Control Barrier Functions for Sample-Based Beliefs

This repository accompanies our paper: "Risk-Aware Robot Control in Dynamic Environments using Belief Control Barrier Functions", accepted at CDC 2025, Rio de Janeiro, Brazil.

Why This Tool?

Real-world robots operate under stochastic uncertainties — caused by unmodeled dynamics, noisy sensors, and partial observability.

This tool is designed to guarantee safety using only i.i.d. samples from belief distributions, without relying on parametric representations.

✅ Key Features

• Provable safety guarantees under uncertainty
• No parametric modeling required — works directly with i.i.d. samples from any Bayesian state estimator
• Real-time performance at kilo-Hz rates, suitable for robotics

Installation

Clone the repository:

git clone https://github.com/KTH-RPL-Planiacs/sample_based_bcbf
cd sample_based_bcbf

Create and activate the virtual environment using Mamba.

mamba env create -f environment.yml
mamba activate sample_based_bcbf

How to Run

Run the main experiment and collect data:

python3 main.py 

Visualize results with animation:

zsh scripts/create_gif.sh

BibTeX

If you find this work useful, please consider citing:

@article{han2025risk,
  title={Risk-Aware Robot Control in Dynamic Environments Using Belief Control Barrier Functions},
  author={Han, Shaohang and Vahs, Matti and Tumova, Jana},
  journal={arXiv preprint arXiv:2504.04097},
  year={2025}
}