Drone Simulation with Neural Network Integration

This project simulates a drone navigating towards targets within a 2D environment. The simulation is designed to integrate human control or neural network models trained by genetic algorithm for autonomous navigation.

Features

• Drone Dynamics: Realistic physics for drone movement including acceleration, angular rotation, and thruster control.
• Target Tracking: The drone aims to reach randomly generated or pre-defined targets.
• Neural Network Integration: Supports custom-trained (with genetic algorithm) neural networks for autonomous navigation.
• Visualization: Real-time simulation using pygame.

---

Getting Started

Prerequisites

• Python 3.9 or later
• Virtual environment (optional, but recommended)

Installation

1. Clone the repository:

   git clone https://github.com/skublin/drone-ea.git
   cd drone-ea

2. Install dependencies:

   pip install -r requirements.txt

3. Place any pre-trained models in the models directory.

---

Running the Simulation

To run the simulation with a pre-trained neural network model:

python main.py --model models/model-1.h5

For manual control (keyboard):

python main.py

---

Key Controls (Manual Mode)

Key	Action
W	Increase thrust (both engines)
S	Decrease thrust (both engines)
A	Decrease left engine thrust
D	Decrease right engine thrust

---

Files and Structure

• main.py: Entry point for running the simulation.
• train.py: Utilities for training or evaluating neural network models.
• simulation.py: Manages the game state and logic.
• drone.py: Handles drone physics and rendering.
• target.py: Defines target behavior and drawing.
• settings.py: Central configuration file for the simulation.

---

Neural Network Integration

You can integrate custom neural networks trained to navigate the drone. The model should:

• Accept a 6-dimensional input vector:
  • [dx, dy, vx, vy, angular_velocity, angle]
• Output 2 values:
  • [force_amplitude, force_difference]

To use a custom model, save it in .h5 or .pkl format and specify its path in main.py.

---

Future Enhancements

• Add pre-trained models.
• Include reinforcement learning examples.
• Add more visual effects and drone states.

---

License

This project is licensed under the MIT License - see the LICENSE file for details.