End to end UAV hover controller using deep reinforcement learning algorithm in wind disturbed environment

Structure

All the source code is in folder "hover_rl/end2end"

Env

We use RotorPy, a Python-based multirotor simulation environment with aerodynamic wrenches, useful for education and research in estimation, planning, and control for UAVs.

Install

Please follow the install requirement in RotorPy

To run the training:

python3 hover_train.py

To run the testing:

python3 hover_test.py

future work:

1. train an end to end controller for path tracking instead of only hovering: on-going, code is in folder "hover_rl/path_trk"
2. train an controller combine with RL residual learning and nomial geometry(SE3) controller: on-going, code is in folder "hover_rl/res_rl"