Self-driving Car AI Simulation

A PyTorch-based implementation of a self-driving car using Deep Q-Learning. The car learns to navigate through a custom environment, avoiding obstacles while trying to reach its target.

Features

• Deep Q-Learning implementation for autonomous navigation
• Interactive environment with drawable obstacles
• Real-time visualization of car sensors and decision making
• Save/Load functionality for trained models
• Performance tracking and visualization

Interface

• Red Circle: The car with its direction indicator
• Blue Lines: Three sensors detecting obstacles
• Green Circle: Target destination
• Dark Blue Lines: Obstacles (sand) drawn by user
• Control Buttons:
  • Clear: Reset the environment
  • Save: Store trained model
  • Load: Use previously trained model

How to Use

1. Run map.py to start the simulation
2. Left-click and drag to draw obstacles (sand)
3. Watch the car learn to navigate to the green target
4. Use buttons to manage the simulation

Requirements

• Python 3.x
• PyTorch
• Pygame
• Numpy
• Matplotlib

Project Structure

• map.py: Main simulation environment
• ai.py: Deep Q-Learning implementation

The car uses three sensors to detect obstacles and learns to navigate through reinforcement learning.