Multialgorithm Fusion Method for State Estimation of USV Based on GS-STSCKF and Credibility Theory Coupling

To address the issue of low accuracy in unmanned surface vehicle motion state estimation under complex and harsh water surface conditions, a state estimation method based on the coupling of Gaussian sum strong tracking square-root cubature Kalman filter (GS-STSCKF) and credibility theory is proposed. Firstly, building upon the traditional STSCKF algorithm, a multiple fading factors matrix is introduced and directly applied to the process noise covariance matrix, achieving dynamic fine-tuning and enhancing model interpretability. Subsequently, an improved Gaussian sum framework based on a Gaussian component pruning, merging, and capping strategy is adopted to significantly enhance the algorithm's capability to handle non-Gaussian disturbances. To overcome the deficiency of existing methods in real-time adjustment of observation noise covariance, credibility theory is introduced from the perspective of the consistency between the filter mean square error and the true mean square error. Trust factors and discrimination conditions are constructed to achieve accurate online estimation of observation noise. Furthermore, the coupling mechanism between strong tracking theory and credibility theory is analyzed, forming a closed-loop optimization iterative process. Finally, a broad learning system optimized by an improved genetic algorithm and a metaheuristic strategy is leveraged to compensate for linear approximation errors caused by pseudo-matrix construction. Simulation and field data tests demonstrate the superiority of the proposed algorithm in terms of estimation accuracy, robustness, and adaptability compared to existing methods.