Objectives To construct an accurate MMG （mathematical model group） model for a water-jet propulsion unmanned surface vehicle, the traditional extended Kalman filter algorithm and improved extended Kalman filter algorithm are combined with the real-world boat data for parameter identification.
Methods First, based on the traditional EKF algorithm, in order to fully utilize the valuable information hidden in the historical data, an improved EKF algorithm integrating multi-innovation theory and dynamic forgetting factor is proposed. Then, using the real-world unmanned surface vehicle data, the unknown parameters in the MMG model are identified. Finally, the identified parameter values are substituted into the established MMG model, and the rudder angle and main engine speed consistent with the real boat data are input. The heading angle, longitudinal velocity, transverse velocity, heading angle rate and position information data are obtained through simulation, and the comparative analysis is carried out.
Results The results indicate that compared with the traditional EKF algorithm, the root mean squared error index and the symmetric mean absolute percentage error index of the improved EKF algorithm are closer to 0. Specifically, the root mean squared error index is reduced by up to 20.02% at the highest, and the symmetric mean absolute percentage error index is reduced by 26.84% at the highest.
Conclusions The simulation results demonstrate that the improved extended Kalman filter algorithm has higher identification accuracy, verifying the accuracy of the MMG model established by the algorithm.
DownLoad: