Research Of Filtering Method For One Class Of No-Linear System With No-Gaussian Noise

Filtering technology is one of the hottest problems in the academic field and engineering field.Most of existing filter methods are designed for systems with Gaussian noise,and works of the linear systems are particularly conspicuous.However,the researches for non-Gaussian systems are less because of the complexity of noise,especially for systems with nonlinear state models and nonlinear measurement models.In practice,most of systems are nonlinear and non-Gaussian.For this reason,this paper will design one class of filter method for non-Gaussian systems with nonlinear state models and measurement models.This thesis mainly consists of the following parts:(1)Analyze the filter method based on characteristic function.Firstly,for the existing filter method,although designed for multi-dimensional measurement models,we obtain the result that it just fit to one-dimensional measurement models.Secondly,analyze the reason why the performance index and the design process of the filter do not fit to multi-dimensional measurement models.Finally,analyze the reason why the filter method can't fit to nonlinear state models.(2)For one class of non-Gaussian dynamical systems with multidimensional state equations and measurement models,we will build a class of filter based on characteristic functions.Firstly,matrix format of performance index will be designed to fit to the multidimensional measurements and constraint conditions of the weighting function will be given to ensure the index's uniformly boundness.Subsequently,the solving method of the filter gain matrix will be given.Finally,simulation examples will be made to illustrate the effectiveness of the proposed filter design scheme.Meanwhile,for the background of industrial electronics,we model the measurement system of the grinding process cylindrical workpieces and illustrate the effective of the designed filter.(3)For the non-Gaussian systems with multidimensional nonlinear state models and measurement models,we will develop the work of the filter method based on characteristic function.Firstly,based on the linearization method of state model,the recursive expression of the estimation error and its characteristic function will be obtained.Then,based on the linearized state model,the performance index and corresponding filter method will be redesigned.Finally,simulation examples will be made to illustrate the effectiveness of the proposed filter design scheme.