RndmVariableQ/Parl-Quadrotor-Velocity-Control

Baidu RL camp reproduction project

Parl-Quadrotor-Velocity-Control

This is a Baidu Reinforcement Learning Camp reproduction project, trying to implement Reinforcement Learning on a Quadrotor Velocity Control task.

Install:

Dependencies

• Python 3.7+ (On Windows, PARL only supprorts the enviroment with python3.6+).
• Parl 1.3.1 (https://github.com/PaddlePaddle/PARL)
• Paddlepaddle 1.6.1 or higher (https://github.com/PaddlePaddle/Paddle)
• RLSchool -

pip install parl==1.3.1
pip install paddlepaddle==1.6.1
pip install rlschool

Demo:

Trained through 500000 episodes + 110000 episodes of fine tuning

• Average Test Score: -19.18