- Download simulator binaries: https://github.com/tawnkramer/gym-donkeycar/releases
- Install Poetry following the official docs: https://python-poetry.org/docs/#installation
- Go to the project folder
cd frl_donkey- Install the indenpandencies to virtual environment
poetry install- Install Flower federated learning:
pip install flwr-nightly- Install stable-baselines3:
pip install stable-baselines3[extra]- Install master version of gym donkey car:
pip install git+https://github.com/arthur-chen0/gym-donkeycar.git- Set up the config.ini
| Key | Description | Value |
|---|---|---|
| exe_path_linux | The path of your donkey car simulator installed on Linux system | |
| exe_path_mac | The path of your donkey car simulator installed on Mac system | |
| host | The federated learning server address | 127.0.0.1 |
| rlAlgo | The algorithm your reinforcement learning using | PPO, TQC |
| aggregationFn | The aggregation function of federated learning | FedAvg, FedProx |
| dp | Use the fixed or adaptive clipping. Set it to 'none' if you do not need differential privacy | dp_fixed_clipping, dp_adaptive_clipping |
| policy | The neural network used in reinforcement learning | Cnn, Mlp |
| env | environment number | 0-3 |
| rounds | The number of federated learning rounds | A number you want |
| timesteps | The number of total learning steps | A number you want |
| clients | The number of federated learning clients | A number you want |
- Start the server, and it will run the strategy that you set up in config.ini
python frl_server.py- Start the client with ID, number of clients at least 2
python frl_client.py --id 1