Toyo Construction Co., Ltd. (President and Representative Director: Kyoji Takezawa) has developed
the “AI forecasting system for long-term course of vessel” which enables to forecastthe course of
general vessels navigating close to the construction sea area by AIS (Automatic Identification System) data.
The AI forecasting system for long-term course of vessel is a system to forecast the course of general vessels
such as tankers or freight vessels, up to next 60 minutes. This system applies the machine learning supervised by
the previous AIS data of each port, by using the Seq2Seq (Sequence To Sequence※), a time-series forecasting model
to handle variable-length time-series data, based on the AI technology of Fujitsu Limited (CEO: Takahito Tokita) that
forecasts the future position of navigating vessel with high accuracy. During the operation, it forecasts the long-term
course of vessel using the real-time AIS data transmitted from the navigating AIS-mounted vessel as its input data.
Previously, the Company used to grasp the general vessels navigating close to the construction sea area in real time
by utilizing information of the AIS, ship radar, and GNSS (Global Navigation Satellite System), and the course up to
next 10 minutes was displayed by linear vector using the current information of vessel’s traveling direction and speed
(“Mihari-chan” (watcher in English), a navigating vessel monitoring system developed by the Company).
This development of the AI forecasting system for long-term course of vessel provides a capacity to forecast the course
up to next 60 minutes by taking into consideration the changes of vessel’s traveling direction and speed due to the
shape of harbor such as topography or shipping route. This will improve the navigation safety, not only because our
work ship will be able to ensure that measures are taken to deal with approaching general vessels at an earlier stage,
but also because the work ship’s master will be able to use the information as a judgment criterion when entering from
the construction sea area into the shipping route where general vessels are drifting.
We will expand this system nationwide by having the system learn AIS data of main ports as well as ship radar information,
which will lead to the development of a system to forecast the course of all ships including non AIS-mounted vessel,
and also will continue to improve safety in marine construction.
※Sequence To Sequence is an AI model developed by Google, modeling the probability of conversion from a series
to another series.
Route forecast by AI forecasting system for long-term course of vessel
Previous forecast by the navigating vessel monitoring system
Conceptual illustration of route forecasting