Implementation of ProBF: Learning Probabilistic Safety Certificates with Barrier Functions presented at NeurIPS 2021 SafeRL workshop. (https://arxiv.org/abs/2112.12210)
This code compares the our ProBF-GP framework with the prior art learned Control Barrier Functions which uses a neural network (LCBF-NN) [2]. We build upon the code in [1]. We add a controller that uses the mean and variance of the predictions to solve the ProBF-convex program.
Please reach out for any questions!
[1] Python simulation and hardware library for learning and control. https://github.com/learning-and-control/core.git. Accessed: 2021-11-17.
[2] A. Taylor, A. Singletary, Y. Yue, and A. Ames. Learning for safety-critical control with control barrier functions. In Learning for Dynamics and Control, pages 708–717. PMLR, 2020.