Online nonparametric identification modeling of ship maneuvering motion based on PSO-optimized incremental Gaussian mixture model

Autonomous navigation of intelligent ships is highly dependent on real-time dynamic modeling technology to cope with the inherent nonlinear and time-varying characteristics of the marine environment. In this paper, an online identification method based on particle swarm optimization(PSO) to optimize the incremental Gaussian mixture model(IGMM) is proposed. This method solves the dilemma of manually tuning by deeply analyzing the significant impact of IGMM hyperparameters on model high fidelity and introducing the PSO algorithm to optimize the IGMM hyperparameters. Under the complex conditions of random steering tests and the introduction of wind and wave interference, the IGMM still shows high prediction high fidelity. To verify the performance of the proposed method, the IGMM optimized by PSO is compared with support vector regression(SVR), random forest regression(RFR) and noise input Gaussian process(NIGP) algorithms based on the free navigation test data of the SR108 container ship. The experimental results show that the IGMM optimized by PSO significantly reduces the root mean square error(RMSE) compared with the traditional algorithm in the free navigation test, and can dynamically adjust the number of model components to ensure robustness.