Journal of Traffic and Logistics Engineering Vol. 2, No.

2, June 2014

Proactive Evaluation of Traffic Safety at An

Unsignalized Intersection Using Micro-
Simulation
Deepak V. Killi and P. Vedagiri
Department of Civil Engineering, Indian Institute of Technology Bombay, Mumbai, India
Email: [email protected], [email protected]


Abstract—Traffic safety is an important area in in addition to creating social awareness among the driver
transportation engineering. Presently, most traffic safety community.
assessment and prediction related work is based on the use Studies in general show that speed enforcement led to
of historical accident data that has known drawbacks significant reduction in frequency of accidents [2].
related to the quality and coverage of data especially in
Several studies solely evaluated traffic safety with respect
developing countries like India. For the assessment of
roadway solutions in the future, it is impractical and
to speed [3], [4]. Experience shows that microscopic
unethical to wait for accidents to occur before being able to simulation is able to predict to a certain extent, the risks
draw statistically sound conclusions regarding safety impact. within a traffic flow [5]. Microscopic simulation helps to
Hence there is a need to develop proactive models based on evaluate traffic conditions without having to realize tests
Surrogate Safety Measures (SSMs) for a more effective in the field. Using these techniques it is very difficult to
safety evaluation. The main advantage associated with the evaluate accidents and find the cause for it because of the
use of safety indicators is that they occur considerably more varying trajectories in the real and virtual simulated cases
frequently than accidents, thereby implying an efficient and as recognized by Darzentas et al. [6].
more statistically reliable proximal measure of traffic safety. This lead to a new strategy to approach this problem
The objective of this study is to assess the level of traffic
safety at an uncontrolled intersection using micro-
using Surrogate safety measures (SSMs) [7]. They
simulation modeling under mixed traffic condition by involve usage of proximal safety indicators that represent
devising a unique strategy of measuring proximal safety the temporal and spatial proximity characteristics of
indicator, Post Encroachment Time (PET). unsafe interactions and near-crashes. The major
advantage of such measures is that they occur more
Index Terms—surrogate safety measures, micro-simulation, frequently than crashes and require short observation
uncontrolled intersections, post encroachment time, periods in order to establish statistically reliable results.
validation As Archer mentioned, these measures include Time-to-
collision (TTC), Time extended TTC (TET), Time
integrated TTC (TIT), Post encroachment time (PET),
I. INTRODUCTION Deceleration rate (DR) etc. [8]-[10]. The most commonly
Safety on roads is a major concern in the developing used are TTC and PET, which were identified from the
world because of its impact on the global economy and study of Gettman and Head [10].
people’s welfare. Due to the rising population, the traffic Pirdavani et al. presented a safety evaluation of
risk has increased especially in developing nations like unsignalized intersections using micro simulation and
India, as the infrastructure is unable to cope up with the proximal safety indicators. They have applied a micro
increasing traffic. In India, 17 lives were lost every hour simulator (S-Paramics) to investigate whether changing
in 2011 and its increase to 21 lives per hour in 2012 speed limits under different traffic volume conditions will
shows the necessity of a reliable strategy to prevent affect traffic safety as measured by PET. The results of
accidents [1]. the simulation show that increasing the traffic volume on
Intersections pose special safety concerns because of both major and minor roadways will lead to a decrease of
unsafe driver actions and maneuvers. The conflicts at mean PET values [2].
intersections include merging, crossing and diverging Klunder et al. developed a new micro-simulation
conflicts apart from the abrupt changes in vehicle speeds model for intersection traffic, which can generate
and unexpected lane changes. As the above data amply accurate SSMs and which represents driver behavior on
show drivers as the major cause for accidents, it is intersections more accurately. The results show that as
essential to further study this from the perspective of the critical gap value decreases, increasingly unsafe
infrastructure development to enable drivers take situations can occur at the intersection [11].
precautionary measures to avoid accidents at intersections Athough these attempts were successful in predicting
the future occurrences, the accuracy has always been
Manuscript received January 10, 2014; revised March 31, 2014. doubtful because of varying trajectories of different

©2014 Engineering and Technology Publishing 140

doi: 10.12720/jtle.2.2.140-145
 Journal of Traffic and Logistics Engineering Vol. 2, No. 2, June 2014

vehicles. The objective of the study is to improve the an instant ‘t’, but using the rhombus geometry, one can
accuracy of findings and also assess the traffic safety identify and differentiate relatively safe cases and
under mixed traffic condition. In the current analysis, the relatively unsafe cases. In the following figures, it can
safety measure Post Encroachment Time (PET) is used as clearly identified that first case will have a negative PET
it is a reliable indicator for both rear-end and transverse at that instant ‘t’ as first vehicle is yet to leave and the
collisions which occur frequently at intersections. This second case will have a positive PET as first vehicle left
measure is used to identify cases in which two road-users already, hence gives accurate predictions.
pass over a common spatial point or area with a temporal
difference. PET is the time between moment t1 when the
first vehicle exits the conflict spot and moment t2 when
the second vehicle enters the conflict spot, defined by
t2−t1. A smaller PET value implies a greater probability
of collision. Micro-simulation modeling software
VISSIM 6.0 has been used to evaluate the safety at Figure 1. Relatively unsafe conflict with greater crash possibility.
unsignalized intersections. For the first time in PET
studies, a concept of negative PET has been introduced
which is detailed in the study methodology section. This
will be followed by sections on PET from field and PET
from simulations and ended with findings from
simulations and conclusions.

II. STUDY METHODOLOGY Figure 2. Relatively safe conflict with lesser crash possibility.
It is known that usage of statistical models based on
historical crash data are most common in traffic Furthermore, the trajectory of a turning vehicle is
engineering today. The approach used is considered never straight and always at an angle that ensures
‘reactive’ in nature rather than ‘proactive’, where a vehicles occupy less number of grids and maximum
significant number of crashes must occur before a portion of grids in current case. This will further improve
problem is identified and suitable corrective measures can the accuracy of the predictions. Similar method is applied
be implemented. Understanding these problems, in simulation using the concept of nodes by detecting the
researchers have recently proposed a framework for time differences when a vehicle enters the defined grids.
‘proactive’ safety planning, i.e. planning that is not
entirely based on historical crash data, but uses other III. POST ENCROACHMENT TIME FROM FIELD DATA
measures such as the use of safety indicators and A video based traffic survey was conducted at Central
predictive models [8]. Also, the use of safety indicators is Avenue-Technology St. intersection in Hiranandani
also a more resource-efficient and ethically appealing Gardens Powai, Mumbai, India for a duration of one hour,
alternative for fast, reliable and effective safety on February 19th 2014 from 1700 hours to 1800 hours
assessment. which is also the peak hour period at that intersection.
The method used in this study is to first calculate PET The roadway characteristics observed from the survey is
values from the field, then simulate and model in micro- that, it is an uncontrolled T intersection with three lanes
simulation software VISSIM 6.0 and use simulation to on major road in each direction and two lanes on minor
identify and predict interdependencies of the vehicle road in each direction. The data set extracted from the
characteristics and evaluate traffic safety at uncontrolled video included traffic volume, vehicle composition along
or unsignalized intersections. each of the directions apart from PET extraction.
Usually PET is calculated on the basis of grids on The grids have been created manually using Autodesk
lanes, assuming vehicles move in a single lane in a Maya 3D animation software. These grids were then
straight path, which is not synonymous to the actual case converted into a transparent picture and then overlaid into
especially in developing nations. Also, traditional method the video using Corel VideoStudio Pro X6 software. The
involves using square or rectangular grids of large grids were numbered lane wise to ease the calculation of
dimensions on the conflict area at an intersection that PET values. After this, the video is run at frame rate of 6
may not give reliable conclusions from the study. frames per second. The least count of this is 0.01 seconds,
Current study involves the concept of area occupancy which gives more accurate results. While running frame
on the grids, which have an advantage of identifying the by frame, when a possible conflict occurs, time t1 i.e.
grid in which there is highest possibility of collision and when a rear end of one vehicle leaves a grid and time t 2
also calculate PET values more accurately. A novel i.e. when a front end of another vehicle touches the same
geometry of the grid is used for the study, using a grid are noted, and PET values are calculated as t2-t1. Fig.
rhombus with dimensions as 2.5m x 2.5m and internal 3 is the snapshot of the overlaid video including the grid
angles of 600 and 1200. The advantages of using this numbers. From the data extracted manually from video,
geometry compared to the traditional method have been the grids 3.4, 3.5, 4.4, 4.5, 5.4, 5.5 had the maximum
showed in Fig. 1 and Fig. 2. In both the cases, traditional number of conflicts. In total 944 vehicle-vehicle conflicts
calculation method gives the same value of zero PET at have been identified from a total of 3204 vehicles at the

©2014 Engineering and Technology Publishing 141

 Journal of Traffic and Logistics Engineering Vol. 2, No. 2, June 2014

intersection. The mean PET value obtained was 4.04 when the average speed of all the vehicles was reduced
seconds. The frequency distribution of the calculated PET by 5 kmph to 10 kmph depending on the vehicle type.
values is tabulated in Table I. The obtained results will be The vehicle dimensions are also adjusted according to the
compared to the simulated values in the simulation Indian traffic with the inclusion of a three-wheeler called
findings section. Auto-rickshaw and a Light commercial vehicle (LCV).
Priority zones are given in accordance to the real
movement of traffic from the field data collected. A total
of 6 vehicle types are considered under study, which
conforms to the requirement of mixed traffic condition.
B. Validation
After calibrating the model and running the simulation
a couple of times, driver behavior characteristics were
changed in order to get the simulated values comparable
to the field values. The shapes of connectors are also
adjusted for better outputs. The final model was then re-
validated with secondary data from another one-hour
period and the results obtained were satisfactory. Hence,
the simulation model was calibrated and validated.
Figure 3. Screenshot of the video with grids.
C. Calculation of PET
Polygon nodes of rhombus shape have been placed
TABLE I. FREQUENCY DISTRIBUTION OF FIELD PET exactly at the same positions as in the video based
Lower Upper Percentage Cumulative
method. Fig. 4 is the screenshot of the intersection during
Mid Conflicts simulation.
Class Class Relative Percentage
point Frequency
Limits Limits Frequency Frequency
-5 0 -2.5 196 20.76 20.76

0 5 2.5 488 51.69 72.45

5 10 7.5 136 14.41 86.86

10 15 12.5 80 8.47 95.33

15 20 17.5 22 2.33 97.66

20 25 22.5 12 1.27 98.93

25 30 27.5 6 0.64 99.57

30 35 32.5 4 0.42 100 Figure 4. Screenshot of the modeled intersection.

IV. POST ENCROACHMENT TIME FROM SIMULATION

TABLE II. FREQUENCY DISTRIBUTION OF BASECASE SIMULATED PET
As mentioned earlier, micro-simulation modeling Lower Upper Percentage Cumulative
software VISSIM 6.0 has been used. In developing the Mid Conflicts
Class Class Relative Percentage
point Frequency
model of the existing situation in VISSIM, the first step Limits Limits Frequency Frequency
was to create a road network with the background of the -5 0 -2.5 221 23.31 23.31
intersection extracted from Google maps. The dimensions
of the road and channelization have been obtained using 0 5 2.5 432 45.57 68.88
wheel meters and the same have been applied in the 5 10 7.5 160 16.88 85.76
simulation model. VISSIM has inbuilt features that can
10 15 12.5 91 9.6 95.36
be partially considered as assumptions of the simulation
model for real time data. It follows Poisson’s distribution 15 20 17.5 28 2.95 98.31
for the arrival process with exponential inter arrival times.
20 25 22.5 12 1.27 99.58
It uses a psychophysical car following model and a rule-
based algorithm for lateral movements. 25 30 27.5 4 0.42 100

A. Calibration 30 35 32.5 0 0 100

After building the road network, the model is
calibrated by assigning vehicle inputs from all the three The nodes are numbered in the same manner as the
directions from the values obtained in field data. Speed case of field data and raw data from the nodes is collected
ranges are assigned to vehicles by conducting a spot in a Microsoft Excel sheet. The raw data contains the
speed survey of 15 vehicles in each vehicle type. details of time when a vehicle enters the node, time when
Reduced speed areas have been used at the intersections a vehicle leaves the node, type of vehicle and path of

©2014 Engineering and Technology Publishing 142

 Journal of Traffic and Logistics Engineering Vol. 2, No. 2, June 2014

vehicle. From this PET values are calculated using Excel or a new facility can be created as an alternative means
functions and this can also be generalized using a for some incoming traffic.
computer program to save some time for data analysis. Keeping all the other characteristics constant, volumes
For the existing case, a total of 948 vehicle-vehicle on major roads were increased by 10%, 20% and also
conflicts have been identified from a total of 3204 decreased by 10%, 20%. One can easily predict that total
vehicles at the intersection. The mean PET value obtained number of conflicts would obviously increase when
was 4.11 seconds. The frequency distribution of the volume is increased and total conflicts would decrease
calculated PET values is tabulated in Table II. Similar when volume is decreased.
procedure was followed with some changes in inputs to The total number of conflicts was 1147, 1267 for 10%
test their influence on PET values, which is discussed in and 20% increase respectively. The conflicts were 837
the upcoming findings from simulation section. and 726 for 10% and 20% decrease respectively. To do a
comparative analysis, relative frequency curve is plotted
V. FINDINGS FROM SIMULATION as in Fig. 7 and it was clearly observed that a huge
proportion of about 88% of conflicts had PET values less
A. Comparision with Field Results than 5 seconds in case of 20% increase in volume and
After obtaining both the field and simulated PET about 53% of conflicts in case of 20% volume decrease
values, they were compared with relative frequency and compared to the existing 69%. The results are
cumulative frequency curves. The graphs are plotted as comparable with the results from previous studies but
shown in Fig. 5 and Fig. 6 respectively. From the figures, with more reliable and foolproof technique [5].
it is obtained that field and simulated values are almost
same and model is accurate. This is also an indicator that
the model has been calibrated and validated. The next
step was to assess these values with varying vehicle
characteristics as discussed in the following subsections.

Figure 7. Frequency curves for change in Major road volume.

C. Effect of Change in Vehicle Composition

The second part of assessment involved changing the
vehicle compositions on roads along all the directions.
Figure 5. Relative frequency curves for field and simulated PET This is a crucial step of evaluation as the entry of heavy
vehicles can be restricted in cities up to an extent, at least
during the peak hours on the roads. The analysis was
done changing the total number of heavy vehicles and
commercial vehicles i.e. buses, LCVs and HCVs which
form a major source of obstruction and also increase the
collision probability.
Keeping all the other characteristics constant, heavy
vehicle compositions were increased by 10%, 20% and
also decreased by 10%, 20%. In this case, surprisingly,
the total number of conflicts was not much higher
compared to the previous case with only change in
volumes on major roads. This is possibly because speeds
Figure 6. Cumulative frequency curves for field and simulated PET.
vary for different type of vehicles and this being a mixed
traffic condition with 6 different vehicle types.
The total number of conflicts was 1169, 1278 for 10%
B. Effect of Change in Major Road Volume and 20% increase respectively. The conflicts were 822
The first part of assessment involved changing the and 652 for 10% and 20% decrease respectively. To do a
volumes of vehicles on the major roads in both the comparative analysis, relative frequency curve is plotted
directions with respect to the existing major road volumes. as in Fig. 8 and it was clearly observed that a huge
With about 10% increase every year in total number of proportion of about 93% of conflicts had PET values less
vehicles, it is essential to evaluate the facility by than 5 seconds in case of 20% increase in heavy vehicle
changing the volumes to future values. It is also essential composition and about 43% of conflicts in case of 20%
to see for any reductions in case of lesser volumes as the heavy vehicle composition decrease compared to the
incoming traffic can be diverted through some other route existing 69%.

©2014 Engineering and Technology Publishing 143

 Journal of Traffic and Logistics Engineering Vol. 2, No. 2, June 2014

This means that although, in case of 20% increase values apart from the fact that overall conflicts would
there is not much difference in the number of conflicts, in increase with increase in vehicle speeds which increases
case of excess heavy vehicles a greater proportion of the total accident probability, but the relative risk in each
conflicts have PET values less than 5 seconds, which is of the case is almost same with similar pattern of
unsafe. But, in case of lesser heavy vehicles which is frequency distribution for different PET values.
possible by restricting heavy vehicle movement along The intersection selected in the study was a typical
such routes by only allowing public transport, less than uncontrolled intersection inside city premises with
half of the conflicts came under 5 seconds PET value. limited speeds. This can be one of the reasons for
observing a similar pattern for all the speeds. But, if the
overall speeds increase, which occurs usually when
volume decreases, there is a possibility of observing a
varied pattern of relative frequency distribution.

Figure 8. Frequency curves for change in heavy vehicle composition.

D. Effect of Change in Vehicle Speeds

The third part of assessment involved changing the Figure 9. Frequency curves for change vehicle speeds.
speeds of vehicles with respect to the speeds along the
existing facility. This part of study evaluates and
identifies the safe speed at the intersections that will be VI. CONCLUSION
different from the design speed of the road. The results of This paper presented a proactive safety evaluation of
this part of study can be used to have restrictions on an unsignalized intersection using proximal safety
vehicle speeds in order to provide a safe road for indicator Post Encroachment Time (PET) and micro-
commuters by preventing accidents. The analysis was simulation modeling. This new method of calculating
done by changing the speeds of vehicles in all the PET can provide useful comparisons for evaluating safety
directions irrespective of the vehicle type. by studying the effect of roadway and traffic
Keeping all the other characteristics constant, speeds characteristics under mixed traffic condition.
were increased by 5 kmph (kilometers per hour), 10 A few conclusions specific to the intersection under
kmph and also decreased by 5 kmph, 10 kmph. In this study can be drawn as follows:
case, the total number of conflicts was much lesser  The PET values obtained are very small, with 25-
compared to both the previous cases where there was 40% values in the negative range and 40-50%
change in volumes on major roads and change in vehicle values in the range of less than 5 seconds.
compositions. This is possible because of many factors  It can be inferred that the intersection is highly
such as: unsafe and there is a need of precautionary
 Vehicles might follow a completely different measures such as using signals, decreasing the
trajectory in order to maintain their speed and overall vehicle volumes etc.
avoid stoppages.  A couple of minor transverse end crashes were
 The duration of vehicles on the conflict spots is observed during the short study period, which
less than before in case of faster movement. supports this conclusion.
 The vehicles may start queuing in case of slower The practical merits of micro-simulation were
movement and hence PET values would demonstrated with varying conditions of speeds, volumes
automatically increase. and compositions. The application shows how the
The total number of conflicts was 1020, 1083 for 5 changes in certain parameters effect the safety situation,
kmph and 10 kmph increase respectively. The conflicts as measured by the Post Encroachment Time (PET). The
were 881 and 750 for 5 kmph and 10 kmph decrease promising results found in this new study methodology
respectively. To do a comparative analysis, relative pointed out an opportunity of expanding the research to a
frequency curve is plotted as in Fig. 9 and it was clearly more complex, comprehensive and extensive traffic
observed that a huge proportion of about 76% of conflicts safety evaluation with different combination of
had PET values less than 5 seconds in case of 10 kmph assessment apart from traffic policy measures.
increase in vehicle speeds and about 62% of conflicts in From the three phases of assessment, change in the
case of 10 kmph decrease in vehicle speeds compared to traffic volumes resulted in a proportional change in
the existing 69%. relative frequencies whereas change in heavy vehicle
From the observed values, we can clearly notice that compositions had a higher impact on relative frequencies.
there is not much dependence of vehicle speeds on PET On, the contrary, change in speeds did not have great

©2014 Engineering and Technology Publishing 144

 Journal of Traffic and Logistics Engineering Vol. 2, No. 2, June 2014

effect on these values. Hence, the best and first [7] P. Songchitruksa and A. P. Tarko, “The extreme value theory
approach to safety estimation,” Accident Analysis and prevention,
precautionary method is decreasing the number of heavy
vol. 38, no. 4, pp. 811-822, 2006.
vehicles on the road during peak hours. [8] J. Archer, “Indicators for traffic safety assessment and prediction
However, it is necessary to point out that this study is and their application in micro-simulation modeling: A study of
limited to the use of PET as a proximal safety indicator, urban and suburban intersections,” Ph.D. dissertation, Dept. of
Infrastructure, Royal Institute of Technology, Stockholm, Sweden,
which is predominantly useful to investigate transverse
2005.
collisions. The use of other safety indicators, like TTC [9] N. El Faouzi, et al., Road Safety Indicators: Results in Vaud
and its derived sub-indicators are more useful in rear-end Canton, Rome, Italy, In RSS 2007.
collisions. Nevertheless, since most of the collisions at [10] D. Gettman and L. Head, “Surrogate safety measures from traffic
simulation models,” Transportation Research Record: Journal of
unsignalized intersections are transverse collisions, the
the Transportation Research Board, vol. 1840, pp. 104-115, 2003.
use of PET for this type of intersection is appropriate. [11] G. Gerdien Klunder, A. Arshad, and B. Immers, “Development of
The outcome of our study and its future scope a micro-simulation model to predict road traffic safety on
promises an accurate and reliable method to identify intersections with surrogate safety measures,” TNO Build
Environment and Geo Sciences, 2004, pp. 1-25.
unsafe intersections and prevent accidents by applying
necessary precautionary methods.
Deepak V. Killi was born in Visakhapatnam, India, in 1993. He is
ACKNOWLEDGMENT currently pursuing his Bachelors in Technology from IIT Bombay, India.
He is doing his majors in civil engineering and minors in electrical
The authors wish to thank Dr. Hiranandani, Managing engineering.
He was a winner in Mathematics Olympiads at both national and
Director and Mr. Joseph Reddy, General Manager of international levels in 2008 and 2010 respectively. He has gained
administration and facilities of Hiranandani Group of operational experience in Oil field and consumer goods industries
Companies, Mumbai, India for providing an opportunity during his summer interns at renowned firms Schlumberger and Coca-
to collect data for our study. This work was supported in cola. His current research is in the field of transportation and traffic
assessment. His past research projects include ‘Real time traffic
part by resources and grant from Indian Institute of monitoring using loop inductors’ and ‘Improvement of major arterials
Technology Bombay, India. using micro-simulation modeling.’
Mr. Deepak Killi is a recipient of prestigious Kishore Vaigyanik
REFERENCES Protsahan Yojana in 2009, the highest form of fellowship awarded by
the Government of India for students with aptitude in research. He was
[1] Road Accidents in India, Ministry of Road Transport and also awarded Institute Organizational Color by IIT Bombay for his
Highways, GOI, 2011-12. exemplary contribution to organizational activities.
[2] A. Pirdavani, T. Brijs, T. Bellemans, and G. Wets, “Evaluation of
traffic safety at un-signalized intersections using microsimulation:
A utilization of proximal safety indicators,” Advances in Dr. P. Vedagiri was born in Chennai, India, in 1980. He is currently an
Transportation Studies an international Journal, Section A 22, pp. Assistant Professor in the Department of Civil engineering at IIT
43-50, 2010. Bombay, India. He did his Doctorate from IIT Madras, India in 2009
[3] R. Elvik, “Effects on accidents of automatic speed enforcement in and was also a Project officer and Research associate during 2004-09.
norway,” Transportation Research Record: Journal of the His research areas include Traffic flow modeling and simulation, Traffic
Transportation Research Board, vol. 1959, pp. 14-19, 1997. monitoring and public transit system design. He is well known for his
[4] S. Hess, “Analysis of the effects of speed limit enforcement co-authored book ‘Planning for Priority for Public Transport: Providing
cameras: Differentiation by road type and catchment area,” Exclusive Bus Lanes on Roads Carrying Heterogeneous Traffic’
Transportation Research Record: Journal of the Transportation published in 2012. He has more than 50 journal and conference
Research Board, vol. 1865, pp. 28-34, 2004. publications at both national and international level.
[5] P. Vedagiri and T. Pragna, “Evaluation of traffic safety at un- Dr. Vedagiri is a recipient of many honors and fellowships, notably the
signalized intersections under mixed traffic conditions,” presented World Conference in Transportation Research society fellowship and
at the Transportation Research Board 92nd annual meeting, Young faculty award from IIT Bombay. He is an active member in
Washington, D.C, Jan 13-17, 2013. professional bodies such as Transportation Research Board, USA and
[6] J. Darzentas, et al., “Simulation of road traffic conflicts at T- Indian Roads Congress etc. He is a recognized road safety auditor
junctions,” Simulation, vol. 34, no. 5, pp. 155-164, 1980. certified by IRF, India and ARRB, Australia.

©2014 Engineering and Technology Publishing 145