A Python Extension in Sumo for Simulating Traffic Incidents and Emergency Service Vehicles

Traffic signal retiming and coordination play a significant role in traffic management, especially when the traffic incidents disrupt the network. Many researchers have developed artificially intelligent (AI) traffic signal controllers based on goal-oriented machine learning frameworks to try and optimize network performance. However, previous efforts have lacked the means to evaluate these networks under traffic incident conditions. To make these AI traffic signal controllers more robust, current research needs to consider AI traffic controller performance under traffic incidents and accompanying emergency response. Obtaining field incident data and converting into inputs for simulation models to evaluate these machine learning models has been a huge hurdle because it is expensive, time-consuming, and sometimes even unfeasible. This paper provides an integrated traffic incident and response simulation tool for a grid network made in Simulation of Urban MObility (SUMO) to overcome this gap. The tool includes random traffic incident generation (location and duration), incident detection, random emergency vehicle generation, and emergency vehicle dispatching. With this tool, users can simulate a road network with traffic incidents and emergency vehicle response to produce substantial amounts of data for training robust reinforcement learning models. In addition, this tool will save future researchers and practitioners both time and effort in testing the impact of their proposed AI traffic control models and allow for a more complete evaluation of performance.

This work was sponsored by the USDOT CAMMSE University Transportation Center, University of North Carolina, Charlotte.

Authors and Affiliations

1. Connecticut Transportation Institute, University of Connecticut, Storrs, CT, USA

   Tianxin Li

2. Cockrell School of Engineering, Department of Civil, Architectural and Environmental Engineering, University of Texas, Austin, USA

   Tianxin Li, Carolina Baumanis, Jennifer Hall & Randy Machemehl

3. Department of Biological System Engineering, University of Wisconsin-Madison, Wisconsin-Madison, USA

   Wei Zhao

Corresponding author

Correspondence to Tianxin Li .

Editors and Affiliations

1. Department of Civil Engineering, University of Victoria, Victoria, BC, Canada

   Rishi Gupta

2. Department of Civil Engineering, University of Victoria, Victoria, BC, Canada

   Min Sun

3. Department of Earthquake Engineering, The University of British Columbia, Vancouver, BC, Canada

   Svetlana Brzev

4. The University of British Columbia, Okanagan Campus, Kelowna, BC, Canada

   M. Shahria Alam

5. University of Regina, Regina, SK, Canada

   Kelvin Tsun Wai Ng

6. Environmental Engineering Program, University of Northern British Columbia, Prince George, BC, Canada

   Jianbing Li

7. The University of Western Ontario, London, ON, Canada

   Ashraf El Damatty

8. Department of Civil Engineering, University of British Columbia, Vancouver, BC, Canada

   Clark Lim

Appendix

1.1 Appendix 1

Below command was run for generating the 4 × 4 grid network illustrated in this paper.

netgenerate --grid --grid.number=4 --grid.length=200 --default.lanenumber=2 --default.speed=20 --no-turnarounds=true --turn-lanes=1 --turn-lanes.length=100 --default-junction-type=traffic_light --grid.attach-length=200 --tls.yellow.time=3 --tls.left-green.time=12 --tls.allred.time = 2 --output-file=net.net.xml

Network generating parameters and their meanings:

--grid: grid network will be generated. SUMO also provides other types of the network to be generated automatically, including spider and random networks.

--grid.length defines the length of each intersection leg in meters

--default.lanenumber defines the number of lanes for each approach

--default.speed defines the edge design speed in meters/second

--no-turnarounds defines whether to allow turn around for the left turn lane

--turn-lanes defines the number of left turn lanes

--turn-lanes.length defines the length of left turn lanes

--default-junction-type defines the intersections in the network are controlled by the pretimed traffic signals

--grid.attach-length defines the length of road attached to the fringe of intersections in the network

--tls.yellow.time defines the duration of yellow phase in seconds

--tls.left-green.time defines the protected left turn movement green time in seconds

--tls.allred.time defines the duration of all red phase in seconds

1.2 Appendix 2

Traffic demand was prepared by calling python randomTrips.py -n net.net.xml -r random.rou.xml --fringe-factor=100000000 --period=0.5 -e 3600.

where randomTrips.py is a Python script tool provided by SUMO.

-n net.net.xml defines the location of the network file.

-r defines the name of the output route file.

--fringe-factor defines the ratios of through and internal traffic demand in the network. An extremely large number is used here to eliminate the internal traffic demand in the network.

--period defines the 1/number of vehicles generated per second. 0.5 used here means two vehicles will be generated per second in the network. 

-e defines the end simulation step of generating trips, so here one hour traffic demand is generated.

Reprints and permissions

© 2024 Canadian Society for Civil Engineering

Policies and ethics