Rolls-Royce: Autonomous vessels can meet COLREGSMar 23, 2018 09:10 AM
Tests show (AI)-based navigation systems can enact the rules to avoid collisions
The following is text of a news release from Rolls-Royce:
(LONDON) — Rolls-Royce has completed the $1.85 million MAXCMAS (MAchine eXecutable Collision regulations for Marine Autonomous Systems) research project, demonstrating that the operation of autonomous vessels can meet, if not exceed, current collision avoidance (COLREGS) rules.
Project partners Lloyd’s Register, Warsash Maritime Academy (WMA), Queen’s University Belfast and Atlas Elektronik (AEUK) found that use of newly developed algorithms allowed existing COLREGS to remain relevant in a crewless environment, finding that artificial intelligence (AI)-based navigation systems were able to enact the rules to avoid collision effectively, even when approaching manned vessels were interpreting the rules differently.
A key aspect of the research was the use of WMA’s networked bridge simulators. These highly immersive simulators, typically used for seafarer training, were used to analyze reactions from the crew when faced with a range of real-world situations and subsequently hone the MAXCMAS algorithms.
Rolls-Royce Future Technologies Group’s Eshan Rajabally, who led the project, said, “Through MAXCMAS, we have demonstrated autonomous collision avoidance that is indistinguishable from good seafarer behavior and we’ve confirmed this by having WMA instructors assess MAXCMAS exactly as they would assess the human.”
During the development project, Rolls-Royce and its partners adapted a commercial-specification bridge simulator as a test bed for autonomous navigation. This was also used to validate autonomous seafarer-like collision avoidance in likely real-world scenarios.
Various simulator-based scenarios were designed, with the algorithms installed in one of WMA’s conventional bridge simulators. This also included Atlas Elektronik’s ARCIMS mission manager “autonomy engine," Queen’s University Belfast’s collision avoidance algorithms and a Rolls-Royce interface.
During sea trials aboard AEUK’s ARCIMS unmanned surface vessel (USV), collision avoidance was successfully demonstrated in a real environment under true platform motion, sensor performance and environmental conditions.
“The trials showed that an unmanned vessel is capable of making a collision avoidance judgment call even when the give-way vessel isn’t taking appropriate action,” said Ralph Dodds, Innovation & Autonomous Systems program manager, AEUK. “What MAXCMAS does is make the collision avoidance regulations applicable to the unmanned ship.”
The MAXCMAS project is now complete, delivering state-of-the-art regulation compliant collision avoidance. The technology and system has been thoroughly tested both at sea and under a multitude of scenarios using desktop and bridge simulators, to demonstrate its robustness, and prove that autonomous navigation can meet existing COLREGS requirements.Edit Module