Fujitsu has announced the results of a field trial conducted with the Maritime and Port Authority of Singapore (MPA) which employed Artificial Intelligence (AI) to analyse marine traffic risks in the Singapore Strait, to predict potential collisions before they happen and increase the lead time in advising vessels on avoidance measures.
The ‘Fujitsu Human Centric AI Zinrai’ system detects ship collision risks based on data collected from the port waterways and predicts areas where those risks are concentrated as ‘dynamic risk hotspots’, the company says.
Fujitsu aims to market the system for use within a Vessel Traffic Service (VTS) infrastructure, to assist maritime traffic controllers to proactively manage the movement of vessels near shore. Commercial services are expected to commence from 2020.
The Japanese company says it will present the outcomes of its research and testing in Singapore at the IALA e-Navigation Information Services and Communications (ENAV) 23rd Committee Meeting in Singapore this week, to introduce the concept to a wider audience.
VTS centres that manage marine traffic commonly rely on anti-collision warnings using closest point of approach techniques, detecting potential problems and notifying ships when they get unusually close to one another.
Fujitsu says that the addition of AI analytics will help to improve management of these collision risks even further, particularly as vessel operations and interactions become more complex in high density vessel traffic areas.
Testing programme
The research and testing programme in Singapore was conducted with the assistance of MPA officers from the Vessel Traffic Management Department and Port Systems Division over the course of 24 months.
Drawing on past traffic data for the Singapore Strait provided by MPA, Fujitsu applied its collision risk prediction technology to predict collisions or near misses involving multiple vessels, as well as examples of developing dynamic risk hotspots.
The prediction technology was then benchmarked against human operators, where Fujitsu says that its system demonstrated an ability to quantify risk in more detail and more quickly than it could be detected by human operators.
For example, during the benchmarking studies the detection technology was able to flag potential risks approximately 10 minutes before the near-miss, and in doing so could theoretically provide approximately 5 minutes additional lead time for human operators to execute necessary follow on actions.
The technology was also able to pick out scenarios where the collision risk may have been overlooked using traditional methods, and could consequently raise an alert.
Separately, the dynamic risk hotspot detection technology showed it could detect risks up to 15 minutes in advance, enabling staff to take specific measures to avoid an incident, the company said.
Fujitsu says it will continue working with authorities in Singapore to improve the AI-based system, in advance of the full launch in 2020.