some experiments on the Torcs environment
http://blog.naver.com/dnddnjs11/221074046378
Just see the state, action, reward of the environment Simple but important step in RL
Make simple actor-critic agent which can predict action but can't learn from interaction.
Make simple DDPG agent. See https://github.com/yanpanlau/DDPG-Keras-Torcs I have made the code simpler.