Research On Dynamic Positioning Particle Filter Method Based On EnKF

Dynamic Positioning System(DP System)is an important class of ship closed-loop control systems.During sailing,the ship can't avoid the effect of wind,waves,streams and other environmental interference.The working principle of the dynamic positioning system is to reduce the interference of the environment on the ship by the thrust of the ship propeller so that the ship could be fixed at the desired position or sailing along the set path.The positioning accuracy of the DP system depends mainly on whether the ship state observer could effectively filter out the noise of the ship's state information and estimate the state of the ship's various degrees of freedom including position,bow and speed.In this paper,we take the low-speed navigation of the dynamic positioning ship as the object of study.The realization of state estimation in dynamic positioning is actually based on the stochastic dynamic system of the state estimation problem of nonlinear filtering theory.In the case of noise statistics changing,we design a dynamic positioning ship's nonlinear state observer based on Ensemble Kalman Particle Filter(EnKPF)to estimate the state of the dynamic positioning ship running at low speed.It could improve the positioning accuracy of DP and overall performance of the DP system.The specific discussion is as follows:Firstly,the mathematical model of the ship dynamic positioning is established based on the movement of the ship in the actual environment.Ship movement could be considered as a mixture of high frequency motion and low frequency motion.The high frequency motion and low frequency motion are respectively established and studied.Modeling the ocean environment,the actual environment is an important factor affecting the accuracy of the ship's dynamic positioning,including wind,waves and streams.Finally,we use Matlab to verify the established model above.Secondly,this paper introduced several common methods used in nonlinear filtering theory in recent years including recursive Bayesian Filter,Extended Kalman Filter,Ensemble Kalman Filter and Particle Filter.Then,combined with the Ensemble Kalman filter and particle filter,the Ensemble Kalman Particle Filter algorithm EnKPF is proposed.Thirdly,in order to prove the effectiveness of EnKPF in the dynamic positioning system,we designed a state observer based on EnKPF.As a result of the output signal obtained by the device measurement is discontinuous in the dynamic positioning system,it is necessary to establish a continuous-discrete dynamic positioning state space model.When the noise is common and non-Gaussian,the ship nonlinear observer is designed with EnKPF.When the noise is unknown time-varying non-Gaussian noise,we introduce the idea of marginalization,and enhance the robustness of the algorithm.Then,the idea of particle swarm optimization is introduced to improve the particle in the algorithm,and the ship nonlinear observer based on PSO-AEnKPF is formed.Fourthly,simulate and analyze the nonlinear state observers above.Then we added the wild value of the measurement to verify its ability of measuring the anti-interference ability.Based on the simulation,we prove the validity of the proposed methods based on EnKF dynamic positioning particle filter.