Technical papers from our work — each with an on-page abstract and a downloadable PDF.
A ship motion prediction model that combines physics-based equations with data-driven parameter optimization — keeping the interpretability of hydrodynamic models while capturing ship-specific behaviour. Validated on two container ships, with predictions over 50% more accurate than traditionally tuned physics-based baselines.
Two data-driven routes to vessel trajectory prediction go head-to-head: optimising the parameters of a physics-based hydrodynamic model versus a pure feed-forward neural network. Trained and 5-fold cross-validated on 90 simulated trajectories, the physics-based-with-optimisation approach wins — roughly 40% lower summarised error and 35% tighter consistency than the black-box model.
Physics-based ship motion models are rarely validated against real-world voyages. This study tailors a physics-based 3D dynamics model to a container ship and evaluates its sub-second predictions against real vessel trajectories — visually and with multiple distance measures.
Fuel consumption models often demand data operators simply don't have. This peer-reviewed study builds main-engine FOC forecasting models for a VLCC using only sensors and simple weather data that are routinely available — with XGBoost predicting within 5% of the true value in more than 86% of cases.
A parametric design tool for fast, robust optimization of SWATH (Small Waterplane Area Twin Hull) ships, developed at NTUA's Ship Design Laboratory and validated against the Aegean Queen ferry design — optimizing the hull form for two different seasonal operating speeds.