A real-time collision avoidance learning system for Unmanned Surface Vessels

A great amount of effort has been devoted to the study on Unmanned Surface Vehicles (USV) due to an increasing demand for their use in a variety of maritime applications. Real-time autonomous collision avoidance system is the pivotal issue here, in which reliable collision risk detection and the adoption of a plausible collision avoidance maneuver play a key role. Existing studies on this subject seldom integrate the COLREGS guidelines, however, and in order to ensure maritime safety, it is of fundamental importance that they should be obeyed at all times. In this paper, we presented an approach to real-time collision avoidance that complies with the COLREGS rules for USV. The Evidential Reasoning (ER) theory is employed to evaluate the collision risks with obstacles encountered and trigger a prompt warning of a potential collision. Then, we extend and adopt the optimal reciprocal collision avoidance (ORCA) algorithm so as to determine a collision avoidance maneuver that is COLREGS compliant. The proposed approach takes into consideration the fact that other obstacles also sense their surroundings and react accordingly, conforming to a practical marine situation when making a decision concerning collision-free motion. A number of simulations have been conducted in order to confirm the validity of the theoretic results obtained.