Reinforcement Learning


Looking for industrial-strength RL baselines? Head over to the RLlib baselines to find full implementations of methods such as Ape-X, PPO and imitation learning algorithms.

Let’s get our hands dirty with some simple reinforcement learning methods! We will solve tasks such as the one pictured below using DQN implementations written from scratch using PyTorch.


  • The single agent tutorial will guide through a basic Dueling Double DQN implementation which controls a single train

  • The multi agent tutorial shows how to extend this method to the multi-agent setting. You will then be able to submit this agent to the NeurIPS 2020 Flatland challenge!

The code for these examples can be found in the starter kit repository.