Welcome to the Lunar Lander Reinforcement Learning Project!

This project demonstrates the application of various reinforcement learning (RL) algorithms to solve the Lunar Lander problem, a classic control problem provided by OpenAI's Gym environment.

Overview

• In this project, we train an agent to successfully land a lunar module on a designated landing pad using reinforcement learning.
• The goal is to achieve a soft landing, minimizing the impact velocity and landing within the designated area.
• The agent receives rewards based on its actions, such as maintaining a proper orientation, landing softly, and avoiding crashing.

Algorithms Implemented

The project includes implementations of the following RL algorithms:

• Deep Q-Network (DQN)
• Double DQN
• Dueling DQN
• Policy Gradient Methods (REINFORCE)
• Actor-Critic Methods (A2C, A3C)
• Proximal Policy Optimization (PPO)

Getting Started

Prerequisites

Ensure you have the following installed:

• Python 3.7+
• pip (Python package installer)
• Git

Results

The results of the training, including the performance metrics and trained models, are saved in the logs/ and models/ directories, respectively. You can visualize the training progress using TensorBoard:

Acknowledgements

This project uses the following open-source libraries:

• OpenAI Gym
• PyTorch
• NumPy
• Matplotlib

Thank you for checking out this project! If you have any questions or feedback, feel free to open an issue or reach out.