Flappy Bird AI using Genetic Algorithm

Welcome to the Flappy Bird AI 🐦 using Genetic Algorithm. This project is a simple implementation of the Flappy Bird game using the Pygame library, Neural Networks, and Genetic Algorithm. The game is played by a population of birds, each with its own Neural Network. The birds are trained using the Genetic Algorithm to learn how to play the game! 🤖.

📚 What's Inside

• main.py: The main file that runs the game and the Genetic Algorithm.
• main.ipynb: The main file that runs the game and the Genetic Algorithm with more detailed explanations.
• GA.py: The Genetic Algorithm implementation.
• model.py: The Neural Network model.
• game.py: The Flappy Bird game using Pygame.
• Best_Weights.npz: The best weights of the Neural Network after training.

🌐 Overview

The project is divided into 3 main parts:

1. The Flappy Bird game using Pygame
2. The Neural Network
3. The Genetic Algorithm

🎥 Sample Video

🚀 Getting Started

1. Clone the repository:

   git clone https://github.com/oyounis19/Flappy-Bird-AI.git

2. Install the required packages:

    pip install -r requirements.txt

3. Run the Evolution:

   python main.py

4. Sit Back and Enjoy:

   The game will start and the birds will start learning how to play the game.

🧠 Neural Network Architecture

Determines whether the bird should jump or not based on the inputs.

• Input Layer: 5 nodes

  • The vertical position of the bird.
  • The horizontal distance to the next pipe.
  • The height of the next gap.
  • The bird's velocity.
  • The gap's center position.

• Hidden Layer: 8 nodes

  • The hidden layer is used to learn the complex patterns in the input data.
  • Activation Function: ReLU

• Output Layer: 1 node

  • The output layer is used to determine whether the bird should jump or not.
  • Activation Function: Sigmoid

🛠️ Customization

You can customize the game by changing the following parameters:

• 🧬 Genetic Algorithm Parameters:

  • POPULATION_SIZE: The number of birds in the population.
  • MUTATION_RATE: The rate at which the Neural Networks are mutated.
  • CROSSOVER_RATE: The rate at which the Neural Networks are crossed over.

• 🕹️ Game Parameters:

  • Bird Speed: The speed at which the bird moves.
  • FPS: The frames per second of the game.
  • Pipe Height Range: The range of the height of the pipes.
  • Number of Pipes: The number of pipes per screen.
  • Gap Size: The gap between the pipes.
  • Pipes Color: The color of the pipes.
  • Screen Width: The width of the game screen.
  • Screen Height: The height of the game screen.
  • Jump Height: The height at which the bird jumps.

• 🧠 Neural Network Parameters:

  • Hidden Layer Size: The size of the hidden layer of the Neural Network.
  • Threshold: The threshold at which the bird jumps.

Testing Your Network

Once trained, put your neural network to the test. First save the weights of the best_nn returned by the Genetic Algorithm. Then, load the weights and run test method by the Genetic Algorithm.

# Save the weights
best_nn.save_weights('Best_Weights')

# Initialize Neural Network
nn = NeuralNetwork()

# Load the weights
nn.load_weights('Best_Weights.npz')

# Test the Neural Network
ga.test(nn, num_episodes=5)

Or don't save the weights and just test the Neural Network.

# Initialize Neural Network
nn = NeuralNetwork()

# Replace weights
nn.weights = best_nn.weights

# Test the Neural Network
ga.test(nn, num_episodes=5)

💡 Contributing

Contributions are welcome! If you've found a typo, want to add more content, or improve something, feel free to open an issue or submit a pull request.

Happy learning! 🚀