Microscopic traffic simulators have become efficient tools to conduct different analytic studies on roads, vehicles, behavior of drivers, and critical intersections, which lead towards a well-planned traffic solution. Devising a realistic and sustainable traffic solution requires replication of the real traffic scenario in a simulator. For example, to simulate the traffic streams of developing and under developed countries, we need to simulate non-lane based heterogeneous traffic stream, i.e., motorized and non-motorized vehicles, road traffic behaviors such as irregular pedestrian, illegal parking, violation of laws pertaining lanes, etc. However, most of the existing traffic simulators are unable to mimic the unstructured road traffic streams of less developed countries with their diversified behaviors. Therefore, in this work, we propose a new microscopic traffic simulator to handle nonlane based heterogeneous traffic stream and on road traffic behaviors that generally occurred in the road networks of cities in less developed countries. Our simulator receives network topology, traffic routes, and traffic demand flow rates as input, visualizes the traffic flows, and provides traffic statistics. To evaluate sustainability of our proposed simulator in real-life scenarios, we calibrate the simulator using real traffic data. Our evaluation reveals 99% accuracy in terms of travel time.
Towards Simulating Non-lane Based Heterogeneous Road Traffic of Less Developed Countries
1. Towards Simulating Non-lane Based
Heterogeneous Road Traffic of Less
Developed Countries
Department of Computer Science and Engineering,
Bangladesh University of Engineering and Technology, Dhaka, Bangladesh
ICT4S 2018
Toronto, Canada
Quazi Mishkatul Alam, Bejon Sarker, Biplob Biswas, Kazi Hasan Zubaer,
Tarik Reza Toha, Novia Nurain, and A. B. M. Alim Al Islam
2. Outline
• Background and motivation
• Our proposed simulator: DhakaSim
– Modeling a road network
– Modeling motorized and non-motorized vehicles
– Modeling pedestrians and non-ideal phenomena
• Experimental evaluation
– Virtual testing
– Real testing
• Conclusion and future work
2
3. Microscopic Traffic Simulation
• The most sustainable way to predict traffic behavior of
any existing or future development on the road network
– Applied for structured traffic systems of advanced countries
• Less-developed countries such as Bangladesh, India, etc.,
often experience unstructured traffic systems
– Motorized vehicles: buses, cars, auto-rickshaws, trucks, etc.
– Non-motorized (human-powered) vehicles: rickshaws, vans,
bicycles, etc.
– Non-ideal phenomena
• Irregular pedestrians on road
• Unwise parking on road
• Mobile market on road
• Intentional traffic rule violation 3
4. Real Traffic Scenario in Dhaka
4
Non-lane
based traffic
Non-motorized
vehicle
On road
pedestrians
Unstructured
traffic stream
5. Existing Commercial Traffic Simulators:
AIMSUN
5Ferrer et al., 1993
Lane-based and
homogeneous
trafficStructured
traffic stream
6. Existing Commercial Traffic Simulators:
VISSIM
6
Fellendorf et al., 1994
Lane-based and
heterogeneous
traffic
Structured
traffic stream
7. Existing Commercial Traffic Simulators:
MITSIM
7Yang et al., 1996
Lane-based and
homogeneous
traffic
Structured
traffic stream
8. Existing Open-source Traffic Simulators:
SUMO
8Behrisch et al., 2011
The architectural difference
resists to extend SUMO to
support unstructured traffic
stream
9. Our Contribution
9
We propose a microscopic traffic simulator,
DhakaSim, to simulate non-lane based heterogeneous
traffic stream and on road traffic behaviors that
generally occur in the road networks of cities in less-
developed countries such as Dhaka city
DhakaSim
10. Block Diagram of DhakaSim
10
Modules of our proposed microscopic traffic simulator
12. Modeling A Road Network
• Environment definition
– Junctions
• Intersections of the roads
• Nodes of the road
network graph
– Links
• Actual roads or corridors
– Segmentation to
represent curved road
– Lateral thin strips to
create virtual lanes
• Edges of the road
network graph 12
Junction file Link file
Strip arrangement of a curved road (link)
13. Modeling A Road Network (contd.)
13
• Environment definition
– Traffic route
• List of links between source
and sink nodes
– Traffic demand
• Traffic flow rate (vehicles
per hour) of a route
– Simulation parameters
• Encounter per accident
– Probability percentage of
being an accident
• Maximum speed of the
vehicles
• Pedestrian mode, etc.
Route file Demand file
Parameters file
14. Modeling Pedestrians and Vehicles
14
Visual representation of our simulator (colored rectangles are
vehicles and green dots are pedestrians)
On road
pedestrians
Non-lane
based traffic
16. Traffic Statistics Generation (contd.)
16
Total number of accidents over all links
Simulator log file describing the state of
pedestrians and vehicles for each
simulation step
17. Experimental Evaluation: Virtual Testing
17
Sample map of a road network used in analyzing link effects by
removing the red colored link
Parameters Value
Demand
flow rate
100
vph
Max.
allowable
speed
10 kph
Pedestrian
mode
On
Accident
probability
0.01%
18. Experimental Evaluation: Virtual Testing
(contd.)
18
Total number of vehicles over all links for
original map and modified map
Average speed of vehicles over all links
for original map and modified map
19. Experimental Evaluation: Real Testing
19
Real GIS road map of Sankar to
Palashi intersections (red colored)
Simulated road map of Sankar to
Palashi intersections
20. Experimental Evaluation: Real Testing (contd.)
20
Parameters Value
Demand flow rate 124 vph
Max. allowable speed
17.3
kph
Pedestrian mode On
Accident probability 0.01%
Travel time from Sankar to Palashi intersections
measuring in simulator and real-world
22. Conclusion and Future Work
• Microscopic traffic simulation has utmost
significance towards designing an effective and
sustainable transportation system
– No such simulator exists that models unstructured traffic
stream of less-developed countries
• DhakaSim incorporates non-ideal road traffic
behavior of less-developed countries
• In future, ee plan to simulate autonomous vehicles to
analyze its sustainability in the less-developed
countries by observing their on road behavior
22
Hello everyone! Welcome to my presentation. Now I am going to present the paper titled as “Towards Simulating Non-lane Based Heterogenous Road Traffic of Less Developed Countries”. The authors are Quazi Mishkatul Alam, Bejon Sarker, Biplob Biswas, Kazi Hasan Zubaer, Tarik Reza Toha, Novia Nurain, and A. B. M. Alim Al Islam. All are from Dept. of CSE, BUET. This work has been funded by ICT division of Bangladesh. Let’s see the outline of my presentation.
At first, I will talk about the background and motivation behind this simulator development. Then, I will talk about the modeling of simulator environment followed by experimental evaluation. Finally, I will conclude with some future work. Let’s start.
To predict traffic behavior of any existing or future development on the road network, one of the most effective and sustainable ways is microscopic traffic simulation. We have to provide real traffic scenario in the simulation environment to get dependable predictions. For example, less developed countries such as Bangladesh, India, etc. have no traditional structured traffic stream that is found in developed countries such as US, Canada, etc. rather an unstructured traffic stream. Here, motorized vehicles such as buses, cars, trucks, etc. and human powered vehicles such as rickshaws, vans, etc. share same road way. Besides, intentional traffic rule violation, irregular pedestrians, unwise parking, mobile market on road, etc. are very common in these countries. Let’s see a real picture of the roads.
This picture has been taken from a road of Dhaka city, capital of Bangladesh. We can see that non-lane based traffic, non-motorized vehicles, on road pedestrians are very common in Dhaka city. Hence, the traffic stream is non-lane based and heterogenous, in short unstructured. Now, we will see some existing popular traffic simulators.
AIMSUN is the first microscopic commercial traffic simulator. We can see the it simulates lane based and homogenous traffic stream, which means it can simulate only structured traffic stream. Let’s see another simulator.
VISSIM is the latest commercial micro simulator. It can simulate heterogeneous traffic, i.e., motorized and non-motorized. However, the traffic is lane-based. Therefore, we cannot simulate the traffic of Dhaka. Now, we will see some open-source simulators.
MITSIM is another commercial simulator, which also simulates only lane based and homogenous traffic that means structured stream. Now, we will see the latest traffic simulator.
SUMO is an open-source simulator, which supports structured traffic stream. However, we cannot extend it to support unstructured traffic stream because the fundamental traffic characteristics of SUMO is lane-based and homogenous. Therefore, we propose a new traffic simulator for less developed countries.
DhakaSim simulates non-lane based heterogenous traffic, i.e., unstructured traffic stream and non-ideal behavior of traffic on road, which is very common in the road networks of cities in less developed countries such as Dhaka. Let’s see the block diagram of our simulator.
DhakaSim has three modules, i.e., core simulator, signaling, and surveillance module. We define the simulation environment, i.e., components of road network such as junctions and links, traffic flow rate and routes and feed them to the core simulator. The core part creates the environment and generates heterogeneous traffic including pedestrians, motorized vehicles, and human-powered vehicles. Then, it simulates the traffic using their real-world behavior.
The surveillance module collects data of current traffic from different parts of the road network and forwards it to signaling module, which imposes traffic signals to thesimulator. When the simulation time is over, surveillance module generates different statistics on simulated road network such as average number of vehicles, average speed of vehicles, accident count, etc. for each link.
Now, we will see the working process of DhakaSim.
Like a normal traffic simulator, we load the road network and generate pedestrians and vehicles and remove them when they get reached their destination. The main difference is we generate the pedestrians from random side of any road and they may have accidents while crossing the road. Besides, we generate the vehicles of random sizes and random speeds to model the heterogenous traffic and they can move along any place of the road to model no-lane based traffic.
After simulation time over, relevant statistics are generated. Let’s see the detailed part of our simulator.
We model the road network as a graph data structure. Here, nodes are the intersections of the roads and the edges are the corridors, i.e., actual roads. We describe them using these text files. We model the curved road by segmentation and model the virtual lane by using lateral thin strip as shown in the figure.
Traffic routes and flow rate are provided by these route and demand files. To customize the simulation environment, we use some simulation parameters such as simulation speed, end time, pixel per strip, etc. For modeling pedestrian behavior, we use a value for encounter per accident, i.e., probability percentage of being an accident. Let’s see the visual representation of DhakaSim.
This is the visual representation of DhakaSim. Here, the colored rectangles are vehicles and green dots are pedestrians and the vehicles and pedestrians do not maintain any structure. Let’s see the traffic statistics.
After finishing simulation, some traffic statistics will be generated. The first graph represents the average vehicle speed in km per hour over all roads and the second one represents the total number of vehicles over all roads.
The third one represents the total number of accidents over all roads. To replay the simulation, we maintain a log file that describe the state of pedestrians and vehicles for each simulation step. Here, object means pedestrians. We describe them with their coordinates. Let’s see the experimental evaluation.
First, we evaluate our simulator in virtual world. We provide a sample map of a road network to analyze link effects by removing red colored (10) link. The parameters are listed in the table. Let’s see the results.
Due to the removal of link 10, the number of vehicles get increased in link 9, 14, and 16. Besides, the number of vehicles in link 15 is zero. Therefore, average speed of vehicles is also zero. Now, we will see a real-world testing of DhakaSim.
For real-world evaluation, we select the red colored road network in GIS map. The simulated map is shown in the second figure. We collect real data of the route by travelling in various vehicles such as buses, legunas, CNGs, etc. Then, we compare it with simulated statistics.
We can see that the simulate travel times are very close to real travel time (99% accuracy). In real data, we took 37 data points. The simulation parameters are listed in the table. These parameters are collected from a paper that presents the number of vehicles and average speed of the vehicles of the route.
To design a sustainable transportation system, microscopic traffic simulator is an inevitable tool. Existing simulators fail to simulate the unstructured traffic of less developed countries. DhakaSim models non-ideal road traffic behavior of these countries. We plan to simulate autonomous vehicle to analyze its sustainability in these countries by observing its on road behavior. Thank you.