NAVISP EL2-191: “ARRIVAL-X – AI-Driven Vessel Arrival Prediction for Ports”

Last Updated: 03/06/2025 08:53     Created at: 03/06/2025 08:51

On 15 May 2025, Space-Ship World unveiled the results of ARRIVAL-X, a NAVISP Element 2 project aimed at cutting port congestion by providing accurate, last-mile estimated times of arrival (ETAs) through a standalone web application that needs no extra hardware on the quayside. 

Key take-aways of the presentation were:

• Multi-layer AI engine

A three-stage approach (statistical look-up, gradient-boosted models, deep neural network) delivers the best ETA accuracy from 48 h down to the crucial final 6 h before berth arrival. 

• Rich data fusion

Live and historical AIS feeds are blended with ERA-5 Copernicus weather data; global models give rapid deployment, while optional local retraining sharpens last-mile predictions. 

• Pilot success

A four-week trial at Spain’s Port of Algeciras showed mean ETA errors well within target in the 0-6 h window, validating the approach for busy container hubs. 

• Business impact

Accurate ETAs cut idle time, curb emissions, and reduce costs linked to late arrivals—addressing an industry where 67 % of ships now arrive behind schedule. 

Next steps include UX refinements, scalable cloud back-end, an open API for port-community systems, and new pilots scheduled for June 2025.

The company will incorporate new data sources such as from VDES-R when it becomes operational in the areas of interest.

This project was co-funded through NAVISP Element 2. The slides are available here.