An OpenAI gym environment fully compatible with OpenAI baselines implementation of various Deep Reinforcement Learning algorithms (DDPG,A2C,TRPO,PPO,ACER,ACKTR...).
Python 3
This environment is intended to use the modified version of TORCS that I made:
https://github.com/DamienLancry/blocking_torcs
You will also need to
pip install gym
git clone https://github.com/DamienLancry/gym_torcs
python -m gym_torcs.gym_torcs.envs.torcs_env
It will not work if your configuration xml file is not correctly edited as explained in https://github.com/DamienLancry/blocking_torcs . So far I only experimented with one scr_server and no opponents.
https://github.com/ugo-nama-kun/gym_torcs (For various reasons, this environment was not compatible with OpenAI baselines.)