An improved real-time collision-avoidance algorithm based on Hybrid A* in a multi-object-encountering scenario for autonomous surface vessels

Abstract

Collision-avoidance algorithms for maritime autonomous surface vessels are increasingly being analyzed for challenging scenarios. However, the trade-off between the computation’s effectiveness and efficiency is always an intractable problem, especially in complex, multi-object-encountering cases. This paper proposes an improved Hybrid A* algorithm that searches a node map with discrete control behaviors as ‘movement options’. The real-time performance in maritime, multi-object- encountering scenarios is improved by narrowing the search space with the proposed pre-processing method, i.e., the collision velocity check (CVC). The following is incorporated within the mentioned approach: the compliance of the relevant rules in International Regulations for Preventing Collisions at Sea and the dynamics of the ship. A comparative study with various multi-target scenarios is conducted with a baseline method. Simulation results show the effectiveness of the proposed method. The use of the CVC significantly reduces the time of the computation with little reduction in the solution’s accuracy. Additionally, the proposed method is applied within a real-time simulation environment, which is fed by logged data collected during an actual sea trial.