On December 9, NYK Group companies MTI Co Ltd and Japan Marine Science Inc conducted a demonstration test of collision avoidance navigation using artificial intelligence in Osaka Bay as part of a research program to develop navigation support for domestic vessels by using AI as core technology.* The research is currently conducted in collaboration with Kobe University and Osaka Prefecture University.Outline of demonstration testIn this R&D project, we aim for to develop an automatic collision avoidance system that determines safe and economical course selection in various circumstances by applying the advanced AI method, deep reinforcement learning, a machine-learning method in which an autonomous program performs a vast number of voyage simulations to learn optimal collision avoidance strategy. The AI navigation support system that we are developing uses radar and Automatic Identification System sensor information to recognize collision risks in ship navigation and automatically selects the optimal course. The selected course is set as the target course of the autopilot for maneuvering.After quantitative evaluations and evaluations by experienced operators of the AI navigation support system by using JMS's ship-handling simulator, the demonstration test was performed by connecting the AI navigation support system with the ship-maneuvering system of the Kobe University training ship Fukae Maru. The collision avoidance behaviors against other ships and obstacles during navigation in Osaka Bay were confirmed and evaluated.
On December 9, NYK Group companies MTI Co Ltd and Japan Marine Science Inc conducted a demonstration test of collision avoidance navigation using artificial intelligence in Osaka Bay as part of a research program to develop navigation support for domestic vessels by using AI as core technology.* The research is currently conducted in collaboration with Kobe University and Osaka Prefecture University.Outline of demonstration testIn this R&D project, we aim for to develop an automatic collision avoidance system that determines safe and economical course selection in various circumstances by applying the advanced AI method, deep reinforcement learning, a machine-learning method in which an autonomous program performs a vast number of voyage simulations to learn optimal collision avoidance strategy. The AI navigation support system that we are developing uses radar and Automatic Identification System sensor information to recognize collision risks in ship navigation and automatically selects the optimal course. The selected course is set as the target course of the autopilot for maneuvering.After quantitative evaluations and evaluations by experienced operators of the AI navigation support system by using JMS's ship-handling simulator, the demonstration test was performed by connecting the AI navigation support system with the ship-maneuvering system of the Kobe University training ship Fukae Maru. The collision avoidance behaviors against other ships and obstacles during navigation in Osaka Bay were confirmed and evaluated.
