Safe Probabilistic Planning for Human-Robot Interaction using Conformal Risk Control

This paper presents a novel probabilistic safe control framework for human-robot interaction that combines control barrier functions (CBFs) with conformal risk control (CRC) to provide formal safety guarantees while considering complex human behavior. Safe deployment of autonomous robots in human environments remains challenging due to the inherent unpredictability in human actions, with traditional approaches either relying on restrictive distributional assumptions or defaulting to overly conservative worse-case bounds. Our approach uses CRC to quantify and control the prediction errors in CBF safety values and establishes formal guarantees on the probability of constraint satisfaction during interaction. We further introduce an online update scheme that dynamically adjusts the safety margins produced by CRC based on the current interaction context. Through experiments in challenging human-robot navigation scenarios, we demonstrate that our approach significantly reduces collision rates and safety violations compared to baseline methods while maintaining high success rates in goal-reaching tasks and efficient control.