Control Barriers in Bayesian Learning of System Dynamics

Demo

Installation

If you are using a Python virtual environment, edit setup.bash to point to the activate script of the virtual environment. One way to do that is:

python3_ver () 
{ 
    python3 -V 2>&1 | sed -e 's/.*3.\([0-9]\).*/py3\1/'
}
mkdir -p .tox/
virtualenv --python=python3 .tox/$(python3_ver)

Install gurobi. Edit setup.bash to set GUROBI_LIB_PATH
Activate environment and install current package in edit mode

source setup.bash
pip install -e .

Run tests

pytest

Unicycle demos

To run experiment where unicycle with mean CBF collides with the obstacle run

python -c 'from bayes_cbf.unicycle_move_to_pose import unicycle_mean_cbf_collides_obstacle; unicycle_mean_cbf_collides_obstacle()'

To run experiment where unicycle with Bayes CBF drives safely between the obstacles

python -c 'from bayes_cbf.unicycle_move_to_pose import unicycle_bayes_cbf_safe_obstacle; unicycle_bayes_cbf_safe_obstacle()'

To run experiment where unicycle gets stuck without learning run

python -c 'from bayes_cbf.unicycle_move_to_pose import unicycle_no_learning_gets_stuck; unicycle_no_learning_gets_stuck()'

To run experiment where unicycle passes safely through obstacles due to learning run

python -c 'from bayes_cbf.unicycle_move_to_pose import unicycle_learning_helps_avoid_getting_stuck; unicycle_learning_helps_avoid_getting_stuck()'

Pendulum demos

To run pendulum example and compare MVGP with CoGP

from bayes_cbf.pendulum import learn_dynamics_matrix_vector
learn_dynamics_matrix_vector()

To compare the computation requirement for MVGP vs CoGP vs diagonal

from bayes_cbf.pendulum import speed_test_matrix_vector
speed_test_matrix_vector()

Please cite

@InProceedings{khojasteh2020probabilistic,
  title = 	 {Probabilistic Safety Constraints for Learned High Relative Degree System Dynamics},
  author =       {Khojasteh, Mohammad Javad and Dhiman, Vikas and Franceschetti, Massimo and Atanasov, Nikolay},
  booktitle = 	 {Proceedings of the 2nd Conference on Learning for Dynamics and Control},
  pages = 	 {781--792},
  year = 	 {2020},
  editor = 	 {Bayen, Alexandre M. and Jadbabaie, Ali and Pappas, George and Parrilo, Pablo A. and Recht, Benjamin and Tomlin, Claire and Zeilinger, Melanie},
  volume = 	 {120},
  series = 	 {Proceedings of Machine Learning Research},
  month = 	 {10--11 Jun},
  publisher =    {PMLR},
  pdf = 	 {http://proceedings.mlr.press/v120/khojasteh20a/khojasteh20a.pdf},
  url = 	 {https://proceedings.mlr.press/v120/khojasteh20a.html}
}

@article{dhiman2021TACControlBarriers,
  author={Dhiman, Vikas and Khojasteh, Mohammad Javad and Franceschetti, Massimo and Atanasov, Nikolay},
  journal={IEEE Transactions on Automatic Control},
  title={Control Barriers in Bayesian Learning of System Dynamics},
  year={2021},
  volume={},
  number={},
  pages={1-1},
  doi={10.1109/TAC.2021.3137059}
  }

vahidhj93/Bayesian_CBF