JSheldon3488/OpenAI_Gym_LunarLanderV2

Udacity Deep Reinforcement Learning OpenAI Gym LunarLander-v2 project.

OpenAI_Gym_LunarLanderV2

Udacity Deep Reinforcement Learning OpenAI Gym LunarLander-v2 project. Original code and project details can be found here.

Agents

Baseline Agent

The baseline agent is a Deep Q-Network with Experience Replay and Fixed Q-Targets. More details found in this paper.

Baseline + Double DQN

This agent is the same as the baseline agent in every way except Agent.learn now uses the Double Deep Q-Network algorithm.

Baseline + Prioritized Experience Replay

This agent is the same as the baseline agent with prioritized experience replay added. (Ran too slow, need faster priority replay buffer)

Baseline + Dueling DQN

Baseline + Prioritized Experience Replay and Dueling DQN

Dependencies

• OpenAI Gym
• PyTorch
• Numpy
• box2d
• matplotlib

For rendering on Windows

• Xming Full explenation here.