Gymnasium Lunar Lander DQN

This repository contains two different approaches to solving the Gymnasium Lunar Lander environment: a simple rule-based algorithm and a Machine Learning (Reinforcement Learning) model.

Demonstration

1. Machine Learning Model (DQN)

The ML implementation uses a Deep Q-Network (DQN) algorithm written with PyTorch.

Usage:

• Train the Model (Headless):

  python main.py

  This trains the model over 1000 fast episodes and saves the progress to lunar_lander_model.pth.

• Watch the Trained Model (Visual):

  python main.py play

  This will open a render window and let the built-in trained agent attempt to land the craft based on its network weights.

---

2. Rule-Based Algorithm (If-Statement Logic)

This was the initial approach containing hand-tuned conditions (if/elif/else) that manually monitor position, velocity, and tilt to guide the module.

Usage:

• Watch the Rule-Based Control:

  python rule_based.py