Multi Agent Environment for TORCS based on https://github.com/openai/multiagent-particle-envs
An example algorithm which uses MPE is Multi Agent DDPG. We have shown MADDPG for TORCS using MPE. Link of code of MADDPG: https://github.com/openai/maddpg Link of paper: https://arxiv.org/pdf/1706.02275.pdf
This environment can be used for training any multi agent algorithm. Inter-vehicle communication support is also provided.