Perception-Aware Control Barrier Function for Safe Navigation Under Uncertain LiDAR Observation

Abstract

This paper presents a discrete-time Perception-Aware Control Barrier Function (PA-CBF) framework for LiDAR-based robot navigation under uncertainty. The method combines beam-wise barrier constraints with uncertainty-aware range-rate estimation from sequential LiDAR measurements and a predictive finite-horizon extension based on constant-input propagation. The resulting safety filter provides probabilistic safety guarantees at discrete sampling instants while preserving convexity and real-time tractability. Simulation results show improved safety, smoother control, and higher task success than distance-only and reactive baselines in dynamic environments.