AI System Tested by Tokyo Bay Vessel Traffic Controllers
Fujitsu Limited and the Japan Coast Guard have field tested an artificial intelligence (AI) system for predicting vessel collisions.
The trial was conducted from December 2019 to March 2020 at the Tokyo Wan Vessel Traffic Service Center, which offers navigation support services, under an outsourcing contract with the Japan Coast Guard.
According to a report by the Japan Transport Safety Board, there were 2,863 marine collision accidents in Japan between 2009 and 2019, with an average of 286 accidents per year.
Using Fujitsu Human Centric AI Zinrai developed by Fujitsu Laboratories, the technology can detect near misses between vessels and predict areas where collision risks are concentrated in Tokyo Bay.
The Japan Coast Guard operates a navigation support system that combines radar and AIS data for high-traffic routes. However, it is not easy to recognize and predict the movements of a large number of vessels in addition to detecting risks. Therefore, the detection of risks and the timing of risk recommendations to vessels depends on the experience and skill of the controllers.
Using the training environment of the Tokyo Wan Vessel Traffic Service Center, with the cooperation of six experienced maritime traffic controllers and instructors, control tasks based on past near miss accidents were simulated to demonstrate the effectiveness of the new system.
Two methods of control operation were evaluated:
1) The conventional method, where operation controllers recognized and predicted the movement of vessels based on their own experience and skills to determine danger.
2) In addition to conventional control operations, risk information was calculated with collision risk prediction technology by utilizing past AIS data and risk recommendation records from the Japan Coast Guard.
It was found that the new technology could shorten the time before a risk warning is sent to a vessel by the operation controller by an average of approximately two minutes, enabling the early detection of vessels at risk of collision.
It was also confirmed that the number of warnings issued for risk-prone vessels nearly doubled with the technology, indicating proactive preventive measures and the potential to enhance safety in Tokyo Bay.
As this technology assists marine control operations by analyzing the qualitative conditions of collision risks, it was found to be particularly effective for newcomers who could in some cases perform the same control actions as those of highly skilled controllers.
Fujitsu will continue to collaborate with the Japan Coast Guard to improve the technology.