Lunar Lander: Deep Q-Learning Approach for Autonomous Lunar Module Landing

This project presents the design and implementation of an autonomous reinforcement learning agent developed to solve the Lunar Lander simulation problem using deep reinforcement learning techniques.The project simulates the soft landing of a spacecraft in a 2D environment, leveraging the LunarLander-v2 environment from the Gym toolkit and a custom training pipeline built in Python.A Deep Q-Network (DQN) approach is employed, in which a neural network estimates the Qvalues for each action based on the current state.The implementation includes experience replay, epsilon-greedy action selection, and periodic target network updates.Through iterative training over thousands of episodes, the agent learns an optimal policy to minimize fuel use and achieve safe landings.Performance evaluation shows a steady improvement in cumulative rewards, and final gameplay results indicate a high landing success rate.This work demonstrates the effectiveness of reinforcement learning in solving dynamic control problems and its applicability to real-world autonomous navigation systems.