Research On Nonlinear State Estimation And Fusion Algorithm For Dynamic Positioning Ships

The nonlinear state estimation of the Dynamic Positioning(DP)ships plays an important role in accurate ships operation.However,the DP ships state estimation process mainly faces the following disturbances.At first,as the electromagnetic interference,thermal noise and cosmic noise exist universally,which are additive to the state,such interference is called additive noise interference.Because recursive relations exist among states,additive state noise has the feature of self-correlation.The measured value of the sensor is obtained according to the system state,therefore the additive measurement noise also has the feature of self-correlation.Moreover,the sensor and DP ships work in the same noise environment,so cross-correlation feature exists between additive measurement noise and additive state noise.Secondly,the exact values of the system model parameters and the statistical characteristics of noise interference are difficult to obtain,and the measured values lose randomly due to the communication failure and electromagnetic interference.This kind of interference is called the parameter uncertainty interference.Thirdly,because of the un-modeled part such as the change of sea condition and the time varying characteristic of the system in the complex ocean environment,the interference with change of signal appears in the system.These interferences are called multiplicative noise interference.Any interference above will affect the accuracy of DP ships state estimation.Furthermore,DP ships are usually equipped with redundant sensor units and the estimated fusion value obtained by the multi-sensor is more accurate than estimated values obtained by single sensor.Accordingly,in order to improve the state estimation accuracy in the presence of disturbances,this thesis develops a nonlinear state estimation and fusion algorithm for DP ships.The main work of the thesis is as follows:Firstly,aiming at the interference including the additive cross-correlation noise,parameter uncertainty and multiplicative noise in the process of state estimation for DP ships,and based on the continuous time state space model of DP ships,the discrete time nonlinear model of DP ships with the additive cross-correlation noise,multiplicative noise and random measurement loss is established respectively in the case of no correlations among the above mentioned interferences in this thesis.Secondly,in view of the fact that the additive measurement noise and additive state noise of DP ships have the property of one step autocorrelation and two step cross-correlation,the state estimation method of single degree of freedom(DOF)direct route in cross correlated noise and the nonlinear state estimation method of 3-DOF motion in the related noise are studied respectively.On one hand,the problem of the estimation of yaw angle and turn rate of ships is settled via the linear state estimation method with the related noise for the direct route of DP ships in single DOF.On the other hand,the Bayesian estimation with the related noise is obtained for the 3-DOF motion of DP ships.And a cubature Kalman filter with cross-correlation noise is designed based on the cubature rules to solve the problem of state estimation for the 3-DOF motion of DP ships.Thirdly,according to the problem of parameters uncertainty for DP ships,the nonlinear state estimation algorithm is studied under the condition of parameters uncertainty of system model,then the cubature smoothing variable structure filter and square root cubature smoothing variable structure filter are proposed.The cubature rule improves the estimation accuracy,and the square root form guarantees the stability of the algorithm.Aiming at the unknown statistical characteristics of additive measurement noise for DP ships,a variational Bayesian variable structure filter is proposed.The variational Bayesian achieves the estimation of noise statistical properties.The accuracy of the smoothing boundary layer of the variable structure filter is improved,and the nonlinear state estimation problem of DP ships with unknown statistical characteristics of the additive noise is solved.Aiming at the random measurement loss and additive colored noise interference of DP ships,a Gauss filtering framework with colored noise and random measurement loss is designed and a cubature mixing Kalman filter based on cubature rule is proposed so that the nonlinear state estimation problem of DP ships is solved in the presence of random loss and additive colored noise.Finally,in the light of the interference with multiplicative noise and additive correlated noise of DP ships,the nonlinear state estimation and fusion method for DP ships is studied under the condition of multi-sensor multiplicative noise.A smoothing variable structure filter with cross-correlation noise is proposed for the single sensor system.The nonlinear state estimation problem is resolved for DP ships with single sensor multiplicative noise.For multi-sensor systems,training set and test set are randomly generated from each sub-sensor's state estimation value.Then,the corresponding support vector machine for regression model is constructed for each state variable separately and the multi-sensor fusion algorithm is proposed based on multiple support vector machine for regression model.At last,the problem of nonlinear state estimation and fusion is settled in the presence of multiplicative noise.