🚁 Drone Control

Drone control project using recurrent neural networks (GRU, LSTM, RNN) with trajectory optimization and Extended Kalman Filtering.

📋 Prerequisites

• Python 3.11
• uv - Ultra-fast Python package manager
• MATLAB (optional, for Simulink usage)

🚀 Installation

1. Install uv

python -m pip install uv

2. Clone the project

git clone https://github.com/solalbaudoincs/dronecontrol.git
cd dronecontrol

3. Create virtual environment and install dependencies

# uv automatically creates a venv and installs dependencies from pyproject.toml
uv sync

🎯 CLI Usage

The project provides a command-line interface (CLI) for model training and optimization report generation.

⚠️ All default parameters correspond to the parameters presented in our report.

Available Commands

1. Train a model

# Basic training with default parameters
uv run scripts/cli.py train --model-name gru

# Training with custom hyperparameters
uv run scripts/cli.py train \
    --model-name gru \
    --epochs 30 \
    --batch-size 64 \
    --lr 0.005 \
    --hidden-dim 128 \
    --num-layers 2 \
    --dropout 0.1

Available options:

• --model-name: Model type (gru, lstm, rnn) - required
• --epochs: Number of training epochs
• --batch-size: Batch size
• --lr: Learning rate
• --hidden-dim: Hidden layer dimension
• --num-layers: Number of RNN layers
• --dropout: Dropout rate
• --seed: Random seed (default: 42)
• --best-model-out: Path to save the best model

2. Generate optimization report

# Report generation with default trajectory
uv run scripts/cli.py run \
    --model-name gru \
    --trajectory default

# Full generation with all parameters
uv run scripts/cli.py run \
    --model-name gru \
    --trajectory 5-step \
    --use-ekf true \
    --use-simulink true \
    --optimize-trajectory true \
    --max-epochs 100 \
    --horizon 30 \
    --lr 0.1

Available options:

• --model-name: Model name - required
• --trajectory: Trajectory type (default, step, multi, smooth) - required (look here for values)
• --model-ckpt: Checkpoint path (default: {model_name}_best.ckpt)
• --use-ekf: Use Extended Kalman Filter (true/false, default: true)
• --use-simulink: Use Simulink for simulation (true/false, default: true)
• --optimize-trajectory: Enable trajectory optimization (true/false, default: true)
• --max-epochs: Maximum number of optimization epochs (default: 100)
• --dt: Time step (default: 0.05)
• --horizon: MPC horizon (default: 30)
• --lr: Optimizer learning rate (default: 0.1)
• --max-speed: Maximum speed (default: 9.81)

📊 Usage Examples

Complete Workflow: Training + Optimization

# 1. Train a GRU model
uv run scripts/cli.py train --model-name gru --epochs 50

# 2. Generate optimization report with smooth trajectory
uv run scripts/cli.py run \
    --model-name gru \
    --trajectory 10-step \
    --max-epochs 150

Test different trajectories

# Default trajectory
uv run scripts/cli.py run --model-name gru --trajectory default

# Step trajectory
uv run scripts/cli.py run --model-name gru --trajectory step

# Multi trajectory
uv run scripts/cli.py run --model-name gru --trajectory multi

# Smooth trajectory
uv run scripts/cli.py run --model-name gru --trajectory smooth

Compare different models

# Train all models
uv run scripts/cli.py train --model-name gru
uv run scripts/cli.py train --model-name lstm
uv run scripts/cli.py train --model-name rnn

# Generate reports
uv run scripts/cli.py run --model-name gru --trajectory default
uv run scripts/cli.py run --model-name lstm --trajectory default
uv run scripts/cli.py run --model-name rnn --trajectory default

📁 Project Structure

dronecontrol/
├── scripts/
│   └── cli.py              # Command-line interface
├── src/
│   └── dronecontrol/       # Main source code
├── data/                   # Training data
├── logs/                   # Training logs
├── models/                 # Trained models
├── optimization_results/   # Optimization results
├── predictions_plots/      # Generated plots
├── pyproject.toml          # Project configuration
└── README.md               # This file

🛠️ Development

Install development dependencies

uv pip install pytest

Run tests

pytest tests/

📝 Notes

• Trained models are saved in the models/ folder
• Intermediate checkpoints are in models_checkpoints/
• Reports and plots are generated in predictions_plots/
• Training logs are in logs/

🤝 Contributing

To contribute to the project:

1. Fork the project
2. Create a branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

📄 License

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