智能船舶的反应型避碰规划研究

对于多船会遇场景而言，在反应型避碰规划中考虑操船者意图具有明显的优越性，可在紧迫局面下更合理地执行《1972年国际海上避碰规则》（即COLREGs）的背离规则。对此，提出一种基于隐马尔可夫模型的操船意图预报方法，将生成的目标船意图预报结果与经典速度障碍法相融合，产生更合理的碰撞危险评估结果，据此建立的反应型避碰规划算法比经典速度障碍法更安全。依托案例三体船完成的自主避碰试验，验证所提改进算法的有效性、可靠性和实时性，为智能船舶的自主避碰研究提供参考。

Reactive Collision Avoidance Planning for Intelligent Ships

Autonomous collision avoidance is the critical technology for intelligent ships. In multi-ship encounter situations, consideration of the target ship intention in reactive collision avoidance planning has the superior advantages, which could generate the urgent collision avoidance strategy beyond COLREGs rules. The human-driven intention prediction method based on Hidden Markov Model is proposed. The prediction results of target ship intention are combined with classical Velocity Obstacle method to form the improved collision risk evaluation. The resulting reactive collision avoidance planning always produce the solutions much safer than those of classical Velocity Obstacle method. Successful on-water tests of autonomous collision avoidance with the trimaran model demonstrate the validity and reliability of the proposed algorithm.