Skip to content

PyBullet CartPole and Quadrotor environments—with CasADi symbolic a priori dynamics—for learning-based control and RL

License

Notifications You must be signed in to change notification settings

ManUtdMoon/safe-control-gym

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Safe-control-gym

Overview

Note that this version is revision of the original repository and we focus on the circle tracking task of a 2D quadrotor in the $xz$-plane.

Major differences include:

  • Adding new environment configurations examples/constrained_tracking_eval.yaml and examples/constrained_tracking_reset.yaml about the initial position/velocity of the 2D quadrotor;
  • norm_act_scale=1.0 in safe_control_gym/envs/gym_pybullet_drones/quadrotor.py because the the original value 0.1 may lead to a limited control over the quadrotor and less violation. Therefore, 1.0 is harder and more capable;
  • Adding info about the tracking error of angular speed in the info variable;
  • The coefficients of different error in the reward function are modified but not essential.

Installation of safe-control-gym

$ sudo apt-get install libgmp-dev                      # Install a necessary lib
$ cd /your/path/to/safe-control-gym/
$ pip install -e .                                     # Install the repository

Please cite the original authors' paper:

@article{brunke2021safe,
         title={Safe Learning in Robotics: From Learning-Based Control to Safe Reinforcement Learning}, 
         author={Lukas Brunke and Melissa Greeff and Adam W. Hall and Zhaocong Yuan and Siqi Zhou and Jacopo Panerati and Angela P. Schoellig},
         journal = {Annual Review of Control, Robotics, and Autonomous Systems},
         year={2021},
         url = {https://arxiv.org/abs/2108.06266}}

About

PyBullet CartPole and Quadrotor environments—with CasADi symbolic a priori dynamics—for learning-based control and RL

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Python 100.0%