Distributed Estimation Of Nonlinear Systems Based On Communication Protocol

The sensor network technology has been widely applied in the fields of military and civil-ian due to its great characteristics(e.g.low cost and easy installment).Considering the limited power and resources of a single sensor node on the sensor networks,the distributed structure is adopted when it comes to the state estimation issue,where the sensor networks deal with the in-formation of itself and neighboring nodes on the basis of the sensor topology.The utilization of the distributed structure can effectively reduce the communication cost between sensor network nodes.In addition,the network communication protocols can effectively alleviate the network congestion and avoid data collision.Therefore,it is of great theoretical significance and appli-cation value to study and discuss the distributed estimation problem of nonlinear systems based on network communication protocols.This thesis aims to investigate the distributed estimation problem of nonlinear systems based on network communication protocols.Many interference phenomena that may occur in practical applications are taken into account,such as network phenomena,quantitative effects and the one-sided Lipschitz nonlinearity in the network systems.At the same time,this thesis analyzes and discusses some relatively novel models and methods,such as characterizing a model to describe the randomly varying topology,presenting a distributed filter model with the partial nodes information,and adjusting the scheduling rules of Round-Robin protocols to adapt to partial nodes which can not obtain measurement information.The main contents of this thesis include the following parts:The first part is the introduction,which mainly introduces the background,significance and research status of the distributed estimation based on sensor networks,influences of net-worked phenomena and quantization effects,network communication protocols,partial nodes information method and randomly varying topology.The second part investigates the distributed filtering problem for a class of Markov jump systems based on the Round-Robin protocol and random nonlinearity.The nonlinearity dis-turbance satisfies the one-sided Lipschitz constraint,which is more general than the Lipschitz condition.Based on the Lyapunov function method,the exponentially ultimate boundedness in mean-square and the H_?performance analysis is considered for the bounded noise and square additive noise,respectively.Then,the corresponding filter is designed by solving the linear matrix inequalities.The distributed H_?filtering problem for a class of nonlinear sensor network systems based on the event-triggered protocol with randomly varying topology and missing measurements is considered in the third part.Firstly,a stochastic topological model is presented,a sequence of random variables is used to characterize the phenomenon of randomly varying connection strength between adjacent nodes.Secondly,the H_?performance analysis of augmented error system is investigated by using Lyapunov function analysis method.Then,it follows from the matrix decomposition method that the direct coupling between the statistical information of interconnection strengths and gains of filters are removed.Finally,the distributed H_?filter based on the event-triggered protocol is designed.The fourth part focuses on the distributed estimation issue of a class of nonlinear stochastic systems with the Round-Robin protocol and partial nodes information.Firstly,the scheduling rule of the Round-Robin protocol is adjusted to accommodate the available sensor nodes.Sec-ondly,based on the Lyapunov function analysis method,the exponentially mean-square ulti-mately bounded analysis results of the augmented error system with the Round-Robin protocol and partial nodes information are given.Finally,the distributed exponential estimator is de-signed by solving the linear matrix inequalities.Based on the event-triggered protocol,the fifth part is concerned with the distributed H_?estimation problem for a class of nonlinear stochastic systems with the Markov switching topol-ogy and partial nodes information.The influences of random delays and random nonlinearity disturbance to the distributed estimation are considered.Based on the event-triggered mech-anism and Markov switching topology rules,the H_?performance analysis of the augmented error system is investigated by using the Lyapunov function method.Then,the design method of distributed H_?estimator is given by solving linear matrix inequalities,and the augmented error system is exponentially mean-square stable satisfying the H_?performance constraint.The sixth part summarizes the main results of this thesis and makes a further prospect to the future research work,respectively.