Adaptive Anti-noise Least-Squares Algorithm for Parameter Identification of Unmanned Marine Vehicles: Theory, Simulation, and Experiment

In this paper, an adaptive anti-noise least-squares algorithm (ANLS) is proposed for parameter identification of an unmanned marine vehicle in the presence of measurement noise. As a basis, a horizontal-plane second-order nonlinear Nomoto model is established and transformed into a discrete-time model for parameter identification. Then, a noise reduction term is added to the loss function to achieve a trade-off between the anti-noise effect and parameter identification accuracy. Furthermore, the Levenberg–Marquardt algorithm is embedded into the parameter identification algorithm to achieve adaptive coefficient optimization. Finally, the simulation and experimental data are utilized for parameter identification and performance validation. By comparing with the recursive least-squares algorithm and least-squares support vector machine algorithm, the excellent anti-noise and maneuvering prediction abilities of the proposed ANLS algorithm are verified, i.e., up to 84% reduction of the identification error in the simulation and less than 4 ∘ of the heading angle prediction error in the experiment.