Design And Performance Analysis Of Characteristic Function Filter For A Class Of Nonlinear System

State estimation have received more and more attention in the fields of target tracking,parameter identification,intelligent driving,and communication navigation.The main methods are: linear Kalman filter,nonlinear extended Kalman filter,unscented Kalman filter and strong tracking filter,non-Gaussian particle filter and so on.Aiming at the bottleneck problem encountered by the above methods in the face of strong nonlinear systems,the characteristic function filtering developed in recent years has achieved satisfactory results in the state estimation problem of strong nonlinear measurement systems.However,in the face of actual systems where the state model and the measurement model are both nonlinear,the existing methods still lack a better solution.To this end,in response to the above-mentioned problems,this paper defines the nonlinear term in the state model as the hidden variable of the system,and combines the newly designed filter performance index to carry out the application research of the newly-built filter to establish a class of nonlinear system characteristic function filter design and performance analysis method.The main research work of this dissertation is as follows:(1)Design of characteristic function filter with hidden variables.First,by introducing the hidden variable of the system state,the nonlinear state model is linearized modeling;Secondly,construct a new performance index with characteristic function as the optimization target,and design a new characteristic function filter;Finally,the nonlinear model approximation ability is used as an indicator to compare the effect of the truncation error in the EKF on the filtering performance,and verify the high accuracy of the built new characteristic function filter.(2)Design of a new characteristic function filter fusion method.First,in the case of multi-sensor measurement transmission without delay,respectively design centralized fusion and efficient parallel fusion methods;Secondly,in the case of multi-sensor communication with time delay and packet loss,a sequential fusion filtering method suitable for the continuous arrival of measurement information is designed.Finally,based on the distributed multi-sensor wireless communication environment,the performance of the three forms of fusion methods is analyzed and studied by experiment simulation.(3)A real-time identification method for parameters of strongly nonlinear input and output systems.First,regarding strongly nonlinear system as a measurement model with the parameters to be identified as the state variables.Secondly,carry out random dynamic modeling of the system state,and establish a system state estimation method based on a new characteristic function filter;Finally,take the deep neural network model parameter identification as the object to verify the powerful application ability of the new filter.