pneumatic tubes were ripped up frequently, only limited pneumatic
tube data were available for the “before” versus “during” and the “during” versus “after”
comparisons. When data were available, the values of the MOE from the same time of
the day and the same day of the week in two conditions (“before” and “during” or
“during” and “after”) were plotted for comparisons. The pneumatic tubes reported the
data in either 16- or 30-speed bin format depending on which setting the MDOT used on
a particular day. Sometimes the formats “before” and “during” the deployment were
different. It should be noted that the number of speed bins available affects the estimation
of the average speed, the 85Lh percentile speed and the speed variance. For the l6-speed
bin format, the individual vehicle speed can be estimated to within 5 mph of the actual
speed, while for the 30 speed-bin format, the individual vehicle speed can be estimated to
within 2 mph of the actual speed. The available pneumatic tube data and the format of
those data are shown in table 5.1.
Table 5.1. Available pneumatic tube data

VCI'SUS VCI'SUS

deployment trailer

1

 

*The numbers in each cell (n/n) indicate the number of speed bins, with the first number indicating
the number of speed bins in the first condition (e.g., "before" in "before" versus "during") and the
second number indicating the number of speed bins in the second condition.

A shaded cell indicates that the data were not available during that time period.

84

 

81

CO

(I.

inf

5.1 Average speed

The hypotheses to be tested are that the average speed will increase with an
increase in the displayed speed and that the average speed will decrease with a decrease
in the displayed speed. The 6-rninute average speeds and 6-minute average displayed
speeds in the “before” and “during” conditions were compared. F-tests within the
AN OVA procedure in the SPSS software were performed to determine whether there
were significant differences in the average speeds between the “before” and “during”
conditions. The average speed at the preceding upstream trailer was used as a covariate in
the F-tests to account for the effect of the “entry” speed from the preceding trailer. The
covariate was used to reduce any possible bias caused by a difference in the entry speed
between vehicles in the “before” and “during” or “during” and “after” data sets. The time
periods that the average speeds and average displayed speeds were compared were
selected based on the time when VSL effects were expected to be interesting, and on data
availability. The four time periods used in the analyses were 6:00AM to 8:30AM (AM
peak), 10:30AM to 12:30PM (off-peak daytime), 4:00PM to 6:00PM (PM peak), and
8:00PM to 10:00PM (off-peak nighttime).

Since there were variations in day-to-day and trailer-to-trailer average speeds in
this work zone, the average speed profiles and average displayed speed profiles were also
compared to determine how the displayed speed reacted to the average speed changes.
5.1.1 First deployment
A. Average speed and displayed speed

The average speed profiles in the “before” and “during” conditions are compared

in figures 5.1 to 5 .4. The 6-minute average speeds at each trailer averaged over the noted

85

{11c

time period were plotted over the length of the VSL deployment area. The horizontal
axis showed the distance through the zone and the trailer locations. Each line in the
figures represents an average speed profile during the noted time period on each day.
Vehicles were proceeding through the zone from trailer 7 to trailer 1 (left to right). It
should be noted that the maximum speed limit allowed on trailers 7 and 2 was 50 mph
because they were located in advance of an on-ramp from I-69 and the traffic crossover,
respectively. The displayed speeds at these 2 trailers in the “during” condition were 50

mph most of the time, which were the same as in the “before” condition.

W 777777 h , ,7 ,,,,,, ,,,, ,.i,.,#.iiig.ia#i7.i7

V!
VI

average speed (mph)
VI
O

 

 

+ Wed, 06/05/02

-*- Thu., 06/06/02

+ Fri., 06/07/02

 

45 '
40 ‘
35
0.0 0.5 1.0 1.5 2.0 2.5 3.0
T7 T6 T5 T4 T3 T2 T1
distance relative to trailer 7 (miles)
1. + Fri., 05/24/02 + Fri., 05/31/02 + Sat, 06/01/02
-*- Sun., 06/0fl02 — Mon., 06/03/02 -°- Tue, 06/04/02 lst deployment

treatment: before/during

 

 

 

+Sat.,06/08/02_7 +Smr.,067t)_9/02 —-Mon., 06/10/02 1

 

 

Figure 5.1. Average speed profiles between 6:00AM and 8:30AM from the first
deployment

In figures 5.1 to 5.4, there was a day-to-day variation both within and between
“treatment groups”. The average speeds were somewhat higher in the “during” condition.
However, the average speeds at the first two trailers encountered by the motorist (trailers
7 and 6) in the “during” condition were slightly lower than the “before” condition. These

average speed differences were more apparent in the off-peak nighttime period. In both

86

 

the “before and “during ’ conditions, the motorists slowed down at trailers 2 and 6 where

there was a ramp and a traffic crossover, respectively in the vicinity.

O
M

 

average speed (mph) 7 “7
‘6 8

M
LII

 

 

 

 

 

 

45
40
l
' 35
0.0 0.5 1.0 1.5 2.0 2.5 3.0
T7 T6 T5 T1
distance relative to trailer 7 (miles)
Jar-170507702 + _Sat, 06/01/02 ... Sun., 06/02/02
— Mon., 06/03/02 + Tue, 06/04/02 + Wed, 06l05/02 ,5, mm,“
+ Thu., 06/06/02 + Fri., 06/07/02 + Sat, 06/08/02 treatment: before/during
+ 51111., 06/09/02: — Mon., 06/10/02 ’
Figure 5.2. Average speed profiles between 10:30AM and 12:30PM
from the first deployment
65
60
55

average speed (mph)
VI
a

 

45
40
35
30
0.0 0.5 I .0 1.5 2.0 2.5 3.0
T7 T6 T5 T4 T3 T2 Tl

distance relative to traila 7 (miles)

 

17+_1K,‘05/23/02' +Fri, 05/31/02 lo—Sar, “0670
-°-Sun..06/02/02 +Mon.,06/03/02 —Tue, 06/04/02

“02
_stdeploym
+ Wat, 06/05/02 + Thu., sz + Fri” 06’07/02 We“ “fore/during
:Sun-z 96/1941}, , _ _,._ H W .

Figure 5.3. Average speed profiles between 4:00PM and 6:00PM
from the first deployment

 

87

averageispgecd (mph)
U!
o

 

1
1 45 1
1 40 V 1
1
35
1 0.0 0.5 1.0 1.5 2.0 2.5 3.0
1 17 T6 T5 1‘4 T3 T2 TI
1 distance relative to trailer 7 (miles) '
1 ’741711705/2—3/077-3—‘FEOTS/317027 T—Jsafiefixfi
1 —«- Sun., 06/02/02 —-— Mon., 06/03/02 — Tue, 06/04/02 ls, depbymcm
1 |

 

—-— Wed, 06/05/02 + Thu., 06/06/02 + Fri., 06/07/02 mam before/during
j'” §3Lu06£08’02, hf Sufi 0834194)? ,m

 

Figure 5.4. Average speed profiles between 8:00PM and 10:00PM
from the first deployment

Besides the graphical comparisons, F-tests were performed on the average speeds
“before” and “during” the deployment. The diflemnces between the displayed speeds
“before” and “during” the deployment were also examined. The results are shown in table
5.2. The average speed difference in each tested time period shown in table 5.2 was the
difference between non-weighted average speeds in the “before” and the “during”
conditions (the traffic volume in each 6-minute period was not used in the calculation of
the average speeds). The F-test results are discussed for each trailer. The order of

discussion was based on the order that the motorists encountered the trailers.

88

Table 5.2. Results of the F-tests on the average speeds from the first deployment

time vari able
weekend

6:00AM average speed
to
8:30AM displayed speed
10:30AM average speed
to
12:30PM displayed speed
average speed
displayed speed

average speed

displayed speed

1 1

A shaded cell indicates that the difference was statistically significant with 95% confidence.
A blank cell indicates that the data was not available.

Cell values are the difference in speed (or displayed speed, during-before) in mph.

 

> In the 6:00AM to 8:30AM period, even though the displayed speed on trailer 7 (the
first trailer encountered by the driver) was 50 mph in both the “before” and “during”
conditions, the average speed exceeded 50 mph for both the weekdays and the
weekends. However, the average weekend speed in the “during” condition was
significantly lower (1.3 mph) than the “before” condition. The average speed
increased during the weekdays but the difference (0.3 mph) was not significant. In
the 10:30AM to 12:30 PM period, the displayed speed (50 mph) was also the same
during the two periods. The average speed exceeded the posted speed both “before”
and “during.” There was a small (but significant) speed reduction in both the

weekdays and the weekends when the VSL signs were activated. In the afternoon

89

peak hour, the average displayed speed was slightly lower in the “during” condition
(49.0 and 48.1 mph) with a relatively large (5.4 and 9.1 mph) reduction in the average
speed of motorists. In the 8:00PM to 10:00PM period, there was no change in the
displayed speed but a small (and significant) reduction in the average speed. At this
trailer, speeds were lower in the “during” period than the “before” period for seven of
the eight comparisons; even though the speed limit remained constant in six of these
conditions. Thus, it appears that a lighted display of the same speed limit as a static
sign resulted in a speed reduction.

The results at trailer 6 are interesting. In each of the four time periods the average
displayed speed increased to between 56.3 and 60 mph, yet the average speed
decreased or was not changed significantly. It appears that drivers are not selecting
their speed based on the display at this location.

The results at trailer 5 are consistent with the hypotheses (both the displayed speed
and the average speed were higher in the “during” condition), but the magnitude of
the change was different. An 8-10 mph increase in the average displayed speed
resulted in only a 1-2 mph increase in the average speed. Trailers 4 and 3 showed this
same phenomenon, although the increase in the average speed was slightly higher
than at trailer 5.

Even though the displayed speed on trailer 2 was 50 mph in both the “before” and
“during” conditions, the average speed “during” the deployment was slightly higher
(0.8 —2.5 mph) than “before” the deployment. However, the motorists slowed down at

this trailer in both the “before” and “during” conditions.

90

> The average speeds at trailer 1 in the “before” and “during” the deployment were only
slightly different and most of the differences were not statistically significant. The
displayed speed difference at trailer 1 is not meaningful because a blank sign was
used to cover the VMS in the “before” condition.

An example of the average speed and average displayed speed profile comparison
from 5:00PM to 6:00PM “during” the first deployment is shown in figure 5.5. The
average speed and average displayed speed profiles from a congested day (06/07/02) and
from a less-congested day (06/06/02) are shown. Figure 5 .5 illustrates that the system
was operating as designed because the displayed speeds responded well to both the day-
to-day and trailer-to-trailer variations in the average speeds. The fact that the displayed

speeds respond to existing traffic condition increases the credibility of the displayed

 

 

 

   

 

 

 

 

 

 

 

 

speeds.
7 75 77 ’77 7_7.— 777—" 77—— 7 — 7
1
1 65 .. —— — ~H~ A — m . 1
1%55 —-. 7 777777777 O 77777777777 l 77777777777 ,1 1
1 :3“ 1' \‘-
1 3 45 1
1 8' 1
1 35 .: - ~ 8 * * .. 1
25 . . . ,
0.0 0.5 1.0 1.5 2.0 2.5 3.0
1 T7 T6 T5 T4 T3 r2 r1
1 distance relative to trailer 7 (miles)
1 1 —-— Avg. Speed on 06/06/02 lst deployment
1 1 - . o- - (Avg. grsplciayeldoipggdogn 06/06/02 1 d.o.w.: Thu., Fri.
1 ——*-- vg. pee o . .
1 _ 1 it “7 i. 61g:Pispla_ys§__Spjs<10n0_6/07!92 1 VSL d’sl’lay' °n

 

 

Figure 5.5. Average speed and average displayed speed profile comparison
from 5:00PM to 6:00PM “during” the first deployment

91

B. Average speed and geometry

Since there was a traffic crossover immediately downstream trailer 2, the average
speed at trailer 2 was expected to decrease significantly from the average speed at trailer
3. The speed reduction at the crossover in the “before” and “during” conditions were
compared as shown in figure 5.6. F-tests were performed to determine whether there
were significant differences in the speed reductions between the “before” and “during”
conditions. The average speed at trailer 3 was used as a covariate to account for the effect
of the “initial” speed. The results of the F-tests are shown in table 5.3. Although the
displayed speed at trailer 2 was constant at 50 mph both “before” and “during” the VSL
operation, the speed reduction at the crossover in the “during” condition was 1.0 to 3.0
mph greater than in the “before” condition and almost all of the differences were
significant. These results indicate that the motorists gave more credibility to the speed
limit displayed on the variable message signs than the speed limit displayed on the static

signs.

 

 

 

speed reduction (mph)

 

 

 

weekday weekend weekday weekend weekday weekend weekday weekend
6-8.5AM 6-8.5AM 10.5AM- 10.5AM- 4-6PM 4-6PM 8-10PM 8-10PM
12.5PM 12.5PM

FE‘EFJ’BE‘PEJ

 

 

 

Figure 5.6. “Before” and “during” speed reduction at the crossover comparison
from the first deployment

92

Table 5.3. Results of the F—tests on the speed reductions at the crossover
from the first deployment

time reduction "
weekend

- 1.2

10:30AM - 12:30PM 1.2
4:00PM - 6:00PM
8:00PM - 10:00PM

A shaded cell indicates that the difference was significant with 95% confidence.

 

5.1.2 Third deployment
Average speed and displayed speed

The third deployment work zone was in the same part of the eastbound I-96 as the
first deployment work zone, but two-way traffic was maintained on the eastbound lanes
instead of the westbound lanes. In this deployment, the RTMS data were only available
during three time periods: 10:30AM to 12:30PM, 5:00PM to 6:00PM, and 8:00PM to
10:00PM. The short-range modem at trailer 7 broke down for a few days during the
“before” period, so the “before” data during the weekend were not available, and
“before” data exist for only one weekday. The trailer 5 data were also limited because the
RTMS at trailer 5 broke down on the second day of the deployment.

The average speed profiles in the “before” and “during” conditions are compared
in figures 5.7 to 5.9. In these figures, the “during” speed profiles were higher than the
“before” speed profile almost all the time. However, the “before” data are limited to one
day. The maximum speed limit allowed at trailers 4 and 6 was limited to 50 mph because
of the on- and off- ramps in the vicinity. This low speed limit resulted in low average
speeds through the first six trailers (trailers 7 to 2). The motorists increased their speeds
between the last two trailers, where the highway was clear and the maximum speed limit

allowed was 70 mph.

93

 

average speed (mph)
8 a 8: 8 a 5'

:18

 

 

 

 

+ Thu., 07/18/02

0.5 1.0 1.5 20 2.5
T6 r5 T4 13 r2 r1
distance relative to trailer 7 (miles)
—-— Fri, 07/12/02 -— Sat, 07/13/02 _._ Sun., o7/14/02
— Mon, o7/15/02 —-— Tue, 07/16/02 —-— Wed, o7/r 7/02 3rd deployment

treatment: before/during

 

 

Figure 5.7. Average spwd profiles between 10:30AM and 12:30PM

from the third deployment

 

\I
3 c

8

average speed (mph)
& Ur U:
U. O U.

#

O
O
O

0.5
T6

 

 

 

0 5
T5 T4 T3 T2
distance relative to trailer 7 (miles)

 

+ 111., 07/11/02
—-— Sun., 07/14/02

1 + Wed, 07/17/02

+ Fri., 07/12/02
— Mon, 07/15/02
+ Thu., 07/ 18/02

+ Sat. 07/13/02

—-— Tue, 07/16/02 3rd deployment
treatment: before/during

 

Figure 5.8. Average W profiles between 5:00PM and 6:00PM

from the third deployment

 

 

\l
O

95
Ur

S

‘JI
0

§
LII

 

&
O

0.0 0.5 1.0 1.5 2.0 2.5
T7 T6 T5 T4 T3 T2 T1

distance relative to trailer 7 (miles)

CREW/17132 :Ffi.,07/12(02— 7 +Sat., 07/13/02

- Sun.. 07/14/02 — Mon., 07/15/02 -— Tue, 07/16/02 3rd dcpbymem
lrcatrnrent before/during
+ Wed, 07/17/02 + Thu., 07/18/02

Figure 5.9. Average speed profiles between 8:00PM and 10:00PM
from the third deployment

44_-__.-nm _ mam.“ 4. “fl
average speed (mph) ‘

UI

UI

 

1
1

The F-test results are shown in table 5 .4. The results from the third deployment,
particularly in the off-peak periods, are similar to those found in the first deployment. In
the off-peak periods (10:30AM-12:30PM and 8:00PM-10:00PM), relatively large
increases in the displayed speeds resulted in small but statistically significant increases in
the average speeds. In the PM peak period (5:00-6:00PM), relatively large increases in
the displayed speeds resulted in small and insignificant decreases in the average speeds.
However, these same small decreases in the average speeds also occurred when there was
no change in the displayed speed. The F-test results indicate that the motorists’ responses
to the changes in the displayed speeds were more consistent during the non-peak hours

than during the peak hours.

95

Table 5.4. Results of the F-tests on the average speeds from the third deployment

location

time variable

weekend
1
12: 30PM
5 :OOPM-

8:00PM-

10:00PM . .
A shaded cell indicates that the difference was statistically significant with 95% confidence.
A blank cell indicates that the data was not available.
Cell values are the difference in speed (or displayed speed, during-before) in mph.

    

An example of the average speed and average displayed speed profile comparison
from 2:00PM to 3:00PM “during” the third deployment is shown in figure 5.10. Figure
5.10 illustrates that the displayed speeds responded well to both the day-to-day and
trailer-to-trailer variations in the average speeds, particularly between trailer 1 and trailer
3. The displayed speeds at trailers 4, 5, 6 and 7 were almost the same on both days, even
though the average speeds at these trailers differed by approximately 2-3 mph. The
responsiveness of the displayed speeds to the day-to-day and trailer—to-trailer average
speed variations at trailers 4 to 7 was not as good as at trailers 1 to 3 because the
displayed speeds at trailers 4 to 7 were dictated by the locations of on and off-ramps in

the vicinity.

96

 

 

\lf
v.1

 

 

 

 

 

 

 

‘ 1
1
: 70-~-~—— — -—~ — —~~ —~ __-- .- 1
1:365e— - fir” - e—A m - — 1
1 o- 1'
1 g 60 .. .. ' - 1
1B 55 L
1 0
1 8‘ 50 w 1
. 45 - ~ , - - — g 1
40 . . 1
1 0.0 0.5 1.0 1.5 2.0 2.5 1
1 T7 . .T5 T4 T3 . T1 1
1 distance relative to trarler 7 (miles) 1
' 1 __.—7 fivg' Isfeeld 0111037 /1 2:102 07/12/02 1 3rd deployment
1 - - '- - vg. rsp aye pee on . .
1 —*— Avg. Speed on 07/15/02 d.o.w...Fr1., Mon.
.1 f e 3.: .égggéiplaasgflggw§/g%_m VSL d'sP'aV °“_

 

 

 

 

Figure 5.10. Average speed and average displayed speed profile comparison
from 2:00PM to 3:00PM “during” the third deployment

5.1.3 Fourth deployment
Average speed and displayed speed

The fourth deployment used two different speed-lirnit-setting algorithms, one for
day (6AM-10PM) and one for night (10PM-6AM) because the construction activities
were only carried on between 6AM and 10PM. The maximum speed limit allowed during
the day was 50 mph, while the maximum speed limit allowed at night was 70 mph. In
this deployment, the RTMS data were only available during two time periods: 10:30AM
to 12:30PM and 8:00PM to 10:00PM. Only one day of the weekend “before” data were
available because the battery at trailer 4 failed during the “before” deployment.

The average speed profiles in the “before” and “during” conditions are compared
in figures 5.11 and 5.12. The average speeds at trailers 4 and 5 ranged from 15 mph to

about 50 mph because there was traffic congestion at the beginning of the VSL

97

deployment area during some of the observation periods. The motorists increased their
speed as they traveled through the work zone. At the end of construction zone, trailer 6
displayed a speed limit of 70 mph in both the “before” and “during” conditions (the static

sign was used in the “before” condition).

ua1
one
1
1
1
1
1
1

 
 

##0-
CU-

u
U.

average speed (mph)

NW
UIO

 

N
O

0.0 0.5 1.0 1.5 2.0 2.5 3.0 1
T5 T4 T3 12 TI T6

distance relative to trailer 7 (miles)

1 + Sun.. 08/04/02 + Mon., 08/05/0i -:- Tue, 08/06/027 1

1 1
1 1
1 — Wed, 08/07/02 «~— Thu, 08/08/02 —- Fri.. 08/09/02 1 4‘“ deP'WmeT“ 1
1 1 mtmem: below/during 1
L 1 —-— Sat, 08/10/02 + Sun, 08/1 [/02 1

Figure 5.11. Average speed profiles between 10:30AM and 12:30PM
from the fourth deployment

The F-test results are shown in table 5.5. There were only a few significant
changes in the average speed “during” this deployment and the inconsistency in the
direction of change was even greater than in the first and third deployment. In the 8-
IOPM period, the average “during” displayed speed at trailer 5 was 6.1 mph lower than
the “before” period, but the average speed was 9.3 mph higher in the “during” period.
During the morning period at this same sign location, a 3.8 mph decrease in the displayed
speed resulted in an 8.8 mph reduction in the average speed. Most of the other changes
were not significant for this deployment. However, these results were based on only one

day of “before” data.

98

 

0.0 0.5 1.0 1.5 2.0 2.5 3.0 ‘
T5 T4 T3 T2 T1 T6 1

distance relative to trailer 7 (miles)

+ Sun. 08/04/02 + Mon.. 08/05/02 ... Tue. 08/06/02 1

— Wed. 08107102 + Thu. 08/08/02 + Fri. 08/09/02 4‘“ deployment
. treatment: before/11mm;
+ Sat. 08/10/02 + Sun.. 08111102 + Fri., 08/02/02 1

Figure 5.12. Average speed profiles between 8:00PM and 10:00PM
from the fourth deployment

Table 5.5. Results of the F-tests on the average speeds from the fourth deployment

time variable weekend 5

- 1
10:00PM . diff. -6.1 -l .8 -0.7 0.0 0.0 0.0
A shaded cell indicates that the difference was statistically significant with 95% confidence.
A blank cell indicates that the data was not available.
Cell values are the difference in speed (or displayed speed, during-before) in mph.

 

An example of the average speed and average displayed speed profile comparison
from 8:00PM to 9:00PM “during” the fourth deployment is shown in figure 5.13. The
average speed and average displayed speed profiles from a congested day (08/09/02) and
a less-congested day (08/10/02) are shown. The displayed speeds and the actual speeds
were both lower on the congested day. Figure 5.13 illustrates that the displayed speeds

responded well to the day-to-day and trailer-to trailer average speed variations. It should

99

 

be noted that the maximum displayed speed at trailer 5 to 1 could not exceed 50 mph

because it was governed by the construction activities.

 

 

 

 

 

 

 

 

 

 

 

i; “7,

80 1
1 701-____ n,-_i 4~ ~ i,ii-,,,_i._1
1 3 60 * ~ , A v — A a .
1 o.
1 g 50
_ B 40 ~~~~~~~~~~ 1

D
1 9a“ 30 .
’ 20 ~ ~
1
1 10 a . _ . 1
1 0.0 0.5 1.0 1.5 2.0 2.5 3.0
1 T5 _ T4 _ T3 T2 T1 T6
1 distance relative to trailer 5 (miles)
7 1*:1337 ST“ 7 08109702 A ‘1

1 1 g. pee on
1 1 - — 0' - Avg. Displayed Speed on 08/09/02 1 4th deployment
1 1 ——-— Avg. Speed on 08/1 0102 1 d-O-W- =1 Fri» Sat-
: m_ 1 _ 4-’ x;_-_Avg_. Displayed Speed on 08/10/02 1 VSL display: on

 

 

Figure 5.13. Average speed and average displayed speed profile comparison
from 8:00PM to 9:00PM “during” the fourth deployment

5.1.4 Average speed analysis conclusion

The results were most consistent with the hypotheses at locations in the work
zone where there were no ramps in the vicinity. In the first deployment, this condition
existed at trailers 3, 4, and 5. The average speeds at these trailers changed in the same
direction as the changes in the displayed speeds, but relatively large increases in the
displayed speeds resulted in only small changes in the average speeds. In contrast, there
were many ramps in the vicinity of trailers 4, 5, and 6 in the third deployment and the
average speeds at these locations were not consistent with the hypothesis that these

changes would be in the same direction as the displayed speeds.

100

The motorists apparently gave more credibility to the speed limit displayed on the
variable message signs than the speed limit displayed on the static signs. For example,
the speed limit at trailer 7 in the first deployment was constant at 50 mph in both the
“before” and the “during” conditions but the average speed in the “during” condition was
significantly lower than the “before” condition. The VMS on the VSL trailer also
resulted in a significantly greater speed reduction at the crossover (trailer 2) in the first
deployment than did the static sign.

The motorists’ responses to the changes in the displayed speeds were more
consistent during the non-peak hours than during the peak hours. For example, in the
third deployment at trailer 7, an increase in the displayed speed resulted in an increase in
the average speed in every time period, except between 5:00PM and 6:00PM. The
explanation is that during the peak hour, the motorists follow the traffic stream without
looking at the displayed speeds. Thus, changes in the displayed speeds do not affect the
average speeds.

The average speed and average displayed speed profile comparison indicates that
the VSL responded to the day-to-day and trailer-to-trailer average speed variations, and
thus the used of the VSL system should improve the credibility of the speed limit.

5.2 Travel time through the work zone

The hypothesis is that the travel time through the work zone will decrease if there
is an increase in the displayed speeds at most of the VSL trailers. Travel times were
calculated from the average speeds at all of the trailers by assuming that vehicles traveled
between any two adjacent trailers at a constant speed, which was the non-weighted

average of the average speeds at those two trailers during that six minute time period.

101

 

The travel times through the work zone in the “before” and “during” conditions were
compared. F—tests were performed using the RTMS data to determine whether there were
significant differences in the travel times between the “before” and “during” conditions.
5.2.1 First deployment

The F-tests were performed using weekday and weekend data from four time
periods: 6:00AM to 8:30AM, 10:30AM to 12:30PM, 4:00PM to 6:00PM, and 8:00PM to
10:00PM. The entry speed at trailer 7 was used as a covariate to account for the effect of
the “initial speed” on the travel time. The results of the F-tests on travel time through the
work zone are shown in table 5.6.

Table 5.6. Results of the F-tests on the travel times through the work zone
from the first deployment

 

Avg. "during" travel travel time difference (during—before)
time time (seconds) weekday weekend

weekday weekend seconds percent seconds percent
6:00AM — 8:30AM 184.9 171.3 5'4106 7" —5.4 if“ f —1.6

 

 

 

 

  

 

 

 

 

 

 

 

 

 

 

 

10:30AM — 12:30PM 188.7 177.7 ,.,_,,,'-5;.0:.;1; -2.6 .29
4:00PM — 6:00PM 192.0 205.6 7118-? 1.0 11.3
8:00PM - 10:00PM 177.7 174.5 Jazz -2.9 —3.2

 

 

A shaded cell indicates that the difference was significant with 95% confidence.

All of the differences in each of the four time periods were significant, but the
travel time between 4:00PM and 6:00PM was significantly longer in the “during”
condition, while the travel times in the other time periods were significantly shorter in the
“during” condition. Although the differences in travel time were statistically significant,
they were not operationally significant. The maximum travel time saved by the VSL
system was only 5.4 percent (10.6 seconds of 195.5 seconds travel time over the 2.7-mile
test site) and the maximum increase in travel time with the VSL system was only 11.3

percent (a 20.9-second increase in a 184.7—second travel time). The average speeds

102

through the work zone “during” the deployment for the periods with the maximum travel
time saved and the maximum increase in travel time were 52.6 and 47.3 mph,
respectively.
5.2.2 Third deployment

The third deployment work zone was similar to the first deployment work zone,
but there were more on-ramps and off-ramps, particularly in the first half of the work
zone (trailers 7 to 4). In the “before” condition, only the weekday data were available so
F—tests were performed with RTMS data during three available time periods on
weekdays: 10:30AM to 12:30PM, 5:00PM to 6:00PM, and 8:00PM to 10:00PM. The
trailer 5 data were only available between 5:00PM and 6:00PM. During the periods when
the average speed at trailer 5 was not available, the travel times were calculated by
assuming that vehicles traveled between trailers 6 and 4 at a constant speed, which was
the non-weighted average of the average speeds at those two trailers. Since trailer 7 was
set up to operate independently in this deployment (because of the short-range modem
problem), the trailer 7 data collection periods were different from other trailers, and the
travel time between trailers 7 and 6 was not included. The travel time considered in this
deployment was only the travel time between trailers 6 and 1. The entry speed at trailer
6 was used as a covariate to account for the effect of the “initial” speed on the travel time.

The results of the F-tests on travel time through the work zone are shown in table 5.7.

103

Table 5.7. Results of the F-tests on the travel times through the work zone
from the third deployment

 

 

 

 

 

 

 

 

Avg. "during" weekday travel time difference
time weekday travel (during-before)
time (seconds) seconds percent
10:30AM - 12:30PM 116.6 g??— ,1}; -3.7
5:00PM - 6:00PM 117.5 -0.6 -0.5
8:00PM - 10:00PM 111.9 fare: £6.55: .34

 

 

 

 

 

 

A shaded cell indicates that the difference was significant with 95% confidence.
All of the average travel times through the work zone in the “during” condition
were significantly shorter than the “before” condition, except 5-6:00PM. The average
travel time through the work zone between 5 and 6PM in the “during” condition was
shorter than the “before” condition, but the difference was not significant. As was the
case in the first deployment, the travel time saved by the VSL system was not
operationally significant. The maximum travel time saved by the VSL system was only
3.7 percent (4.4 seconds of 121 seconds travel time) over the 1.7-mile test site. The
average speed through the work zone “during” the deployment for this period was 52.5
mph.
5.2.3 Fourth deployment
The fourth deployment used two different speed-limit-setting algorithms, one for
day (6AM-10PM) and one for night (10PM-6AM) because the construction activities
were only carried on between 6AM and 10PM. The maximum speed limit allowed during
the day was 50 mph, while the maximum speed limit allowed at night was 70 mph. In the
“before” condition, only the weekend data were available so F—tests were performed with
RTMS data during two available time periods on weekends: 10:30AM to 12:30PM and

8:00PM to 10:00PM. Because the maximum speed limit allowed during the day was 50

mph, the displayed speeds “during” the deployment were either equal to or lower than the
displayed speed “before” the deployment (50 mph). Since trailer 6 was set up to operate
independently in this deployment (because the speed limit-setting algorithm at this trailer
was different from other trailers), the trailer 6 data collection periods were different from
other trailers and the travel time between trailers l and 6 (last 2 trailers) cannot be
calculated. The travel time considered in this deployment was only the travel time
between trailers 5 and l. The entry speed at trailer 5 was used as a covariate to account
for the effect of the “initial” speed on the travel time. The results of the F-tests on travel
time through the work zone are shown in table 5.8.

Table 5.8. Results of the F-tests on the travel times through the work zone

 

 

 

 

from the fourth deployment
Avg. "during" 1 weekend travel time difference
time weekend travel (during-before)*
time (secondS) seconds 1 percent
10:30AM — 12:30PM 194.7 24.5 | 14.4
8:00PM - 10:00PM 213.8 -129 | -5.7

 

 

 

 

 

*No difference was significant with 95% confidence.

Since the displayed speeds “during” the deployment were either equal to or lower
than the displayed speed “before” the deployment, the travel times were expected to be
longer “during” the deployment. However, the results indicate that the VSL system
resulted in an increase in travel time in one time period, and a decrease in travel time in
the other time period. The travel time differences in the fourth deployment were longer
than the previous two deployments, but the differences were not statistically nor
operationally significant. There was also a large variation between the two travel time

differences. The variation may result from the limited “before” data.

105

5.2.4 Travel time analysis conclusion

When the displayed speeds at most of the VSL trailers increased in the first and
third deployments, the travel time through the work zone decreased in six out of seven
time periods tested. Among the six decreases, five were statistically significant. However,
all of the statistically significant travel times saved were less than 6 percent and are not
considered to be operationally significant. In contrast, when the displayed speeds at the
VSL trailers either decreased or stayed the same in the fourth deployment, the effect of
the displayed speed decrease on travel times is not clear because of the limited data.
5.3 85th percentile speed

The hypotheses to be tested are that the 85‘h percentile speed (like the average
speed) will increase with an increase in the displayed speed and decrease with a decrease
in the displayed speed. The 85m percentile speed in the “during” condition was compared
with similar data from the same time periods in the “before” and “after” conditions.
Mann—Whitney Tests were performed during the same four time periods that the average
speeds were tested (6:00AM to 8:30AM, 10:30AM to 12:30PM, 4:00PM to 6:00PM, and
8:00PM to 10:00PM) to determine whether the differences in the 85th percentile speed
were statistically significant. According to Norusis (1999), the Mann-Whitney test is the
nonparametric test alternative to the independent-samples T-test. The nonparametric test
was used because it does not require detailed assumptions about the populations from
which the samples are selected.
5.3.1. First deployment

The results of the Mann-Whitney tests are shown in table 5.9. It should be noted

that the pneumatic tube data were in the 30-speed bin format in the “before” condition

106

and in the l6-speed bin format in the “during” condition. For the l6-speed bin format, the
85th percentile speed can only be estimated to within 5 mph of the actual value, while for
30 speed—bin format, the 85th percentile speed can be estimated to within 2 mph of the
actual value.

Table 5.9. Results of the Mann-Whitney tests on the 85'h percentile speeds
“before” and “during” the first deployment

weekday/

trailer time
weekend

1 (right)

weekend
10:30AM—12:30PM
weekend
4:00PM-6:00PM
weekend
8:00PM-10:00PM

6:00AM-8:30AM
10:30AM-12:30PM
4:00PM—6:00PM
8:00PM—10:00PM
6:00AM-8:30AM
10:30AM—12:30PM
4:00PM-6:00PM
8:00PM-10:00PM
*different is (during - before)
A shaded cell indicates that the difference was statistically significant with 95% confidence.
A blank cell indicates that the data was not available.

 

From the results of the Mann-Whitney tests in table 5.9, the 85th percentile speeds
at trailer I changed in the same direction as the changes in the displayed speeds as
hypothesized. The decreases in the average displayed speeds which ranged from 4.71 to
7.59 mph resulted in decreases in the 85‘h percentile speeds ranging from 2.98 to 5.75
mph. However, when the displayed speeds increased at trailer 3, the changes in the 85th

percentile speeds were inconsistent. The 85th percentile speeds at trailer 3 increased in the

107

two AM periods and decreased in the two PM periods. All of these changes in the 85th
percentile speeds at trailer 3 were small, but two of them were statistically significant.
However, the pneumatic tubes at trailers 1 and 3 did not provide consistent speed
measurements.

At trailer 7 where the displayed speed was constant at 50 mph, the 85th percentile
speeds were lower “during” the deployment. The decreases in the 85th percentile speeds
at trailer 7 were statistically significant in three out of four time periods. The results at
trailer 7 support the point that the motorists give more credibility to the speed limit
displayed on a variable message sign than the same speed limit displayed on a static sign.
5.3.2. Third deployment

The results of the Mann-Whitney tests in the “before” and “during” conditions
and the “during” and “after” conditions are shown in tables 5.10 and 5.11, respectively. It
should be noted that the data at trailer 1 were in the l6-speed bin format in the “before”
condition and in the 30-speed bin format in the “during” condition and that the speeds
measured by the pneumatic tubes at trailer 2 in the “during” condition may be about 3
mph higher than the actual speeds.

From the results of the Mann-Whitney tests in table 5.10, the 85'h percentile
speeds at trailer 1 (left lane) “during” the deployment were higher than “before” the
deployment, while the displayed speed was 70 mph both “before” and “during” the
deployment. Only the increases in the 85th percentile speeds between 4PM and 6PM

were statistically significant.

108

Table 5.10. Results of the Mann-Whitney tests on the 85th percentile speeds

“before” and “during” the third deployment

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

trailer time weekday/ avg. 85th percentile avg. displayed speed
weekend before during diff.* before during diff.*

l(left) 4:00PM-6:00PM weekday 67.50 70.00 " 230 70.00 70.00 0.00
8:00PM-10:00PM weekday 69.38 70.50 1.13 70.00 70.00 0.00

2 6:00AM-8:30AM weekday 58.20 60.10 '1 1.90; 50.00 66.79 16.79
10:30AM-12:30PM weekday 58.88 59.38 0.50 50.00 66.18 16.18
4:00PM-6:00PM weekday 58.50 59.00 0.50 50.00 66.79 16.79
8:00PM-10:00PM weekday 61.50 61.25 -0.25 50.00 67.89 17.89

3 4:00PM—6:00PM weekday 57.50 56.88 —0.63 50.00 55.00 5.00
8:00PM-10:00PM weekday 58.75 59.38 0.63 50.00 57.19 7.19

4 4:00PM-6:00PM weekday 55.00 56.88 1.88 50.00 50.00 0.00
8:00PM-10:00PM weekday 58.75 58.75 0.00 50.00 50.00 0.00

7 4:00PM-6:00PM weekday 56.25 52.50 :-3.75" 50.00 60.00 10.00
8:00PM-10:00PM weekday 57.50 55.631 -l.88 | 50.00 60.00 10.00

 

*different is (during - before)
A shaded cell indicates that the difference was statistically significant with 95% confidence.

Table 5.11. Results of the Mann-Whitney tests on the 85th percentile speeds
“during” and “after” the third deployment

 

 

 

 

 

 

 

 

 

 

 

 

 

. . weekda/ av .85th rcentile av .dis la eds ed
"all“ "me weekeni; durifig aft: | diff.* durigng in; dlifff
3 8:00PM-10:00PM weekend 60.63" 61.251063 58.56 50.00 8.50
4 8:00PM-10:00PM weekend 57.50 61.88 ‘-'S4.38i.' 50.00 50.00 0.00
7 8:00PM-10:00PM weekend 57.50 56.88| 0.63 60.00 50.00 10.00

 

*different is (during - after)
A shaded cell indicates that the difference was statistically significant with 95% confidence.

While the displayed speeds “during” the deployment were higher than the

displayed speeds “before” and “after” the deployment at trailers 2, 3, and 7, most of the

changes in the 85th percentile speeds were small and inconsistent. There were only two

statistically significant changes. The increase in the average displayed speed at trailer 2 of

16.79 mph resulted in the increase in the 85th percentile speed of 1.90 mph and the

increase in the average displayed speed at trailer 7 of 10.00 mph resulted in a decrease in

the 85th percentile speed of 3.75 mph.

109

 

 

At trailer 4 where the displayed speed was constant at 50 mph, the changes in the
85th percentile speeds were also inconsistent. The 85th percentile speeds “during” the
deployment were higher than “before” the deployment between 4PM and 6PM and equal
to “before” the deployment between 8PM and 10PM. The 85th percentile speed between
8PM and 10PM “during” the deployment was significantly lower than “after” the
deployment.

The results from the third deployment indicate that the changes in the displayed
speeds did not have a consistent impact on the changes of the 85th percentile speeds. The
85th percentile speeds changed in the same direction as the changes in the displayed
speeds only in some of the tested time periods and most of the changes in the 85th
percentile speeds were small and not significant.

5.3.3. Fourth deployment

Only the data at trailer 6 (left lane) were available for the “during” versus “after”
comparison, and the displayed speed at this trailer was fixed at 70 mph. Thus no
comparison of the 85th percentile speed was possible at this deployment site.

5.3.4. 85“1 percentile speed analysis conclusion

The 85‘h percentile speeds at almost all of the trailers did not change in the same
direction as the changes in the displayed speeds as hypothesized. There was only one
trailer (trailer 1 in the first deployment) that the changes in the displayed speeds resulted
in consistent changes in the 85th percentile speeds in the same direction. However, the

pneumatic tubes at that trailer did not provide consistent speed measurements.

110

5.4 Speed variance
The hypothesis is that the variable speed limit will lower the speed variance. The
literature review indicated that an accident involvement rate increases with an increase in
the speed variance. Thus, the lower speed variance may lead to a lower accident
involvement rate and an improvement in traffic safety. The “before” versus “during” and
the “during” versus “after” speed variance were plotted over time for comparisons.
5.4.1 First deployment
The examples of “before” versus “during” speed variance comparisons at trailers

1 (right lane) and 3 are shown in figures 5.14 and 5.15. The horizontal axis in these
figures was labeled “start time” because the times that were used to plot the speed
variances were the start time of the data collection updated periods. It should be noted
that the data were in the 30-speed bin format in the “before” condition and in the 16-
speed bin format in the “during” condition. For the l6-speed bin format, the individual
vehicle speed can only be estimated to within 5 mph of the actual speed, while for 30
speed-bin format, the individual vehicle speed can be estimated to within 2 mph of the
actual speed.
> At trailer 1 (right lane), the speed variance on both the weekday and weekend

“during” the deployment was lower than "before" the deployment.
> At trailers 3 and 7 (figure not shown), the speed variances on the weekday “during”

the deployment were higher than "before" the deployment half of the time and lower

the other half.

> The variable speed limit had no consistent impact on the speed variance.

lll

speed variance

 

L_

7—_# h M M #1148 ““88 *7

 

 

 

 

 

 

 

 

 

 

 

 

 

140 j
120 - —~ ,__.#_____
100 ”man- I”; — i_ I»; --- “in--- e e e
80 4...... l.nz2....4_..#i.r l, i...__,,.’.4,v —#.7 .27..- ...’__7 .r -2.-.“
6° 71’"—
40 -
20 -.- _ -
0 1 T 1 1 1 1 1 1 1 1 1m? .1 T T
O O O o o O o O o o o o
‘2 9 c_>_ $3. 0 o 9 9 o o ‘5’. 9
start time
lst deployment
:be
1+ Fri., 05124102 +Fri., 06107102 Segment fm/
uring

 

 

Figure 5.14. “Before” versus “during” speed variance comparison

between 12:00AM and 12:00PM from trailer I (right lane) in the first deployment

 

speed variance

 

 

 

 

 

45
40
35
30
25
20
15
10
5
0 .
888888888888
N M V ‘5 ‘0 l‘ 00 as 5 .‘-i 6i (s;
—u v—1 v—I '— v—t .- ,_. F‘ N N N N
starttime
lstdeployment
1 +Thu., 05/23/02 +Thu., 06/06/02 1tr:natment: before/
..—,- .7 7 .7... fi fi ng

 

 

 

Figure 5.15. “Before” versus “during” speed variance comparison
between 12:00PM and 12:00AM from trailer 3 in the first deployment

112

5.4.2 Third deployment
The examples of “before” versus “during,” and “during” versus “after” speed
variance comparisons at trailers 7 and 4 are shown in figures 5.16 and 5.17, respectively.
It should be noted that the data at trailer 1 were in the 115-speed bin format in the “before”
condition and in the 30-speed bin format in the “during” condition.
> At trailers 1 (lefi lane), 2, 3, 4 and 7, with the exception of a few “spikes” in the speed
variance, the speed variances on the weekday “during” the deployment were slightly
higher than "before" the deployment half of the time and slightly lower the other half.
> The results at trailers 3 and 7 on the weekend “during” the deployment were slightly

higher than " er" the deployment half of the time and slightly lower the other half.

 

 

 

 

 

 

y—"*' *8— s_,,_r,_,i, #77 _fi ,, mm
‘ so .. p _ _
70 r ‘
8 60 '
.5 50 .
§ 40
if: 30
D-
“ 20
10
0
8 8 8 8 8 8 8 8 8
"‘ ‘° K °° <34 5 a as a
v—I '— v—n .— _. N N N N
starttime
W 1 3rd deployment
-°-Thu., 07111102 —+—Thu., 07/18/02 ‘Wftmem- before/
WW ,_7 W 777 _7 _ dunng

 

 

 

 

Figure 5.16. “Before” versus “during” speed variance comparison
between 3:00PM and 12:00AM from trailer 7 in the third deployment

113

speed variance

 

 

 

o
9
o
N

 

 

 

 

 

o o o o o o o o o o
O ‘2 9 9 c3 9 9 C3 C3 9

— N m o .— N m v m to

N N N

start time
.___ r - mi if . -__ -i. 3rd deployment
+ Fri.-Sat., 07/19-20/02 1treatrnent; during]
77 : Fri-Sat, 07/ 1213102 , 7 afier

 

 

 

Figure 5.17. “During” versus “after” speed variance comparison
between 8:00PM and 7:00AM from trailer 4 in the third deployment

> At trailer 4, the speed variances on the weekend “during” the deployment were
slightly lower than "afier" the deployment with the exception of a few “spikes” in the
speed variance.
> The variable speed limit had no consistent impact on the speed variance.
5.4.3 Fourth deployment
Only the data at trailer 6 (left lane) were available for the “during” versus “afier”
comparison, and the displayed speed at this trailer was fixed at 70 mph. Thus no
comparison of the speed variance was possible at this deployment site.
5.4.4 Speed variance analysis conclusion
The variable speed limit had no consistent impact on the speed variance. At many

of the trailers tested, the speed variances “during” the deployment were slightly higher

114

than "before" and “after” the deployment half of the time and slightly lower the other
half.
5.5 Percentage of higher speed vehicles

The percentage of higher speed vehicles was used as a surrogate measure of speed
limit compliance. The hypothesis is that the percentage of higher speed vehicles will be
lower with the deployment of VSL, which indicates an improvement in speed limit
compliance. The speed limits in the “before” and “after” conditions were constant at 50
mph, while the speed limits in the “during” condition varied from 40 mph to 70 mph as
shown in table 5.12 (only trailers with available pneumatic tube data are shown). The
percentage of vehicles with a speed higher than 60 mph and 70 mph in the “during”
condition was compared with similar data from the same time periods in the “before” and
“after” conditions. Mann-Whitney Tests were also performed during the same four time
periods that the average speeds were tested (6:00AM to 8:30AM, 10:30AM to 12:30PM,
4:00PM to 6:00PM, and 8:00PM to 10:00PM) to determine whether the differences in the

percentage of vehicles with a speed higher than 60 mph and 70 mph was statistically

 

 

 

 

 

 

 

 

 

 

significant.
Table 5.12. Speed limits “during” the VSL deployments
deployment trailer speed limit

1 1 Varied between 60 and 65 mph most of the time
3 60 mph most of the time
7 Constant at 50 mph almost all the time

3 1 Constant at 70 mph
2 Varied between 65 and 70 mph most of the time
3 Varied between 55 and 60 mph most of the time
4 Constant at 50 mph almost all the time
7 Constant at 60 mph

4 6 Constant at 70 mph

 

 

 

 

 

115

 

5.5.] First deployment

The examples of the “before” versus “during” percentage of vehicles with a speed
higher than 60 mph and 70 mph at trailer 1 (right lane) are shown in figures 5.18 and
5.19, and the results from all trailers are summarized in table 5.13. The results of the
Mann-Whitney tests are shown in table 5.14. It should be noted that the speeds measured
by the pneumatic tubes at trailers 1 and 3 in the “before” condition may be about 4 mph
higher than the actual speeds (based on the comparisons with the RTMS data as discussed
in chapter 4). This inaccurate measurement resulted from the fact that the devices were
re-set after the tubes were ripped up by traffic.

The summary of results shown in table 5.13, the percentage of vehicles with a
speed higher than 60 mph and 70 mph “during” the deployment at trailers 1 and 7 was
lower than “before” the deployment, with the differences ranged from 2 to 30 percent.

From the results of the Mann-Whitney test in table 5.14, the percentage of
vehicles with a speed higher than 60 mph and 70 mph at trailer 1 was significantly lower
“during” the deployment. The percentage of vehicles with a speed higher than 60 mph at
trailer 3 “during” the deployment only changed significantly (higher) in the 4:00PM to
6:00PM time period. However, the pneumatic tubes at trailers 1 and 3 did not provide

consistent speed measurements.

116

 

 

 

 

 

 

 

 

a 100
g 90 ’1‘
f 80 V
8 70'
g 60
fi 50
g .0
.3 30
8 20
“a 10
3 0
888388388338
N M V If) ‘0 l\ 00 O\ O H N M
:— '— v—I I—I '— v—I v—a v— N N N N
start time
If lst deployment
-°-Thu., 05123102 +Fri., 05124102 1munem, before]
I itl‘flr Q6/06/0LT‘Ffifl6/9L0A during

 

 

 

Figure 5.18. “Before” versus “during” percentage of vehicles with a speed higher than 60
mph comparison between 12:00PM and 12:00AM
from trailer 1 (right lane) in the first deployment

 

% of vehicles w/speed > 70 mph

 

 

O O O O O O
o o 9 o o o
N M <- W \O l\
VI-t 1—1 v—l Ft Fl! 1.!

 

 

o o o o o o
9 ° 9 9 9 9
a 0‘ o v-! N m
"‘ "' N N N N
start time
"" “‘ lst deployment
-+- Thu., 05/23/02 + Fri., 05/24/02 t tment: be fore /
T _PAU-zwt’z + Ffi-r 06/92/02 during

 

 

 

 

Figure 5.19. “Before” versus “during” percentage of vehicles with a speed higher than
70 mph comparison between 12:00PM and 12:00AM
from trailer 1 (right lane) in the first deployment

117

Table 5.13. Results of the “before” versus “during” percentage of vehicles with a speed

higher than 60 mph and 70 mph comparisons from the first deployment

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

trailer
results speed > 60 mph speed >70 mph
weekday weekend weekday weekend
"During" was slightly higher than "before" half of the 3*
time and slightly lower the other half.
"During" was less than 2% lower than "before" most 7
of the time.
"During" was about 5% lower than "before" most of 1(right)
the time.
"During" was about 10% lower than "before" most of 7 1(right)
the time.
"During" was about 20% lower than "before" most of l(right)**
the time.
"During" was about 30% lower than "before" most of 1(right)
the time.
Only a few drivers drove faster than 70 mph "before" 3
and "during" the deployment.
The data were not available. 2, 4, 5, 6 2, 3, 4, 5, 2, 4, 5, 6 2, 3, 4, 5,
6, 7 6, 7

 

* The numbers in the table indicates the trailer numbers.

** The direction in the parenthesis next to the trailer number indicates the lane used in the speed

comparison, while the trailers with no parenthesis indicate that there was only one lane of travel at those

trailers.

At trailer 7 where the displayed speed was 50 mph almost all the time, the

percentage of vehicles with a speed higher than 60 mph and 70 mph was lower “during”

the deployment. The differences in the percentage of vehicles with a speed higher than

60 mph at trailer 7 was statistically significant in three out of four time periods. The

results at trailer 7 support the point that the motorists give more credibility to the speed

limit displayed on a variable message sign than the same speed limit displayed on a static

sign.

118

 

 

Table 5.14. Results of the Mann-Whitney tests on the percentage of vehicles with
a speed higher than 60 mph and 70 mph “before” and “during” the first deployment

weekday/ > >
weekend

trailer time

1
10:30AM-12:30PM
4:00PM-6:00PM
8:00PM-10:00PM

6:00AM—8:30AM
10:30AM-12:30PM
4:00PM—6:00PM
8:00PM-10:00PM
6:00AM-8:30AM
10:30AM-12:30PM
4:00PM-6:00PM
8:00PM-10:00PM
*different is (during - before)
A shaded cell indicates that the difference was statistically significant with 95% confidence.

5.5.2 Third deployment

The examples of the “before” versus “during,” and the “during” versus “after”
percentage of vehicles with a speed higher than 60 mph and 70 mph at trailers 2 and 7 are
shown in figures 5.20 to 5.23, and the results from all trailers are summarized in tables
5.15 and 5.16. The results of the Mann-Whitney tests in the “before” and “during”
conditions and the “during” and “after” conditions are shown in tables 5.17 and 5.18,
respectively. It should be noted that the speeds measured by the pneumatic tubes at trailer

2 in the “during” condition may be about 3 mph higher than the actual speeds.

119

 

 

 

N
U!

y—nu—t
OM

% of vehicles w/speed > 60 mph
u. 8

 

O

 

 

 

 

 

8 8 8 8 8 8 8 8 8 8 g g
O' ‘“ V" "7 ‘5 P~ <w G\ <3 ;: 55 g;
start time
1 3rd deployment
... Wed, 07/10/02 + Wed, 07/17/02 Smell“ before/
unng

 

 

 

 

Figure 5.20. “Before” versus “dining” percentage of vehicles with a speed higher than
60 mph comparison between 12:00PM and 12:00AM
fi'om trailer 2 in the third deployment

 

 

70 mph
U)
o

2.5

 

% of vehicles w/speed >
S

 

 

 

 

 

0.5
0.0 . ,
O O O O O O O O O O o o
N éa <r‘ 01‘ \o (x 66 64 5 .'_: a; 5;
start time
[ ' 7 7 T 3rd deployment
—.— Wed., 07/10/02 _.— Wed, 07,17/02 treatment. before/
I dunng

 

 

 

Figure 5.21. “Before” versus “during” percentage of vehicles with a speed higher than
70 mph comparison between 12:00PM and 12:00AM
from trailer 2 in the third deployment

120

 

NNu
OUIO

 

% of vehicles w/speed > 60 mph
5

 

 

 

 

10
5
0
O o o C
o ... N M
N N N N
start time
3rd deployment
LP“ 07/19/02 -°- Fri., 07/12/02 wtreatinent: dunng/
_-____ w _j alter

 

 

 

 

 

 

 

Figure 5. 22. “During” versus “afier” percentage of vehicles with a speed higher than
60 mph comparison between 8: 00PM and 12: 00AM
from trailer 7 in the third deployment

l
l

 

% of vehicles w/speed > 70 mph

 

 

 

 

o o o o o o o o
o O —1 ... N M M
N N N N S" N N N
start time
3rd deployment
L—o— 121i 07/19/02 + Fri., 07/12/02 treatment: dunng/

 

 

 

 

 

 

 

Figure 5.23. “During” versus “afier” percentage of vehicles with a speed higher than
70 mph comparison between 8:00PM and 12:00AM
from trailer 7 in the third deployment

121

 

 

Table 5.15. Results of the “before” versus “during” percentage of vehicles with a speed

higher than 60 mph and 70 mph comparisons from the third deployment

 

results (weekday only)

trailer

 

speed>60

speed>70

 

"During" was about 10% higher than "before" most
‘ of the time.

l(left)*,**

l(left)

 

"During" was less than 5% higher than "before" most
of the time.

4

 

"During" was slightly higher than "before" half of the
time and slightly lower the other half.

2,3

2,3

 

"During" was less than 5% lower than "before" most
of the time.

7

 

Only a few drivers drove faster than 70 mph "before"
and "during" the deployment.

4,7

 

The data were not available.

 

5,6

5,6

 

* The numbers in the table indicates the trailer numbers.

** The direction in the parenthesis next to the trailer number indicates the lane used in the
speed comparison, while the trailers with no parenthesis indicate that there was only one lane

of travel at those trailers.

Table 5.16. Results of the “during” versus “after” percentage of vehicles with a speed

higher than 60 mph and 70 mph comparisons from the third deployment

 

results (weekend only)

trailer

 

speed>60

 

"During" was slightly higher than "after" half of the
time and slightly lower the other half.

3, 7*

speed>70
3

 

"During" was at least 5% lower than "after" most of
the time.

4

 

"During" was at least 10% lower than "after" most of
the time.

 

Only a few drivers drove faster than 70 mph "during"
and "after" the deployment.

7

 

The data were not available.

 

 

l(left)**, 2, 5,
6

l(left), 2, 5, 6

 

* The numbers in the table indicates the trailer numbers.

** The direction in the parenthesis next to the trailer number indicates the lane used in the
speed comparison, while the trailers with no parenthesis indicate that there was only one lane

of travel at those trailers.

 

 

 

Based on the summary of results from the “before” versus “during” comparisons
in table 5.15, the percentage of vehicles with a speed higher than 60 mph and 70 mph at

trailer 1 (left lane) “during” the deployment was higher than “before” the deployment.

122

However, the results of the Mann-Whitney tests indicate that only the difference in the

percentage of vehicles with a speed higher than 70 mph between 4PM and 6PM was

statistically significant.

Table 5.17. Results of the Mann-Whitney tests on the percentage of vehicles with
a speed higher than 60 mph and 70 mph “before” and “during” the third deployment

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

*different is (during - before)
A shaded cell indicates that the difference was statistically significant with 95% confidence.

 

 

 

 

 

trailer time weekday/ % speed > 60 mph % speed >70mph I
weekend before during diff.* before during diff.*

l(left) 4:00PM-6:00PM weekday 87.67 93.45" 5.78 9.37 15.93“
8:00PM-10:00PM weekday 86.15 91.97 5.82 10.82 18.57 7.76

2 6:00AM-8:30AM weekday 10.28 17.90 ,;,"'7.62 0.25 0.27 0.01
10:30AM-12:30PM weekday 9.85 11.22 1.36 0.15 0.09 -0.07
4:00PM-6:00PM weekday 9.69 9.89 | 0.20 0.05 0.09 0.05
8:00PM-10:00PM weekday 23.58 22.31 E127 0.35 0.57 0.23

3 4:00PM-6:00PM weekday 5.14 3.39 "‘.-_1.75‘ 0.23 0.08 015
8:00PM-10:00PM weekday 15.04 11.29| -3.75 0.76 0.51 0.26

4 4:00PM-6:00PM weekday 2.26 4.65 2:40" 0.00 0.00 0.00
8:00PM-10:00PM weekday 11.10 0.31 0.31

7 4:00PM-6:00PM weekday 4.36 0.00 0.00
8:00PM-10:00PM weekday 8.14 0.00 -0.35

 

 

 

Table 5.18. Results of the Mann-Whitney tests on the percentage of vehicles with
a speed higher than 60 mph and 70 mph “during” and “after” the third deployment

 

 

 

 

 

 

 

 

 

 

 

 

 

trailer time weekday/ % speed > 60 mph % speed >70mph I
weekend during after I diff.* during after diff.*

3 8:00PM-10:00PM weekend 18.11" 20.53 -242 2.33’ 0.99 #1341
4 8:00PM-10:00PM weekend 8.25 22.48 :51423; 0.40 1.17 0.7.7-
7 8:00PM-10:00PM weekend 4.47 6.12 -l.66 0.00 0.17 -O.%

 

 

*different is (during - after)
A shaded cell indicates that the difference was statistically significant with 95% confidence.

The percentage of vehicles with a speed higher than 60 mph and 70 mph at

trailers 2, 3, 4, and 7 in both the “before” and “during” conditions was only slightly

different (about 5 percent or less). Nevertheless, half of the differences in the percentage

of vehicles with a speed higher than 60 mph at these trailers were statistically significant.

123

 

 

For example, the percentage of vehicles with a speed higher than 60 mph at trailer 2 in
the “during” condition between 6:00AM and 8:30AM was 7.62 percent higher than the
“before” condition. However, the speed limit at trailer 2 varied between 65 mph and 70
mph so this increased percentage does not mean there was a decrease in the speed limit
compliance. At trailer 7 where the displayed speed was fixed at 60 mph, the percentage
of vehicles with a speed higher than 60 mph “during” the deployment was significantly
(statistically) lower than “before” the deployment in both of the time periods tested.

Based on the comparisons, the percentage of vehicles with a speed higher than 60
mph and 70 mph at trailers 3, 4, and 7 in both the “during” and “after” conditions was
also only slightly different, with the exception of the percentage of vehicles with a speed
higher than 60 mph at trailer 4. The percentage of vehicles with a speed higher than 60
mph at trailer 4 between 8PM and 10PM “during” the deployment was 14.23 percent
lower than “after” the deployment and the difference was statistically significant. Almost
all of the percentage of vehicles with a speed higher than 60 mph and 70 mph decreased
“during” the deployment (relative to “after”).

The results indicate that the VSL resulted in an improvement in the percentage of
higher speed vehicles at trailer 3, 4, and 7. The magnitude of improvement varied and
depended on the time periods and the conditions under which the data were compared
(i.e., “before” versus “during” or “during” versus “after”). However, the VSL also
resulted in a higher percentage of higher speed vehicles at some locations, particularly at

trailer 1.

124

 

‘filfi' 1 :"

5.5.3 Fourth deployment

Only the data at trailer 6 (left lane) were available for the “during” versus “after”
comparison, and the displayed speed at this trailer was fixed at 70 mph. Thus no
comparison of the percentage of higher speed vehicles was possible at this deployment
site.
5.5.4 Percentage of higher speed vehicles analysis conclusion

Even though the results from the first deployment were promising, the pneumatic
tube data used at 2 of the trailers were not reliable. However, the result at trailer 7 where
the data were reliable supported the point that the motorists give more credibility to the
speed limit displayed on the variable message signs than the speed limit displayed on the
static signs. In the third deployment, the VSL resulted in an improvement in the
percentage of higher speed vehicles at trailer 3, 4, and 7, but the VSL also resulted in the
higher percentage of higher speed vehicles at trailer 1. Among all of the data tested, the
effects of the VSL on the speed limit compliance were somewhat positive but the VSL
also had negative impact at a few trailers. However, this conclusion was based on very
limited data.
5.6 Effect of police presence

The police patrol vehicles were only present in the work zone during part of the
fourth deployment. The average speeds, the speed variances, and the percentage of higher
speed vehicles at the two trailers near the patrol vehicle locations, which were trailers l
and 6, were compared. The RTMS data were used in the average speed comparison,
while the pneumatic tube data were used in the speed variance and the percentage of

higher speed vehicles comparisons. The hypotheses are that the average speed, the speed

125

 

variance, and the percentage of higher speed vehicles will be lower with the presence of
the police enforcement. The police patrol vehicles were present one day “during” the
VSL operation and one day “after” the VSL operation, unfortunately data were only
available for comparison “during” the VSL operation. The data “with” and “without” the
police presence came from different days of the week because data from the same day of
the week were not available.

The locations of the patrol vehicles are shown in figure 3.10 (page 47) and the
times that the patrol vehicles were present are shown in table 5.19. The patrol vehicles
were supposed to be stationary in each time period, but the patrol vehicles were actually
moving back and forth on I-96 between M-43 and Grand River Avenue “during” the VSL
deployment between 6PM and 10PM.

Table 5.19. Times that the patrol vehicles were present in the fourth deployment

 

date time location
8/7/2002 10:00 AM - 2:00 PM I-96WB at Grand River Avenue (1*)
("during") 6:00 PM - 10:00 PM Moving back and forth on I-96 between
M-43 and Grand River Avenue.
10:00 PM - 2:00 AM I-96WB at 1-69 overpass (2*)
8/13/2002 10:00 AM - 2:00 PM I-96WB at Eaton Highway (3*)
("after") 6:00 PM - 10:00 PM I-96WB at I-69 overpass (2*)

10:00 PM - 2:00 AM I-96WB at Eaton Highway (3*)
* Map of the locations can be found in figure 3.11.

 

 

 

 

 

 

 

 

 

 

 

The speed limit on trailer 1 “during” the fourth deployment was 50 mph most of
the time between 6AM and 10 PM, and 60 mph most of the time between 10PM and
6AM. Thus the percentage of vehicles with a speed higher than 60 mph was used to
measure the effect of police presence at this trailer. The speed limits at trailer 6 “during”
the fourth deployment were constant at 70 mph. Thus the percentage of vehicles with a

speed higher than 70 mph was used to measure the effect of police presence at this trailer.

126

WW. m

 

 

Examples of the effect of police presence on the average speed, the speed variance, and
the percentage of vehicles with a speed higher than 60 mph from trailer 1 are shown in
figures 5.24 to 5.26.

At trailer 1, the average speed with the police present was ~2 mph higher than
when they were not present. The percentage of vehicles with a speed higher than 60 mph
with the police present was also higher than without the police, but the speed variance
with the police present was slightly lower (than without).

At trailer 6 (left lane), the average speed with the police present was about 3 mph
higher than without the police. The percentage of vehicles with a speed higher than 70
mph with the police was also higher than without the police, but at this trailer the speed
variance with the police was slightly higherthan without the police about halfof the time

and slightly lower the other half.

f

60 ,
55
50 4.
45 if

40

average speed (mph)

35 j

 

30 ' 4 . .
l0:00 12:00 14:00 16:00 18. 00 20: 00 22:00 200
start time

4th _eployment
IL: Wed, 8/07 (_wNSL, w/police) —-Thu., 8/08 (WI—VSL, w/o policefl treatmzent _uring

Figure 5.24. Effect of police presence on the average speed
between 10:00AM and 2:00AM from trailer 1 in the fourth deployment

 

 

 

 

127

 

variance

 

12:00 14:00 16:00 18:00 20:00 22:00
start time

 

l

 

 

C Wed, 8/07 (wNSL, w/police) — Thu., 8/08 (w/VSL, w/o police)

 

0:00

4th deployment
treatment: during

 

Figure 5.25. Effect of police presence on the speed variance
between 10:00AM and 2:00AM from trailer 1 in the fourth deployment

 

12

__"4 ‘1

percentage of vehicles with speed > 60 mph .

10:00

 

12:00 14:00 16:00 18:00 20:00 22:00

 

 

 

 

 

 

0:00

start time
—~ —- -- 4th d 1 t
I [... Wed., 8/07 (wNSL, w/police) — Thu., 8/08 (wNSL, w/o policél "”2“??ng

Figure 5.26. Effect of police presence on the percentage of vehicles with a speed higher

than 60 mph between 10:00AM and 2:00AM
from trailer 1 in the fourth deployment

128

 

 

In conclusion, the presence of60 the police resulted in higher average speeds and
an increase in the percentage of higher speed vehicles. The police presence had little
effect on the speed variance. The results were not the same as hypothesized. The values
of the average speeds, the speed variances, and the percentage of higher speed vehicles
“with” the police presence relative to “without” the police presence were the same
regardless of the different locations of the police patrol vehicles or the different displayed
speeds. However, this conclusion is based on very limited data and the pneumatic tubes
provide inconsistent speed measurements. Moreover, the differences in the average
speed, the speed variance, and the percentage of higher speed vehicles may result from
the fact that the data “with” and “without” the police presence came from different days

of the week.

129

 

Chapter 6

CONCLUSIONS AND RECOMMENDATIONS

6.1 Conclusions

This study of the effectiveness of the VSL system under various work zone
conditions is based on data collected during a field test in work zones in Michigan. The
goals of the VSL system deployment were to increase the speed limit compliance,
increase the credibility of the speed limits, improve safety, and improve traffic flow in the
work zone. The measures of effectiveness used in this evaluation were changes in the
average speed, the travel time through the work zone, the 85th percentile speed, the speed
variance, and the percentage of higher speed vehicles. Data “before,” “during,” and
“after” the VSL operation were compared. The data came from two sources and each
was used differently. The RTMS (system) data were used in the evaluation of the
average speeds and the travel time through the work zone, while the MDOT pneumatic
tube data were used in the analysis of the 85th percentile speed, the speed variance, and
the percentage of higher speed vehicles.

Based on the analysis of the average speed and the travel time through the work
zone, it was determined that the VSL system works well at locations in the work zone
where there were no geometric features that limited the speed choice of motorists, like

on-ramps and off-ramps. The average speeds at these locations changed in the same

130

 

direction as the changes in the displayed speeds. However it required relatively large
increase in the displayed speed to generate small change in the average speed. For
example, an 8 to 10 mph increase in the average displayed speed resulted in only a 1-2
mph increase in the average speed at trailers 3, 4, and 5 in the first deployment. The
increase in the average speed at these locations resulted in statistically significant
reductions in travel times through the work zone. However, these travel time reductions
may not be operationally significant because the magnitude of these reductions ranged
from 2.8 seconds in deployment one to 12.9 seconds in deployment four. The motorists
responded to the changes in the displayed speeds more consistently during the non-peak
hours than during the peak hours. They also gave more credibility to the speed limit
displayed on the variable message signs than the speed limit displayed on the static signs
during the non-peak hours. The VSL responded to the day-to-day and trailer-to-trailer
average speed variations, and thus improved the credibility of the speed limit.

Based on the analyses with the pneumatic tube data, the 85th percentile speeds at
almost all of the trailers did not change in the same direction as the changes in the
displayed speeds as hypothesized. The effect of the variable speed limit on speed
variance was inconsistent. However, the percentage of higher speed vehicles at most of
the trailers either stayed the same or became lower “during” the VSL deployments. This
occurred even when the variable speed limit was often higher than the static speed limit.
This analysis of the percentage of higher speed vehicles supports the conclusion that the
motorists give more credibility to the speed limit displayed on the variable message signs

than the speed limit displayed on the static signs.

131

 

All of these conclusions were based on very limited data from the pneumatic
tubes, which did not provide consistent speed measurements. Thus some of the
differences in the 85th percentile speed and the percentage of higher speed vehicles may
be at least partly due to errors in the pneumatic tube data.

The effect of police presence along with the VSL system was tested in the fourth
deployment. The presence of police resulted in higher average speeds and a higher
percentage of higher speed vehicles. The police presence had no significant effect on the
speed variance. The values of the average speed, the speed variance, and the percentage
of higher speed vehicles “with” the police relative to “without” the police were the same
regardless of the different locations of the police patrol vehicles or the different displayed
speeds. However, these conclusions are based on very limited data. Moreover, the
differences in the average speed and the percentage of higher speed vehicles may result
from the fact that the data “with” and “without” the police presence came from different
days of the week.

6.2 Recommendations

Although the ability of a VSL system to alter the average speed and the travel
time through the work zones appears promising, particularly at locations in the work zone
where there are no ramps in the vicinity of the trailer, the impact of the VSL system on
the 85th percentile speed, the speed variance, and the speed limit compliance is still
unclear. Recommendations to resolve these issues in future researches on the evaluation

of the VSL system follow.

132

 

> The pneumatic tubes that were used to collect traffic data were ripped up frequently
and when reset provided inconsistent speed measurements, more reliable systems that
can provide accurate individual vehicle speed measurements should be used.

> The data “before” the deployment in this study were collected, with the VSL trailers
present but with the VMS covered by static speed limit signs. Thus, the “before”
versus “during” comparisons actually tested the trailers with VMS against trailers
with static signs. A better comparison of static signs with a variable speed limit
system would have the trailers present in the work zone only “during” the
deployment. This would require a separate system (independent from the VSL trailer)
to collect individual vehicle speed and traffic data “before,” “during,” and “after” the
deployment.

> Deploy and evaluate the VSL system in work zones that last at least 4 weeks. The
“before” data could then be collected in the first week. The VSL system could then be
installed and calibrated, and the speed limit-setting algorithm checked during the
second week. The “during” data could be collected in the third week and the “after”
data in the fourth week after the trailers are removed from the work zone.

In addition to the recommendations for when the VSL system is used for research
purpose, the International Road Dynamics Inc., the manufacturer of the VSL system,
should consider the following recommendations to improve the operation of the VSL
system.
> The VSL system communication should be improved. This includes the

communication between trailers, the communication between the master trailer and a

remote computer, and the communication between the master trailer and pagers. The

133

 

VSL system frequently experienced communication problems between trailers and
between the master trailer and the remote computer. Moreover, it took the current
VSL system about one minute for the master trailer to transmit the updated speed
limit to the pagers. During this one minute, the speed limit that the police enforce
may be different from the speed limit displayed on the VSL trailers.

The consistency in the speed limit update period should be improved. In this study
the speed limit update period was set to 5 minutes, but the actual update time was
approximately 6 minutes instead. The length of the speed limit update period was
occasionally shorter or longer than 6 minutes, and ranged from 2 minutes to 15
minutes.

The flexibility of the speed limit-setting algorithm should be increased. Users should
be able to set different algorithms for each slave trailer from the remote computer.
The system should allow users to set different algorithms for different time periods
and the system should be able to change the algorithms automatically at preset times.
The speed limit setting mechanism should be more reliable. The VSL system
occasionally did not take some specified operational parameters (e.g., a maximum
positive differential) into consideration when the speed limit was set.

The battery used to provide power to the VSL system should be more reliable or there
should be a reserve battery. The battery at trailer 4 failed in both the third and fourth

deployment.

134

 

BIBLIOGRAPHY

Aljanahi, A. A. M., A.H. Rhodes, and A.V. Metcalfe (1999) Speed, speed limits and road
traffic accidents under free flow conditions. Accident Analysis and Prevention,
Vol. 31, PP 161-168.

Bell, CA. (2001) Oregon Green Light CVO evaluation final report executive summary.
Transportation Research Report No. 00-21, Transportation Research Institute,
Oregon State University, Corvallis, OR.

Cirillo, J .A. (1968) Interstate system accident research study H, Interim report 11. Public
Roads, Vol. 35, August, pp 71-75.

Committee for Guidance on Setting and Enforcing Speed Limits (1998) Managing speed:
review of current practice for setting and enforcing speed limits. Transportation
Research Board Special Report 254. National Research Council, Washington,
DC.

EIS Electronic Integrated Systems Inc. (2000) RTMS user manual.

Fildes, B.N., G. Rumbold, and A. Leening (1991) Speed behavior and drivers’ attitude to
speeding. Report 16. Monash University Accident Research Center, Monash,
Victoria, Australia, June.

French, A. and D. Solomon (1986) Traffic data collection and analysis: methods and
procedures. Synthesis of Highway Practice 130. National Cooperative Highway
Research Program, Washington, DC.

Garber, NJ. and R. Gadiraju (1989) Factors affecting speed variance and its influence on
accidents. Transportation Research Record 1213, pp 64-71.

Graham-Migletz Enterprise, Inc. (1996) Procedure for determining work zone speed
limits. Research Results Digest, No. 192. National Cooperative Highway
Research Program, Washington, DC.

Harbord, B. and J. Jones (1996) Variable speed limit enforcement — the M25 controlled
motorway pilot scheme. Computing and Control Division Colloquium on Camera
Enforcement of Traffic Regulations, Institution of Electrical Engineers, London,
England.

Harbord, B. (1998) M25 Controlled Motorway — results of the first 2 years. Road
Transport Information and Control, Conference Publication No. 454, 21-23 April.

135

 

Harkey, D.L., H.D. Robertson, and SE. Davis (1990) Assessment of current speed
zoning criteria. Transportation research Record 1281, pp 40-51.

Harwick, N. K. (1989) Albuquerque variable speed limit system. Proceedings of the 26"I
Paving and Transportation Conference and Symposium on Chip Seal Practice,
University of New Mexico.

Hauer, E. (1971) Accidents, overtaking and speed control. Accident Analysis and
Prevention, Vol. 3, pp 1-13.

Hines, M. (2002) Judicial enforcement of variable speed limits. Legal Research Digest,
No. 47. National C00perative Highway Research Program, Washington, DC

Hool, J.N., S. Maghsoodloo, A.D. Veren, and DB. Brown (1983) Analysis of selective
enforcement strategy effects on rural Alabama traffic speeds. Transportation
research Record 910, pp 74—81.

International Road Dynamics Inc. (2002) Speed Ranger Safety System operator manual.

Janson, EN. (1999) Evaluation of downhill truck speed warning system on I-70 west of
Eisenhower Tunnel. Transportation Research Center, Department of Civil
Engineering, University of Colorado at Denver.

Kloeden, C.N., A.J. McLean, V.M. Moore, and G. Ponte (1997) Traveling speed and the
risk of crash involvement. Vol. I-findings. NIHVIRC Road Accident Research
Unit, University of Adelaide, Australia, November.

Lave, C. A. (1985) Speeding, coordination, and the 55 mph limit. The American
Economic Review, Vol. 75, Issue 5, pp 1159-1164.

Luoma, J. and P. Rama (1998) Effects of variable speed limit signs on speed behaviour
and recall of signs. Trafiic Engineering and Control, April.

Lyles, R.W., W.C. Taylor, and V. Sisiopiku (2002) Field test of variable speed limits in
work zones: Michigan variable speed limit deployment plan, Michigan State
University.

Manual on Uniform Traffic Control Device Millennium Edition (2001). US. Department
of Transportation, Washington DC.

Michalopoulos, P., R. Fitch, and B. Wolf (1989) Development and evaluation of a

breadboard video imaging system for wide area vehicle detection. Transportation
Research Record 1225, pp 140-149.

136

 

Michalopoulos, P., B. Wolf, and R. Benke (1990) Testing and field implementation of the

Minnesota video detection system (Autoscope). Transportation Research Record
1287, pp 176-184.

Migletz, J., J .L. Graham, I.B. Anderson, D.W. Harwood, and KM. Bauer (1999) Work
zone speed limit procedure. Transportation Research Record 1 65 7, pp 24-30.

Montagne, PE. and CA. Bell (2000) Oregon Green Light CVO evaluation: evaluation of
the downhill speed information system. Transportation Research Report No. 00-
14, Transportation Research Institute, Oregon State University, Corvallis, OR.

Norusis, M.J. (1999) SPSS® 9.0 Guide to data analysis. Prentice Hall, Upper Saddle
River, NJ.

Nuttall, I. (1995) Slow, slow, quick, quick, slow: taking the ‘stop-start out of the London
Orbital. Trafi‘ic technology International, Winter.

Park, BB. and SS. Yadlapati (2003) Development and testing of variable speed limit
logics at work zones using simulation. Transportation Research Board 2003
Annual Meeting CD-ROM, January.

Perrin, J., P.T. Martin, and W. Cottrell (2000) Effects of variable speed limit signs on
driver behavior during inclement weather. ITE 2000 Annual Meeting and Exhibit:
August 6-9, Opryland Hotel, Nashville, Tennessee, USA.

Placer, J. and A. Sagahyroon (1998) Fuzzy variable speed limit device project. US.
Department of Transportation, Federal Highway Administration, Washington
DC

Placer, J. (2001) Pre-deployment review of critical issues for variable speed limits.
http://www.cet.nau.edu/~iDl/VSLProiect/VSLPropPhaseIII.htm, Northern Arizona
University.

Rama, P. and J. Luoma (1997) Driver acceptance of weather-controlled road signs and
displays. Transportation Research Record 1689, pp 72-75.

Rama, P. (1998) Effects of weather-controlled variable speed lirrrits and warning signs on
driver behavior. Transportation Research Record 1 5 73, pp 53-59.

Rama, P., J. Luoma, and V. Harjula (1999) Distraction due to variable speed limits.
Traffic Engineering and Control, September.

Richards, S.H., R.C. Wunderlich, and CL. Dudek (1985) Field evaluation of work zone
speed control techniques. Transportation Research Record 1035, pp 66-78.

137

 

Robinson, M. (2000) Examples of variable speed limit applications. US. Department of
Transportation, Washington DC.

Solomon, D. (1964) Accidents on main rural highways related to speed, driver, and
vehicle. US. Department of Commerce, Bureau of Public Roads, July.

Smulders, S. (1992) Control by variable speed signs: the Dutch Experiment. Proceedings
of the 6“ International Conference on Road Traffic Monitoring and Control,
Institution of Electrical Engineers, London, pp. 99-103.

Swindler, K., E. Hardouin, J. Soderstorm, R. Raub, and J. Schofer (2001) Review of
literature and recent practice on variable speed limits and automated speed
enforcement. National Cooperative Highway Research Program Project 3-59:
Assessment of variable speed limit implementation issues. Northwestern
University, March.

Tudor, L.H., A. Meadors, and R. Plant (2003) Deployment of Smart Work Zone
technology in Arkansas. Transportation Research Board 2003 Annual Meeting
CD-ROM, January.

Ulfarsson, G.F., V.N. Shankar, P. Vu, F.L. .Mannering, L.N. Boyle, and M.H. Morse
(2002) Summary: Travel Aid, Washington State Transportation Center, University
of Washington, February.

Ullman, G.L. (1991) Effect of radar transmissions on traffic operations at highway work
zones. Transportation Research Record 1304, pp 261-269.

Warren, D. (20010) Variable speed limits. Speed Management Workshop, Dallas, Texas,
March 6 .

West, LB. and J.W. Dunn (1971) Accidents, speed deviation and speed limits. Traffic
Engineering, pp 52-55.

Wilkie, J.K. (1997) Using variable speed limit signs to mitigate speed differentials
upstream of reduced flow locations. Compendium: Graduate student papers on
advanced surface transportation systems. Southwest Region University
Transportation Center, College Station, TX. -

Zackor, H. (1979) Self sufficient control of speed on freeways. IT E Proceeding of the

International Symposium on Traffic Control System, Vol. 2A, Berkeley,
California, pp 226-247.

138

 

 

11311111111111