Skip to content

Latest commit

 

History

History
17 lines (11 loc) · 1.59 KB

README.md

File metadata and controls

17 lines (11 loc) · 1.59 KB
Raw

Control Barrier Function refinement with HJ Reachability

This repository contains the implementation of refineCBF, accompanying the paper Refining Control Barrier Functions using HJ Reachability by Sander Tonkens and Sylvia Herbert, IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022.

This project combines CBFs and reachability, to refine learned or analytical CBFs and for making backup policy-based CBFs explicit.

In particular:

  • The refine_cbfs directory contains code to define a tabular CBF (a CBF defined over a grid) and provides an interface with hj_reachability and cbf_opt to define its dynamics.
  • The examples folder provides the simulation results for the paper mentioned above

Requirements

  • hj_reachability: Toolbox for computing HJ reachability leveraging jax: pip install --upgrade hj-reachability. Requires installing jax additionally based on available accelerator support. See JAX installation instructions for details.
  • cbf_opt: Toolbox for constructing CBFs and implementing them in a safety filter (using cvxpy). Run pip install "cbf_opt>=0.6.0" or install locally using the Github link and run pip install -e . in DIR to install.
  • experiment_wrapper: Self-contained toolbox for running experiments that have analytically defined dynamics and measurement models. Run pip install "experiment-wrapper>=1.1" Github link and run pip install -e . in DIR to install