We saw previously how to train a DDPG agent to drive a car on TORCS. How to use a PPO agent is left as an exercise for the interested reader. This is a nice challenge to complete. The PPO code from Chapter 7, Trust Region Policy Optimization and Proximal Policy Optimization, can be reused, with the necessary changes made to the TORCS environment. The PPO code for TORCS is also supplied in the code repository (https://github.com/PacktPublishing/TensorFlow-Reinforcement-Learning-Quick-Start-Guide), and the interested reader can peruse it. A cool video of a PPO agent driving a car in TORCS is in the following YouTube video at: https://youtu.be/uE8QaJQ7zDI
Another challenge for the interested reader is to use Trust Region Policy Optimization (TRPO) for the TORCS racing car problem. Try this too, if interested! This is one way to master RL algorithms.
...
