Real-Time Automated Evaluation of COLREGS-Constrained Interactions Between Autonomous Surface Vessels and Human Operated Vessels in Collaborative Human-Machine Partnering Missions

摘要

This paper explores an extension of the real-time evaluation of COLREGS-based collision avoidance interactions between autonomous surface vessels and human-operated surface vessels. Our previous work developed the algorithms that evaluate and quantify a ship's compliance with the collision regulations, safety, and mission efficiencies with respect to the overall goal(s). Our previous work is extended in this paper by establishing a light-weight program to assign penalties to offenders of safety or protocol violations during human-machine collaborative on-water interactions. Vessels interacting in this DARPA-sponsored Aquaticus mission are grouped into teams consisting of both human and robot counterparts. These teams play a “capture the flag” like game while being required to obey the maritime collision avoidance regulations. This paper is a first step in the field toward evaluating collision avoidance rules in the context of a human or robotic vehicle cheating the COLREGS against its opponent to gain a mission advantage. The problem is representative of interactions likely seen on the open ocean using a combination of autonomous and human-operated multi-vehicle collision avoidance interactions to larger scale maritime vessel traffic interactions operating under COLREGS protocol constraints. The vessels deploy in a distributed collaborative pattern to compete against the opposing team. Upon detection of a violation, the offending vessel(s) are required to complete their penalty actions prior to being allowed to proceed in their mission goal. Human offenders are able to be linked to haptic devices that give real-time feedback using vibrations or similar queuing.