Model Predictive Control (MPC) is applied to track control of unmanned surface vessels. To adapt to dynamic changes in speed and water depth, a water depth and speed dependent predictive model, with coefficients identified from free running tests (e.g., acceleration and zigzag trials), is developed. The nonlinear least square method is used for parameter estimation of the model. Furthermore, a novel self-tuning method adapts MPC’s essential parameter, prediction horizon, to diverse operating conditions. Validation via manoeuvring simulations of the full-scale benchmark KVLCC2 ship covers under keel clearances (UKC) from 20 to 100 % ship’s draft (from shallow water to medium deep water) and vessel speeds of 7-15 knots, spanning 25 distinct scenarios. The designed controller achieves high-accuracy path following across all conditions, demonstrating exceptional adaptability to parameter variations. |