Using reinforcement learning algorithm TD3 from Parl to solve quadrotor velocity control based on RLSchool environment.
RLSchool is a group of reinforcement learning simulation environments。
This repository includes the user-friendly jupyter notebook version of the solution. The code is written using PaddlePaddle. The TD3 algorithm from PARL is used.
After training for 1000 episodes, an average score of -20 is achieved and being stable.
