Risk Euclidean Distance-based Model Predictive Path Integral to Safety-Critical Obstacle Avoidance
Zihao Huang, Ruocheng Li, Weili Chen, Zicong Lin, Zhipeng Wu, Bo Zhang
Abstract
Sampling-based Model Predictive Control (MPC) algorithms such as Model Predictive Path Integral (MPPI) excel in managing nonlinear constraints and complex systems. However, their conventional sampling strategies often result in suboptimal local solutions. To address this problem, we propose RESM-MPPI, a novel dynamic obstacle avoidance algorithm that integrates the Risk Euclidean Safety Metric (RESM), which is an enhanced version of the Conventional Euclidean Safety Metric (CESM), to more effectively quantify collision risks between autonomous mobile robots (AMRs) and dynamic obstacles. Our approach extends the classical Control Barrier Function (CBF) framework by introducing the Risk Control Barrier Function (RCBF) and integrating a Control Obstacle Avoidance Annealing (COAA) sampling strategy to enhance obstacle avoidance performance. This combination enables the generation of safe and smooth trajectories for AMRs in dynamic environments. Sufficient simulations and real-world experiments demonstrate the effectiveness of the proposed method. Experimental videos are available at: https://youtu.be/WUchIzz_0wU.
BibTeX
@inproceedings{iros2025_riskeuclideandis,
title = {Risk Euclidean Distance-based Model Predictive Path Integral to Safety-Critical Obstacle Avoidance},
author = {Zihao Huang and Ruocheng Li and Weili Chen and Zicong Lin and Zhipeng Wu and Bo Zhang},
booktitle = {IROS 2025},
year = {2025}
}