/Pybullet-Gym-Drones

Drone reinforcement learning with multiple tasks in pybullet and OpenAI Gym environment

Primary LanguagePython

Pybullet-Gym-Drone - 强化学习无人机

Drone auto-navigation stimulation developed in Pybullet + OpenAI Gym Environment. Trust Region Policy Optimization (TRPO) and Proximal Policy Optimization (PPO) is implemented.

drone_moving

Result

Sucess Rate DroneNavigationV0 DroneNavigationV1
TRPO 50.0% 39.2%
PPO 91.2% 11.2%

result

Run

  • You can run & see the real time successful rate by:
python run.py --model PPO --version 0
  • To train the model, you can run:
python PPO_trainer.py

or

python TRPO_trainer.py
  • You can modify the GUI in drone_envs/config "display": p.GUI,

Thanks Medium series on creating OpenAI Gym Environments with PyBullet.