완화된 제어장벽함수 제약이 적용된 MPC 기반 동적 장애물 회피 알고리즘의 최적화 연구

Abstract

This paper presents an rCBF–MPC (relaxed Control Barrier Function–Model Predictive Control) framework for robust collision avoidance of autonomous surface vessels (ASVs) in congested maritime environments. In coastal waters, multiple ASVs often operate in close proximity, dynamically changing their positions and significantly increasing the risk of collision. While advanced predictive control approaches can ensure safe navigation, complex multi-obstacle interactions and external uncertainties necessitate additional constraints to guarantee reliable safety margins. The proposed method integrates a relaxed Control Barrier Function into a discrete-time MPC structure. By defining and enforcing a mathematically rigorous safe set within the prediction horizon, the system effectively maintains adequate separation from potential hazards, simultaneously mitigating computational overhead typically associated with strict barrier functions. Model tests conducted with a twin-hull ASV demonstrate that the rCBF –MPC framework effectively maintains safe distances and achieves stable maneuvering. Compared to traditional velocityobstacle- based methods, the proposed integrated approach offers enhanced robustness against sensor noise and rapidly changing traffic conditions, effectively addressing the limitations inherent in single-step avoidance strategies. In summary, this study provides an improved collision avoidance methodology with enhanced reliability, reduced computational complexity, and scalability suitable for real-world operations of autonomous surface vessels.