AirSim Unity Reinforcement Learning Quadrotor Pytorch

Reinforcement learning for AirSim Unity Quadrotor environment and DQN pytorch

• Youtube

Environment (Setup)

• Microsoft Airsim on Unity
• Unity-2018.2.7f1 (Linux version)
• Ubuntu 18.04
• CPU Intel Core i7-8750H

Test

Simulation test of agent_v0.py is completed, but not in agent.py.

agent_v0.py is simple version of dqn script to test work well in Airsim Simulator.

To use

Put 3 files env.py , agent.py , run.py in directory ~/Airsim/PythonClient/multirotor.

To run

• Launch Unity Editor.

  cd ~/Unity-2018.2.7f1/Editor
  ./Unity
  

• Choose your own environment.

• Default example UnityDemo is given in directory ~/Airsim/Unity .

• Select Play button.

• If simulation is playing, select Drone Mode.

• Now run script.

  cd ~/Airsim/PythonClient/multirotor
  python run.py
  
  # or run
  python agent_v0.py
  

Author

• Subin Yang
• [email protected]