Hajiseyedjavadi, McKinnon, Fitzpatrick &Knodler| 1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
Application of Microsimulation to Model the Safety of Varied Lane Configurations at Toll
Plazas
Foroogh Sadat Hajiseyedjavadi (Corresponding Author)
Graduate Research Assistant
Department of Civil and Environmental Engineering
University of Massachusetts Amherst
139A Marston Hall
Amherst, MA 01003
Telephone: (413) 531-9400
fhajiseyedja@engin.umass.edu
Ian McKinnon
Transportation Engineer
Tetra Tech, Inc.
1 Grant Street
Framingham, MA 01701
Telephone: (508) 903-2052
ian.mckinnon@tetratech.com
Cole Fitzpatrick
Graduate Research Assistant
Department of Civil and Environmental Engineering
University of Massachusetts Amherst
139B Marston Hall
Amherst, MA 01003
Telephone: (503) 709-1727
cfitzpat@umass.edu
Michael A. Knodler Jr., Ph.D.
Department of Civil and Environmental Engineering
University of Massachusetts Amherst
142B Marston Hall
Amherst, MA 01003
Telephone: (413) 545-0288
mknodler@ecs.umass.edu
Paper Number: 15-3480 Review Committee: ANB25
Prepared for the 94th Annual meeting of the
Transportation Research Board, Washington, D.C.
January, 2015
Length of Paper:
4748 words, 5 tables and 4 figures @ 250 words each
6998 equivalent words
Hajiseyedjavadi, McKinnon, Fitzpatrick &Knodler| 2
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
ABSTRACT
Different configurations of electronic toll collection (ETC) lanes and cash lanes in toll plazas affect
drivers‟ behavior. ETC arrangements predicate lane changing behavior, path decision making and toll
plaza comprehension in addition to the overall safety of the toll plaza. The objective of this research
was to evaluate the effect of toll plaza lane configuration on safety. A secondary objective was to
validate the feasibility of using microsimulation safety analyses in a toll plaza environment. Field
video data was captured and subsequently a microsimulation study was conducted using VISSIM.
Conflicts and events that were captured from the video were used for model calibration and validation
and safety analyses. Surrogate Safety Assessment Model (SSAM) provided by Federal Highway
Administration was used as supplementary software. Vehicle trajectories from a simulated network in
VISSIM were integrated into the SSAM software to generate surrogate safety measures. Distribution
of traffic volumes, stop delays at cash lanes and reduced speed distribution at ETC lanes were used as
calibration variables. The number of conflicts was used as a validation parameter to verify the model‟s
accuracy. The effect on safety of different lane configurations was studied under five scenarios using
the toll plaza that connects Interstate 90 to Interstate 91 and Route 5 in West Springfield,
Massachusetts as a case study scenario. The results identified the safest lane configuration was the one
consisting of only ETC lanes. When ETC was not present in all lanes, the safest configurations were
ones that separated ETC lanes from other lanes.
Hajiseyedjavadi, McKinnon, Fitzpatrick &Knodler| 3
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
INTRODUCTION
Toll plazas are one of the most critical components of a roadway system for capital financing and
ongoing infrastructure maintenance revenue. In some instances toll plazas have additionally served as
traffic maintenance and congestion control strategies. Toll plazas are amongst the most complex road
structures as drivers are exposed to a large amount of information and have a short amount of time to
make a decision. Since electronic toll collection (ETC) technology has been introduced, the
complexity of toll plazas has greatly increased. According to Ayman et al.(1), Drivers now have more
decisions to make in terms of lane selection with the second payment option as they approach the
plaza. A greater mental workload is placed on the drivers and more attention is needed, which may
have a direct correlation with a higher crash and near-miss rate(1). One mitigation effort that could
alleviate the effect of this increased complexity would be optimizing the lane configuration. The term
lane configuration means placing different toll lane types in a specific order at a toll plaza(1).
Since the advent of ETC lanes, many studies have been focused on the efficiency and
performance rate of the electronic toll collection systems, but fewer works have investigated the safety
impacts regarding electronic toll collection. Each agency has its own approach on deciding where to
place ETC lanes in a toll plaza. In some states, such as New Jersey, the ETC lanes are placed in the
middle to reduce number of lane changes and potential conflicts. Some other agencies may put ETC
lanes in the farthest right and left lanes of the roadway to avoid low speed cash costumers having to
cross the ETC lanes to reach their desired lanes and exit ramps. In Florida, Texas, and Colorado, there
are All-Electronic toll booths in some cities. This configuration reduces the number of choices
available to the drivers. So fewer lane changing incentives and fewer potential conflicts are supposed
to exist in this condition. In order to service the cash customers with this configuration, the license
plates of non-ETC customers are captured and later used to send a bill for the toll. This study
investigates the different lane configuration scenarios described above in order to determine the safest
lane configuration for a ramp plaza with close incoming and outgoing ramps.
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
BACKGROUND
There are few roadway elements which rival the complexity and diversity of toll plazas. Toll plazas
rank among the most challenging driving environments, in terms of the number of conflicts areas and
the potential for conflicts and events. The driving difficulty of toll plazas stems from the large amount
of stimuli presented to drivers. The numerous signs and pavement markings guide drivers in making
lane decisions but result in a high mental workload. Adding Enclaves to traditional toll plazas has
improved the efficiency of toll collection. However, it has increased the involvement for drivers and
has impacted roadway safety. To date, there are few studies investigating safety issues at toll plazas.
One common safety analysis method involves conducting computer based simulation models. This
section details the previous work that has addressed toll plaza performance and safety.
Although ETC systems provide several advantages such as an increase in the traffic volume
served and reduction in the total amount of emissions, an increase in the probability and severity of
collisions is probable due to the higher speed within these lanes(2).
An analysis conducted using New York Thruway crash data from 1992 to 1998 by the New
York State Thruway Authority showed the number of crashes increased with the increase prevalence
of ETC lanes. The structure of the toll plaza remained the same as the time before implanting ETC
lanes. The only change to the plaza structure was the signage that was changed from Cash to E-ZPass.
However, the rate of crashes per throughput traffic volume has decreased or not changed. The
common crash types are 1-Rear-end crash, 2- Sideswipe crash, 3- Fixed object collision, 4- Back-into,
and 5- Pedestrian related crash. Rear-end crashes has the highest frequency, they occur more
frequently during peak hours and within lanes with queues. The most common reason for sideswipe
crashes and for fixed object crashes is merging movements and high speed, respectively. Usually
pedestrian related crashes have the lowest frequency at toll plazas. Above all, the effect of drivers‟
familiarity with ETC lanes has not been fully recognized (2).
Hajiseyedjavadi, McKinnon, Fitzpatrick &Knodler| 4
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
McKinnon (3)used a computer based static evaluation to conclude that drivers try to reduce
their travel time, and even a small queue at toll plaza would be an incentive for drivers to change
lanes. The research also found that drivers‟ lane decision at toll plazas is based on the relative
transaction time at ETC and cash lanes. For combination lanes, that accept both cash and ETC
costumers, motorists instinctively weigh the risk of waiting behind a cash costumer versus the risk of
waiting behind slower moving heavy vehicles in an ETC lane. Combination lanes may increase
drivers‟ inattention while at the same time reducing vehicle throughput and increasing delays (3).
According to Ayman et al.(1), toll lane type, vehicle deceleration rates, final velocity, the
number of toll lanes and volume of cross traffic between the lanes would affect where the conflict
points occur at a toll plaza. They also stated that the number of conflicts would decrease by increasing
the number of ETC lanes at a plaza resulting in more organized traffic flow through the toll plaza (1).
They acknowledged that designing the optimum lane configuration for a toll plaza is one of
the most difficult tasks in toll plaza design. Each configuration should provide services to all payment
types and not to be confusing for drivers (1).
Rear-end collisions occur where queues form, especially during peak periods. Another factor
influencing rear-end collisions, in addition to normal driver error, is that motorists lose their forward
attention to decelerating anterior vehicles while they are making their lane choice(1).
Ayman et al. furthermore discussed the benefit of additional ETC lanes at a toll plaza and
consequentially fewer queues. However, there were two major problems with ETC lanes. The first
problem was unfamiliarity among motorists who often stop at the plaza in attempt to understand the
payment methods. The other issue was speed variation between cash lanes and ETC lanes that
increased the probability of conflicts. All things considered, the ETC system decreased the level of
safety at toll plazas(1).
The throughput performance is higher in ETC lanes compared to cash lanes but the lane
changing movements between these lanes increases the probability of conflicts. To account for this
effect, Sze et al. (4) introduced a „weaving ratio‟ which is the number of lane changing movements
across ETC lanes compared to the total possible lane changing movements. They found that as the
traffic volume increased, the crash risk increased for inbound traffic and decreased for outbound
traffic. All in all, the increase in the total number of traffic crashes was less than the increase in total
traffic volume, so crash risk decreased as traffic volume increased. This may be due to a reduction in
average speed during congested conditions. Sze et al. also stated that the crash likelihood when
leaving the plaza is not sensitive to the traffic volume, because the number of interactions when
leaving the plaza is small (4).
Drivers‟ lane change behavior is a contributing factor in toll plaza conflicts and events. As a
result, it is an important component of the computer-based microsimulation study of toll plazas. As
Mudigondaet al. (5) mentioned in their study, the lane choice decision making process for a driver
depends on complex inter-vehicle conditions. The exit lane destination and queue lengths at each lane
affect drivers‟ decisions. Mudigondaet al. also stated that the utility of each lane for drivers depends
on the travel time at the plaza and the total number of decisions drivers have already made before
choosing that lane. Macroscopic simulation softwares could not capture drivers‟ lane changing
behavior. Microscopic models, such as SimTraffic, PARAMICS, and VISSIM employ driver behavior
models, but they do not have a built-in toll plaza model (5).
Russo et al. (6) utilized a toll plaza queuing model, SHAKER, to represent traffic
characteristics observed in the field. They collected demand, throughput, queue lengths, vehicle types,
lane choice, processing time, payment type, whether the vehicle arrived during a queue or not, arrival
time, departure time, and inter-arrival time between vehicles. They first selected a Measure of
Effectiveness (MOE), and then made an initial evaluation. The MOE they selected was throughput and
capacity of the toll plaza per hour. If the MOE was different in the simulated model and the field data,
key parameters were examined and calibration parameters were determined. After multiple trial and
error runs the calibration was completed. To validate the model, field MOEs of different toll plazas
were compared to MOEs from the simulation (6).
Hajiseyedjavadi, McKinnon, Fitzpatrick &Knodler| 5
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Wong et al. (7)reported that the lane searching process was the main cause of crashes. They
defined conflicts as a situation in which a vehicle needs to brake or steer suddenly to avoid a collision.
Number of conflicts and lane changing maneuvers were stated as measures of crash risk in their study
(7).
As it is stated in Smith and Wilbur Smith Associates study(8), to increase level of safety, the
speed difference corresponding to ETC and cash lane needs to be reduced and lanes with the same
payment methods should be clustered(8).
Benda et al. (9) illustrated the challenges and achievements of Illinois‟ road tolling system
switching from conventional tolling to Open Road Tolling (ORT). The main challenge was to inform
motorists as to which lanes were for conventional tolling and which lanes were for ETC. To address
these challenges, a new signing system was developed (9). According to Pisano (10), three measures
to assess the effectiveness of signs are conspicuity, sign comprehension, and sign recognition.
Measures of effectiveness for signs introduced by Illinois study were recognition distance and
comprehension time (9).
The microsimulation guidelines provided by Dowling et al. (11)is a comprehensive document
explaining the procedures and details of microsimulation modeling. According to the authors,
“microsimulationis the modeling of individual vehicle movements on a second or sub-second basis for
the purpose of assessing the traffic performance of highway and street systems, transit, and
pedestrians”(11). As a supplement to this document, Oregon Department of Transportation provided
guidelines for the VISSIM simulation software (12).
However there are not much safety analyses done using SSAM to investigate safety at toll
plazas. Some researches have been conducted using this software for safety analyses at intersections or
roundabouts and the results of SSAM showed an acceptable fit to the field data for those studies.
SSAM processes vehicles trajectory data file that is generated by microsimulation softwares in a
trajectory file format. SSAM can support the data outputs of four simulation softwares including PTV
(VISSIM), TSS (AIMSUN), Quadstone (Paramics), and Rioux Engineering (TEXAS). It has two
thresholds to define vehicle-to-vehicle conflicts. One is Time-to-Collision (TTC) with a default value
of 1.5 seconds and the other one is Post-encroachment time (PET). The values for the thresholds can
be changed by the user to fit the reality. The results would be displayed in a table representing number
of conflicts categorized in tree types (including rear-end, crossing and lane changing conflicts). The
results could also be presented in a map. The statistical comparison of two datasets of trajectory files
could be done in SSAM using t-test tab (13).
33
34
35
36
37
38
39
40
41
42
METHODOLOGY
Due to the existing need to study safety issues at toll plazas, the objective of this study was to
investigate the effect that lane configuration has on the number of conflicts and events, which was
chosen as the primary measure of effectiveness to evaluate safety.
In order to evaluate toll plaza safety, amicrosimulation model was created in VISSIM. The
base case of this study was based on the West Springfield toll plaza at Exit 4 of Massachusetts
Turnpike, Figure 1. The location provides an ideal base case given both the simplicity of the plaza (4
exiting and 4 entering lanes) as well as the intersection of two major interstates and a primary state
route (Interstate 90, Interstate 91 and State Route 5) which results in two different upstream and
downstream access points.
Hajiseyedjavadi, McKinnon, Fitzpatrick &Knodler| 6
1
2
3
4
5
FIGURE 1 West Springfield toll plaza.
The existing lane configuration at the subject toll plaza, as shown in Figure 2, is made up of
two traditional cash lanes in the far right and the far left lanes of the plaza, and two dedicated ETC
lanes in the middle.
6
7
FIGURE 2West Springfield toll plaza lane configuration.
8
9
10
11
12
13
14
15
16
17
18
19
20
21
The VISSIM model was calibrated using traffic volume distribution, traffic composition of
heavy vehicles and passenger cars, stop delay distribution at cash lanes and speed reduction at ETC
lanes. The model was then validated by comparing the number of conflicts that occurred in the
simulation versus field video data. After the model was created and validated, four scenarios
consisting of different lane configurations of ETC and cash lanes were created and compared to the
base case.
The initial conflict and event study evaluated potential safety issues at the toll plaza. Conflicts
and events were captured and defined from video collected in the field. Conflicts and events that were
observed in the video were as follows:
1.immediate lane changing maneuvers
2.hesitation to make lane decisions
3.driving slowly in E-ZPass lanes
4.stopping before the plaza and changing lanes
5.driving in reverse gear (backing up) and
Hajiseyedjavadi, McKinnon, Fitzpatrick &Knodler| 7
1
2
3
4
5
6
7
8
6. secondary conflicts (e.g. braking because of an intruding vehicle entering from another lane
which could lead to a rear end collision or lane changing collision).
After the VISSIM model was created and calibrated, a surrogate safety assessment was
conducted using Surrogate Safety Assessment Model (SSAM). This software, which was provided by
the Federal Highway Administration (FHWA), was utilized because VISSIM does not have the ability
to conduct safety analyses. Although conflicts defined in SSAM are limited to rear end conflicts, lane
changing conflicts, and crossing conflicts, the software was able to fairly represent the traffic safety
conditions and the conflicts observed at the plaza.
9
10
11
12
13
14
Data Collection
Vehicle-by-vehicle origin-destination data was collected from recorded videos from two traffic
cameras at the West Springfield off-ramp toll plaza, Exit 4 of the Massachusetts Turnpike in
December of 2012.One of the two cameras was facing towards the plaza and the other was facing
upstream toward the merging lanes entering the plaza as shown in Figure 3.
15
16
FIGURE3Camera placement and range of vision(3).
17
18
Independent variables which were gathered from the video to build and calibrate the model are
as follows:
19
20
21
22
23
24
Traffic volume and vehicle composition
Number of vehicles entering the plaza and percent of heavy vehicles (HVs) coming from each of two
entry lanes was extracted separately. In one hour,840 vehicles entered the plaza from I-90 Westbound
and 748 from I-90 Eastbound. About 6% of I-90 Westbound entering traffic and 16% of I-90
Eastbound entering traffic consisted of HVs. Additionally, 62% and 69% of the total entering traffic
from each lane used E-ZPass lanes, respectively.
25
26
27
28
Origin-Destination matrix
The two videos were recorded simultaneously, vehicles originating from each entrance lane on the first
camera were tracked to the other camera. Their lane choice and then their exit lane were documented.
An O-D matrix was created from that video.
Hajiseyedjavadi, McKinnon, Fitzpatrick &Knodler| 8
1
2
3
Dwell time
For vehicles using cash lanes, dwell time was recorded. The average dwell time for passenger cars was
3.78 seconds and21.0 seconds for heavy vehicles.
4
5
6
7
8
9
Speed
The reduced speed limit for ETC lanes is 15 mph (24 kph). The average speed of passenger vehicles
and HVs using these lanes was 18.6 mph (30 kph) and 15.5 mph (25 kph), respectively. The speeds
were collected from the field video data. The length of some pavement markings were extracted from
the field‟s map, then the timing of the vehicles travelling along those lines was recorded. The speed
was calculated using those data.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
Modeling
Since VISSIM does not have a native feature to model toll plazas, the structure of the plaza and
vehicle behavior around the plaza needed to be custom made. A group of four parallel links was
created to emulate the four lanes of the plaza and stop signs were used in the middle of cash lanes to
force vehicles to stop for a certain amount of time which was adjusted using the aforementioned dwell
time in the form of a stochastic normal distribution. Reduced speed limit zones were used in the ETC
lanes to replicate the speed reduction of vehicles in those lanes.
Static routing was used based on the traffic distribution taken from video files. This resulted in
the distribution of traffic in the model being strictly determined to match the real world conditions
observed.
Each model was simulated a total of seven times, each with a duration of ten minutes, after 30
seconds seeding period. Vehicle trajectory files were generated as an output of simulation and were
later used in SSAM to conduct conflict analyses. The runs with the maximum and minimum amount
of conflicts were excluded from the analyses so a total of five runs were reported as the result of the
model. To validate the model, the average number of rear end, lane changing and crossing conflicts
from the five SSAM runs was compared to the conflicts observed from video files. The average
number of rear end conflicts before reaching the toll plaza for a ten minute run was 8.6 and the relative
number from the video was 9 conflicts. The number of lane changing conflicts for the same situation
in the simulation was 4.4and in the reality was 5 conflicts. Neither in simulation nor in the video had
any crossing conflicts occurred. Since there was about 92.3 percent match between the number of
conflicts from simulation and from video, the model calibration parameters were accepted and the
model was considered representative of real traffic characteristics in regards to safety.
After the base model was built and validated, four additional scenarios were implemented
which varied the lane configurations. In Scenario 2, all of the lanes were dedicated ETC lanes as
shown in Figure 4. In Scenario 3, lanes 1 and 3 were dedicated ETC lanes and lanes 2 and 4 were cash
lanes. In Scenario 4, lanes 1 and 2 were dedicated ETC lanes and lanes 3 and 4 were cash lanes.
Finally in Scenario 5, lanes 1 and 4 were combined ETC and cash lanes while lanes 2 and 3 were
dedicated ETC lanes. The scenarios represented the effect of clustered payment methods of ETC and
cash lanes and the interaction zones between them. Scenario 2 was used to analyze the border of that
clustered payment method.
Hajiseyedjavadi, McKinnon, Fitzpatrick &Knodler| 9
1
2
FIGURE 4 Lane configuration of all the scenarios built in VISSIM.
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
RESULTS and CONCLUSIONS
A conflict and event study was conducted in SSAM using the trajectory output data files from VISSIM
for five different scenarios with ten minutes of simulation time. The surrogate safety measures that
were defined in SSAM are the following
 TTC: minimum time to collision value observed during the conflict
 PET: minimum post encroachment time. It is the time that takes place from when the
first vehicle involved in the conflict passes a point until the second vehicle reaches
that point.
 MaxS: maximum speed of either vehicle throughout the conflict, i.e. while the TTC is
less than the specified following distance time threshold, which is 1.5 seconds
 DeltaS: the difference in vehicle speeds at the simulation time where the minimum
TTC (time-to-collision) value for this conflict was observed
 DR: initial deceleration rate of the second vehicle
 MaxD: maximum deceleration of the second vehicle.
 MaxDeltaV: maximum difference in speed between two vehicles in the conflict. In
other words, it is the maximum difference between the speeds of the two vehicles
involved in the conflict while a conflict exists based on the SSAM thresholds that
define a conflict.
Scenarios with a higher TTC and PET and lower DR have a lower crash probability. Also,
scenarios with a lowerMaxS and lower DeltaS are expected to have a lower crash severity. A higher
value of MaxDeltaV predicts a higher severity assuming the hypothetical collision occurs between the
two vehicles in the conflict. Table 1, Table 2,Table 3, and Table 4show the results of t-tests between
the base scenario and each of the four other scenarios.
Hajiseyedjavadi, McKinnon, Fitzpatrick &Knodler| 10
1
TABLE 1 T-test Results From SSAM between Base Scenario and Scenario2
Scenario 2
E-E-E-E
SSAM
Measur
es
Mean
TTC
(Sec)
PET
(Sec)
MaxS
(m/s)
DeltaS
(m/s)
DR
(m/s2)
MaxD
(m/s2)
MaxDel
taV
(m/s)
Base Scenario
C-E-E-C
Variance
Mean.
Varianc
e
t value
t critical
Signific
ant
Mean
Differen
ce
Better
Perform
ed
Scenario
0.917
0.298
0.524
0.429
2.517
1.66
YES
0.393
2
1.36
1.257
1.057
2.348
1.139
1.66
NO
0.303
N/A
6.185
8.903
6.92
5.18
-1.665
1.66
YES
-0.735
2
2.983
2.448
4.524
7.393
-3.753
1.66
YES
-1.541
2
-0.981
4.475
-0.244
3.719
-2.074
1.66
YES
-0.737
1
-2.994
10.666
-0.702
5.9
-4.836
1.66
YES
-2.293
1
1.808
1.162
2.589
2.718
-2.961
1.66
YES
-0.78
2
Note: N/A= not applicable
2
3
4
5
6
The significance level used to conduct the t-test is 0.05. The results show that Scenario 2
which was All ETC Lane Scenario resulted in a higher value of TTC, and lower values of MaxS,
DeltaS and MaxDeltaV. This would predict that less severe collisions would occur if the lanes were
configured according to Scenario 2.
7
TABLE 2 T-test Results from SSAM between Base Scenario and Scenario3
Scenario 3
E-C-E-C
SSAM
Measur
es
Mean
TTC
(Sec)
PET
(Sec)
MaxS
(m/s)
DeltaS
(m/s)
DR
(m/s2)
MaxD
(m/s2)
MaxDel
taV
(m/s)
8
Base Scenario
C-E-E-C
Variance
Mean.
Varianc
e
t value
t critical
Signific
ant
Mean
Differen
ce
Better
Perform
ed
Scenario
0.688
0.519
0.524
0.429
1.13
1.66
NO
0.164
N/A
1.33
2.583
1.057
2.348
1.171
1.66
NO
0.273
N/A
6.285
4.95
6.92
5.18
-2.03
1.66
YES
-0.635
3
4.073
5.175
4.524
7.393
-1.302
1.66
NO
-0.451
N/A
-0.232
5.022
-0.244
3.719
0.043
1.66
NO
0.013
N/A
-0.669
7.649
-0.702
5.9
0.094
1.66
NO
0.033
N/A
2.324
1.919
2.589
2.718
-1.154
1.66
NO
-0.265
N/A
Note: N/A= not applicable
Hajiseyedjavadi, McKinnon, Fitzpatrick &Knodler| 11
1
2
3
4
The only significant difference observed between Scenario 3, which has ETC lanes in lanes 1
and 3, and the Base Scenario is that MaxS is lower in Scenario 3. The value of all other measures did
not have any significant differences in the SSAM outputs. This implies no difference in the probability
of collisions exists between these two cases.
5
TABLE 3 T-test Results from SSAM between Base Scenario and Scenario4
Scenario 4
E-E-C-C
SSAM
Measur
es
Mean
TTC
(Sec)
PET
(Sec)
MaxS
(m/s)
DeltaS
(m/s)
DR
(m/s2)
MaxD
(m/s2)
MaxDel
taV
(m/s)
6
7
8
9
10
11
12
13
Base Scenario
C-E-E-C
Variance
Mean.
Varianc
e
t value
t critical
Signific
ant
Mean
Differen
ce
Better
Perform
ed
Scenario
0.48
0.39
0.524
0.429
-0.337
1.66
NO
-0.044
N/A
0.917
1.778
1.057
2.348
-0.816
1.66
NO
-0.14
N/A
5.549
7.195
6.92
5.18
-3.841
1.66
YES
-1.372
4
3.3
4.279
4.524
7.393
-3.318
1.66
YES
-1.212
4
-0.403
2.738
-0.244
3.719
-0.807
1.66
NO
-0.159
N/A
-1.298
6.906
-0.702
5.9
-2.313
1.66
YES
-0.596
1
1.83
1.219
2.589
2.718
-3.532
1.66
YES
-0.759
4
Note: N/A= not applicable
Table 3 shows that MaxS, DeltaS, and maximum speed difference (MaxDeltaV) are
significantly lower in Scenario 4, that has 2 ETC lanes at the farthest left lanes, than in the Base
Scenario. This shows that the severity of collision in Scenario 4 is significantly less than the Base
Scenario. However, MaxD, which could represent the probability of crashes, is less in the Base
Scenario in comparison to Scenario 4. In summary, in Scenario 4 we expect to have a higher number
of collisions. However, we expect these collisions to have a less severity compared to the Base
Scenario.
Hajiseyedjavadi, McKinnon, Fitzpatrick &Knodler| 12
1
TABLE4 T-test Results from SSAM between Base Scenario and Scenario 5
Scenario 5
Comb-E-E-Comb
SSAM
Measur
es
Mean
TTC
(Sec)
PET
(Sec)
MaxS
(m/s)
DeltaS
(m/s)
DR
(m/s2)
MaxD
(m/s2)
MaxDel
taV
(m/s)
Base Scenario
C-E-E-C
Variance
Mean.
Varianc
e.
t value
t critical
Signific
ant
Mean
Differen
ce
Better
Perform
ed
Scenario
0.725
0.455
0.524
0.429
1.259
1.66
NO
0.201
N/A
1.372
2.219
1.057
2.348
1.132
1.66
NO
0.315
N/A
6.12
8.678
6.92
5.18
-1.843
1.66
YES
-0.8
5
3.673
3.815
4.524
7.393
-1.969
1.66
YES
-0.851
5
-0.519
3.02
-0.244
3.719
-0.828
1.66
NO
-0.275
N/A
-1.447
6.552
-0.702
5.9
-1.741
1.66
YES
-0.745
1
2.35
1.799
2.589
2.718
-0.842
1.66
NO
-0.239
N/A
Note: N/A= not applicable
2
3
4
5
6
7
8
9
10
11
12
As represented in Table 4, Scenario 5, which has two ETC lanes in the middle and two Combo
lanes on the sides, has significantly less severe conflicts compared to the Base Scenario. Although
MaxD shows the Base Scenario may have a lower probability of collisions than Scenario 5.
From the results of the t-test, it is found that considering both crash probability and crash
severity, All ETC Lane Scenario is the best scenario.
Three conflict types that have been studied in SSAM are crossing conflicts, rear end conflicts,
and lane changing conflicts. The result of the number of conflicts for 600 seconds of simulation time
for each scenario is provided in Table 5.The number of conflicts represented in the table below is the
sum of conflicts that takes place both before reaching the plaza and after the plaza, before divergence
of the road.
13
TABLE 5SSAM Conflicts Results for 600 Seconds Simulation
14
15
16
Base
Scenario
Scenario2
SSAM
Measures
Mean
Mean
Significant
difference
Mean
Significant
difference
Mean
Significant
difference
Mean
Significant
difference
Crossing
0
0.4
NO
0.2
NO
1.2
NO
0
NO
Rear-end
9.4
2.4
YES
7.2
NO
10
NO
5
NO
Lane
changing
5.6
4.2
NO
4.6
NO
13.4
NO
2.2
YES
Total
15
7
YES
12
NO
24.6
NO
7.2
NO
Scenario3
Scenario4
Scenario5
The number of rear-end conflicts in Scenario 2 and number of lane changing conflicts in
Scenario5are significantly lower than the Base Scenario. Since all lanes in Scenario2 are E-ZPass
lanes, drivers are not forced to make any decisions, so less confusion would occur. As all E-Zpass
Hajiseyedjavadi, McKinnon, Fitzpatrick &Knodler| 13
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
lanes have a similar throughput, there would be less incentive for drivers to change lanes. In Scenario
5, ETC customers have more lane choices, therefore the number of lane changing maneuvers was
reduced and lane changing conflicts decreased. Additionally in Scenario 2, the speed variance is lower
as compared to other configurations due to E-ZPass lane availability. As several studies have shown in
the past, the performance of E-ZPass lanes are better than other lane types, as they result in less
congestion, which is expected to decrease conflicts. This research validates these past studies and
provides further evidence that a configuration consisting of solely E-ZPass lanes would be safer than a
configuration consisting of a mixture. Although in practice with this configuration with all E-ZPass
lanes, open road tolling gantries would be used instead of a toll plaza structure so there would be no
changes in highway operation. The second best scenario could be Scenario 4 which had a lower
probability of collision as compared to Scenario 3 and the Base Scenario. This could be because,
unlike Scenario 3 and the Base Scenario, this scenario has only one ETC lane and cash lane adjacent to
each other so the speed variance in adjacent lanes are minimal. It seems that the fewer joint borders
between ETC lanes and cash lanes provide a safer configuration. However, this configuration may be
infeasible in some conditions because of the downstream weaving required to make the proper exit.
In summary, an All-ETC Lanes Scenario performs best in terms of safety for this study
location. Scenario 5, with a combination of ETC lanes and Combo lanes would be the second safest
scenario. In general, it seems that fewer lane choices and fewer incentives to change lanes would
increase safety at the site. Although for a real world implementation, a feasibility study should also be
considered before deciding on lane configuration.
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
DISCUSSION
This study proved the feasibility of modeling traffic conditions at a toll plaza and evaluating
its safety using VISSIM and SSAM. Also traffic safety has been evaluated in different lane
configurations at the toll plaza. All ETC lanes and combination of Combo lanes and ETC lanes are
found as the safest and second safest conditions, respectively. The third safest conditions belongs to
the conditions with fewer joint borders between ETC lanes and cash lanes. The results of this study
could help having a safer trend of transition if a traditional cash lane toll plaza is going to be modified
to serve ETC customers in the future.
The data used to validate and calibrate this model was from a limited period of time taken
from only one toll plaza. To validate the results of this study and extend the results to other toll plaza
conditions, more data could be collected and the analysis could be re-conducted. Different conditions,
such as in/out ramp distance and number of lanes could affect the results. The road surface and
weather conditions may play a role in drivers‟ lane choice. The video used for analysis was collected
during clear, dry conditions but drivers may drive more conservatively in more hazardous conditions.
Sensitivity analysis is another task that could be done in the future works. Thus the effect of
adding one extra lane to the road, adding one unit to the traffic volume, removing the split after the toll
plaza or changing other variables could be determined.
Conducting the same analysis with dynamic traffic assignment could be another topic of
research to be investigated in the future.
Lack of data on driver behavior is a point that needs a comprehensive research. It would be an
independent topic that its results can affect the parameters used as an input in the micro-simulation
models. Variables such as queue length, vehicle compositions in a queue, origin-destination of a
vehicle could affect drivers‟ lane choice. Micro-simulation analysis is unable to see those details,
hence a simulation study in a virtual reality world would clarify those points.
Hajiseyedjavadi, McKinnon, Fitzpatrick &Knodler| 14
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
REFRENCES
1. Mohamed, A. A., M. Abdel-Aty, and J. G. Klodzinski. Safety considerations in designing
electronic toll plazas: Case study. ITE Journal, Vol. 71, no. 3, 2001, pp. 20-33.
2. Ding, J., F. Ye, and J. Lu. Impact of ETC on Traffic Safety at Toll Plaza. Plan, Build, and Manage
Transportation Infrastructure in China. Presented at Seventh International Conference of Chinese
Transportation Professionals Congress (ICCTP), Shanghai, China, 2007, pp. 695-701.
3. McKinnon, I. A. Operational and Safety-Based Analyses of Varied Toll Lane Configurations.
MSc Thesis, University of Massachusetts Amherst, Amherst, MA, May 2013.
4. Sze, N. N., S. C. Wong, and W. F. Chan. Traffic Crashes at Toll Plazas in Hong Kong.
Proceedings of the ICE-Transport, Vol. 161, no. TR2, May 2008, pp. 71-76.
5. Mudigonda, S., B. Bartin, and K. Ozbay. Microscopic Modeling of Lane Selection and Lane
Changing at Toll Plazas. In Transportation Research Board 88th Annual Meeting. CD-ROM.
Washington, D.C., 2009, p. 18.
6. Russo, C. S. The Calibration and Verification of Simulation Models for Toll Plazas. MSc Thesis,
University of Central Florida, Orlando, Florida, 2008.
7. Wong, S. C., N. N. Sze, W. T. Hung, B. P.Y. Loo, and H. K. Lo. The Effects of A Traffic
Guidance Scheme for Auto-Toll Lanes on Traffic Safety at Toll Plazas. Safety Science, Vol. 44,
no. 9, 2006, pp. 753-770.
8. Smith, R. F., and Wilbur Smith Associates. State of the Practice and Recommendations on Traffic
Control Strategies at Toll Plazas. Manual of Uniform Traffic Control Devices. Federal Highway
Administration, Federal Highway Administration. FHWA, U.S. Department of Transportation
June 2006. http://mutcd.fhwa.dot.gov/rpt/tcstoll/index.htm. Accessed July 27, 2014.
9. Benda, J., J. Hochmuth, R. Kowshik, and L. Medgyesy. Open Road Tolling Signing Studies. In
Transportation Research Board 88th Annual Meeting Compendium of Papers. CD-ROM.
Transportation Research Board of National Academies, Washington, D.C., 2009, p. 16.
10. Pisano, P. A. Developing a Standard Approach for Testing New Traffic Control Signs. Public
Roads, Vol. 56, no. 1, June 1992, pp. 1-8.
11. Dowling, R., A. Skabardonis, and V. Alexiadis. Traffic analysis toolbox volume III: Guidelines
for applying traffic microsimulation modeling software. Final Report FHWA-HRT-04-040,
Federal Highway Administration, U.S. Department of Transportation, 2004.
12. Mai, C., C. McDaniel-Wilson, D. Norval, D. Upton, J. Auth, P. Schuytema, S. Abbot, and R.
Delahanty. Protocol for Vissim Simulation. Oregon Department of Transportation, June 2011.
http://www.oregon.gov/ODOT/TD/TP/APM/AddC.pdf. Accessed April 21, 2014.
13. Gettman, D., Pu, L., Sayed, T., & Shelby, S. G. Surrogate safety assessment model and validation:
Final report. FHWA-HRT-08-051, Federal Highway Administration, U.S. Department of
Transportation, 2008.