25. mai 2026
Completed 👉🏼 The deliverable D2.3 report, which is related to the task T2.3 Digital twin simulations of voyage performance (M6-M30) [Lead: NVT, Contributor: SRL]
The GASS (Green AI for Sustainable Shipping) project addresses the challenge of inefficient voyage planning and execution by applying AI-powered digital twins to optimise vessel performance and enable smarter, greener operations.
At the core of this work is the development of digital twin models of vessels, combined with advanced weather routing algorithms.
Together, these form the basis of a future Weather Routing Optimisation Service (WROS)—one of the first commercial outcomes of the project.
Evaluating Vessel Models for Real-World Use
The report evaluates four different modelling approaches for predicting vessel performance:
Heuristic models based on expert rules
Physics-assisted machine learning (KAN)
Parametric models
NAVTOR’s data-driven noon report model (v3)
Each model is assessed based on data requirements, accuracy, robustness, and suitability for integration into commercial routing solutions.
The findings show that NAVTOR’s noon-report-driven model (v3) is currently the most practical option for deployment. Its main advantage is that it relies on widely available operational data, unlike sensor-based approaches that require more advanced infrastructure. That said, a steady feed from sensor data will provide the most detailed and updated Ship Model for the actual MetOc conditions. And ideally, as also researched in GASS, it is possible to combine sensor data AND NoonReport and also other input (e.g. from Logbooks) to have the best possible data foundation.
Key Challenge: Performance in Extreme Conditions
While the v3 model performs well under normal conditions, the study highlights a critical limitation: reduced reliability in out-of-distribution scenarios, such as extreme weather.
For example, simulations show that the model can produce non-physical predictions—such as increased vessel speed in high waves—when operating outside its training data range.
Path Forward
Despite these challenges, the results confirm that digital twin-based models have strong potential to transform voyage optimisation. The v3 model is already suitable for weather avoidance and routing in typical conditions, but further work is needed to improve robustness in extreme scenarios and expand validation across vessel types.
Next steps focus on strengthening model reliability, enhancing system integration, and enabling full end-to-end validation of routing performance.
Enabling Sustainable Shipping
By combining AI, digital twins, and real-world vessel data, the GASS project is laying the foundation for more efficient, data-driven shipping operations.
The ultimate goal is clear: reduce emissions, improve profitability, and support the industry’s transition to greener maritime transport.
The GASS-project is made possible by The Green Platform Initiative (“Grønn Plattform”) funding scheme by Norges forskningsråd, Innovation Norway and Siva SF under the Grant Agreement No. 346603.