Abstract:

The stability of collision avoidance is directly relate to the safety of unmanned surface vehicle (USV). However, the uncertainty of perception about the velocity of moving obstacles seriously undermine the stability of collision avoidance. Thus, an uncertain velocity obstacle (UVO) method is proposed to solve this problem. In order to improve the stability of collision avoidance at the macro level, an adaptive threshold-based closest point of approach (CPA) is adopted to assess collision risk while a boundary buffer is used to calculate the type of International Regulations for Preventing Collisions at Sea (COLREGS). To prevent changes in collision avoidance strategy, the UVO is modeled in velocity space of the USV, and a gradient descent method is used to determine local optimization of the cost function. Contrast experiments in simulation platform between UVO and VO indicate that the UVO has better performance on three indicators: strategy changes, success rate, and safe distance. Typical encounters such as head-on, crossing, and overtaking are conducted in sea trials. The USV successfully avoids each moving obstacle in the experiment. The results demonstrate the stability and safety of the UVO method.

Key words: unmanned surface vehicle (USV), dynamic collision avoidance, uncertainty of obstacle velocity, velocity obstacle, collision risk assessment