Simulation of Urban Mobility (SUMO) for Multi-Agent Reinforcement Learning (MARL)

This codes has only been confirmed on Windows.

Install SUMO

. Installing SUMO from website creates SUMO files under "C:\Program Files (x86)\Eclipse\Sumo".

Install TRaCI package

For SUMO, need to install TRaCI for traffic control interface

$ pip install traci

Run the code

python main.py

R flag, which is intended to mean 'rendering', can show how the agents actually move.

References

Multi-Agent papers

• Lowe, Ryan, et al. "Multi-agent actor-critic for mixed cooperative-competitive environments." Advances in neural information processing systems 30 (2017).

Traffic Environment

• Reward definition
  • Egea, Alvaro Cabrejas, et al. "Assessment of reward functions for reinforcement learning traffic signal control under real-world limitations." 2020 IEEE International Conference on Systems, Man, and Cybernetics (SMC). IEEE, 2020.