Constrained Parameter Estimation for Ship Maneuvering Models to Increase Physical Reliability

Paper in proceeding, 2025

The reliable modelling of the dynamical behavior of ships became a fundamental task for various applications ranging from maneuverability analyses to trajectory planning and automatic control. With this work, the modelling of ship motion is improved by ship independent filtering and estimation of hidden states in a first step. On top of that, physics informed constraints are applied to the linear regression parameter estimation. The paper shows the improvement in open-loop simulations compared to the conventional unconstrained estimation. It increases interpolation and extrapolation capability of the model and prevents from overfitting. This holds also for different propulsion systems and model structures.