Research On Ship Nonlinear State Estimation Method Based On Particle Filter

Along with the gradual development of world economy and the shortage of land resources,Marine resources development and utilization of the earth has become the focus of the economic development of all countries.So it is high request to the technical equipment of ship equip in resource development and assignment engineer projects.Ships need to equip with Dynamic Positioning System(DP Systems)when sailing on the sea according to the trajectory or keeping the desired position.Positioning accuracy of DP system depends on whether ship state estimator can effectively estimate position,heading angle and speed of the ship.In this article,the object of study is the low speed ship equip with Dynamic Positioning System.In order to improve the positioning precision and the overall performance of positioning system,aiming at the case of noise statistical properties change,we designed an nonlinear state estimator based on improved particle filter to estimate the ship status.Firstly,the mathematical model of ship motion is analyzed and established,introduces common ship motion reference frame,and then the movement can be seen as low-frequency and high-frequency motion movement superimposed,so as to build high frequency and low frequency motion model of ship respectively,meanwhile the system measurement model is built.In order to describe the real motion of ship more accurately,according to the characteristics of the external environment,the three kinds of the Marine environment disturbance force model — wind model,wave model and current model are developed respectively.Then,according the theory of Bayesian framework we research on nonlinear filtering methods,introduced commonly nonlinear Gaussian filtering methods: Extended Kalman Filter(EKF)and Unscented Kalman Filter(UKF),which based on Kalman Filter framework,and give advantages and disadvantages of two kinds of filtering method respectively.In nonlinear non-Gaussian situation,the standard Particle Filter is introduced.And in order to improve the accuracy of Particle Filter,in view of the problem of particle degen-erace,Unscented Kalman Filter is used to generate the importance density function and proposed an Unscented Gaussian Particle Filter(UGPF)in this paper,meanwhile its convergence is proved and analyzed.Next,UGPF method is used for the design of state estimator,due to the process of establishing the model equation is stochastic differential equations,in order to applied to the discrete filter model,It? 1.5 order Taylor expansion is utilized and process equation can beconverted to random discrete equation.In order to improve the filtering accuracy and considering the effects of non-Gaussian noise,using statistical characteristics of the unknown and time-varying Gaussian distribution to fitting of non-Gaussian noise,basing on the idea of marginalized Particle Filter,the unknown parameters of noise can be estimate.Considering the rapidity problem of the Particle Filter algorithm as well,introducing the KLD online adjust the sampling population,eventually proposed RB-AUGPF(Rao-Blackwellization-Automatic Unscented Gaussian Particle Filter)state estimator design method of the ship.Finally,in order to verify the performance of the design of nonlinear state estimation based on the established ship model,the simulation is carried out to verify the effectiveness of the proposed nonlinear state estimator.Under the condition of Gaussian noise to simulate the UGPF estimator,and simulate RB-AUGPF estimator in case to join measurement outliers noise,which both choose the good performance SRCKF(Square Root Cubature Kalman Filter)estimator for comparison.The simulation results prove that the designed nonlinear state estimator can effectively estimate the ship’s position,yaw angle and speed information,RB-AUGPF estimator has good robustness to measurement outliers.