- On-policy model free Sarsa reinforcement learning algorithm
- Linear function approximation using tilecoder
- Solving OpenAI Gym environments MountainCar and CartPole (over 10k timesteps for cartpole)
- It seems it's not enough for LunarLander, a neural network may be needed
Algorithm is from Sutton & Barto's RL Book
Tile coder is from http://incompleteideas.net/tiles/tiles3.html