Research On Ship Motion Prediction Based On Robust Fusion Kalman Filter

Ship motion and attitude prediction is an important branch in the field of ship research.If the rolling angle of ship motion can be predicted a few seconds in advance,it will be of great significance to the design of anti-rolling device and the improvement of anti-rolling effect.The prediction of ship motion based on time series does not need to know the prior information of wave or to establish the specific kinematic equation of ship.It is only necessary to use the historical data of the motion attitude of the ship to find the law for modeling and prediction.Kalman filtering algorithm has been applied to the parameter estimation of the ship motion AR prediction model because of its recurrence relationship and fast convergence speed.But when the system has a In deterministic,AR modeling prediction based on classical Kalman filtering may deteriorate or even lead to divergence.This leads to the problem of AR modeling and prediction of robust fusion Kalman filter.There are two significant advantages over classical Kalman filtering:1)enhanced robustness of the system to uncertainty 2)smaller mean square error and higher prediction accuracy.In this paper,the robust fusion Kalman filter of discrete time-invariant systems with various uncertainties is studied and applied to ship motion AR modeling and prediction.First of all,for linear discrete multisensor systems(system 1)with uncertain noise variance and linear correlation of white noise,and linear discrete multisensor systems(system 2)with state dependent multiplicative noise and uncertain noise variance,The virtual noise method is used for model transformation,and then based on the minimax robust estimation principle,the conservative state fusion Kalman filter is derived according to the matrix weighting criterion.The model is applied to the recurrence estimation of the parameters of AR model,and the ship motion model is established,and the prediction effect of ship rolling motion in the next 10 seconds is simulated and analyzed.Secondly,for linear discrete sensor systems(system 3)with the same state dependent noise and uncertain noise variance and white noise linear correlation,the virtual noise method is used to transform the model,and then based on the minimax robust estimation principle,The conservative state fusion Kalman filter is derived according to the matrix weighting criterion,and it is proved that the actual state fusion Kalman filter is robust.The model isapplied to the recurrence estimation of the parameters of AR model,and the ship motion model is established,and the prediction effect of ship rolling motion in the next 10 seconds is simulated and analyzed.And the robust fusion K of system 2 based on system 3 and system 1is compared.The prediction effect of AR prediction model is established by alman filtering method.Finally,for linear discrete multisensor systems(system 4)with lost observations and uncertain noise variances,and linear discrete multisensor systems(system 5)with lost observations,multiplicative noise and uncertain noise variances,respectively,The virtual noise method is used for model transformation,and then based on the minimax robust estimation principle,the conservative state fusion Kalman filter is derived according to the matrix weighting criterion,and it is proved that the actual state fusion Kalman filter is robust.The model is applied to the recurrence estimation of the parameters of AR model,and the ship motion model is established.The prediction of ship rolling motion in the next 10 seconds is simulated and analyzed.Effect.The prediction effect of AR prediction model based on system 5 and system 4 robust fusion Kalman filtering method is compared.