KohlerHECTOR/LunarLander-v2

Using Stable-Baselines 3 library to train a LunarLander-v2 agent. With Yannis Elrharbi-Fleury : https://github.com/sohio92

LunarLander-v2

REQUIRES PYTHON<3.9 Using Stable-Baselines 3 library to train a LunarLander-v2 agent. Methods used: DQN and CEM. Train a dqn agent on LunarLander-v2:

python3 train_dqn.py

Use Tensorboard to visualize learning:

tensorboard --logdir dqn_lunarLand2_tensorboard/

Key plots:

Gif of the evolution of the policy learned with DQN (reward 260 +- 30):

Gif of the evolution of the policy learned with CEM (reward 280 +- 18):

Visulaizations of the policy learned by CEM (reward 280 +- 18):

Visulaizations of the policy learned by DQN (reward 260 +- 30):