https://github.com/karthikeya-akhandam/snake-gama-using-deep-q---learning
Using Deep Q - Learning to implement reinforcement learning in classic Snake game
https://github.com/karthikeya-akhandam/snake-gama-using-deep-q---learning
dee pygame pyth pytorch reenforcementlearning
Last synced: 5 months ago
JSON representation
Using Deep Q - Learning to implement reinforcement learning in classic Snake game
- Host: GitHub
- URL: https://github.com/karthikeya-akhandam/snake-gama-using-deep-q---learning
- Owner: Karthikeya-Akhandam
- Created: 2025-05-06T13:56:20.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2025-05-06T14:04:20.000Z (over 1 year ago)
- Last Synced: 2025-05-12T08:07:02.304Z (over 1 year ago)
- Topics: dee, pygame, pyth, pytorch, reenforcementlearning
- Language: Python
- Homepage:
- Size: 13.7 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# 🐍 Snake AI using Deep Q-Learning
This project implements an AI agent to play the classic Snake game using Deep Q-Learning (DQN), a reinforcement learning algorithm. The game environment is built using PyGame, and the AI is trained using PyTorch.
---
## 📌 Problem Statement
Can an AI agent learn to play and master the classic Snake game using reinforcement learning techniques?
This project explores how Deep Q-Learning can be used to teach an AI agent to:
- Navigate the game grid
- Eat food to increase score
- Avoid collisions with walls and its own body
- Maximize performance through training
---
## 🎮 Game Overview
- Snake moves in a grid environment.
- Eats food to grow and increase score.
- Dies on collision with wall or itself.
- The agent must learn efficient and safe navigation.
---
## 🧠 Deep Q-Learning (DQN)
- DQN is a value-based reinforcement learning algorithm.
- The agent uses a neural network to approximate Q-values.
- Key components:
- State: 11-dimensional binary vector representing current game state
- Action: [straight, right, left]
- Reward:
- +10 for eating food
- -10 for dying
- -0 per step (to encourage faster solutions)
- Experience Replay: Stores past experiences and trains on batches
- Target Network: Helps stabilize learning
---
## 🎲 Epsilon-Greedy Strategy
- Balances exploration vs exploitation.
- Starts with high ε (exploration), gradually decays over time.
- Random action with probability ε.
- Best-known action with probability 1 - ε.
- Ensures the agent explores enough early on and exploits learned knowledge later.
---
## 🔧 Technologies Used
| Technology | Purpose |
|------------|---------|
| Python | Programming language |
| PyTorch | Neural network and deep learning |
| PyGame | Game development and environment |
| NumPy | Numerical operations |
| Matplotlib | Plotting training graphs |
---
## 🏗️ Architecture Diagram (8 Steps)
1. Game Environment (PyGame)
2. State Representation (11 features)
3. Neural Network Model (PyTorch)
4. Experience Replay Buffer
5. Epsilon-Greedy Policy for action selection
6. Q-value Calculation using Bellman Equation
7. Network Optimization (Backpropagation)
8. Performance Visualization (Matplotlib)
---
## 🛠️ Implementation Steps (6 Steps)
1. Create the Snake game using PyGame.
2. Represent the game state using an 11-dimensional binary vector.
3. Build the Deep Q-Network using PyTorch.
4. Use Epsilon-Greedy strategy for exploration.
5. Implement Experience Replay and train with Q-learning updates.
6. Plot training performance to visualize learning.
---
## 📁 Project Structure
```plaintext
.
├── agent.py # Contains DQNAgent class for decision-making
├── model.py # Neural network definition using PyTorch
├── game.py # Snake game environment using PyGame
├── train.py # Training loop and performance tracking
├── utils.py # Helper functions (optional)
├── screenshots/ # Graphs and game visuals
├── requirements.txt # List of dependencies
└── README.md # Project documentation