Here is my fun side project to train an agent to walk in the BipedalWalker environment using Reinforcement Learning. The environment is obtained from Gym OpenAI (here). Deep Deterministic Policy Gradient (DDPG) method with epsilon-decay exploration is used in this work.
Here is the result after training:
The agent is also trained on a more complex version of the BipedalWalker environment (BipedalWalkerHardcore) in another repository.