Study On State Estimation Of Discrete Memristive Neural Networks With Network-induced Phenomena

In this paper,we mainly discuss the state estimation problem for a class of discrete-time memristive neural networks(DMNNs)with network-induced phenomena including network-induced time delay and packet dropouts.Two different descriptions are adopted in this paper for the network-induced packet dropouts.The multiple partial missing measurement phenomenon is characterized by a set of independent random variables with a specific probability distribution over the[0,1]interval,which is so-called stochastic model;Some appropriate norm-bounded uncertainties is given to account for another type of packet dropouts,which is so-called nonstochastic model.Different from the traditional neural network model,the system we study is a kind of DMNNs model because of the special physical electronic characteristics of memristors,and it is a kind of special state-based switching system.In the process of research,on one hand,we try to extend some existing network packet dropouts models to a kind of DMNNs,and then design a state estimator for a class of DMNNs,which serves to guarantee the stability of the system.Furthermore,by introducing theH_?performance index,we give a description for the disturbance attention level of the designed state estimator,quantitatively.Finally,by solving a class of convex optimization problems,we obtain the sufficient condition for the existence of the state estimator satisfying the expected performance index.At the same time,the conclusion is presented in the form of linear matrix inequality(LMI).On the other hand,the event-triggered based set-membership state estimator could effectively reduce the network communication burden in the band-limited environment,and gives the upper bound of the state estimation set satisfying the requirement.We verify the validity of the designed state estimation algorithm with several numerical simulations.Specifically,the framework of this article generally includes the following chapters:First of all,according to the different actual occurrence forms of packet dropouts in the network,the following two cases should be considered,such as the corresponding problems of random packet dropouts and the non-random case.Meanwhile,the corresponding system models are constructed,then the relevant state compensation method is designed.Secondly,with respect to the two types of network packet dropouts phenomena,the corresponding state estimation algorithm for DMNNs is proposed.According to the theory of set-membership state estimation,the state estimation of the DMNNs is completed use the semidefinite programming method.Thus the optimal state estimation ellipse can be obtained.Furthermore,in actual application process,because of the limited network bandwidth and sensor energy,several changes might occur in model parameters and system structure,which will lead to network-induced phenomena.Therefore,the event-triggered communication mechanism is adopted in given chapter,so as to reduce the communication burden of MNNs and diminish the network-induced phenomena.Finally,combined with the convex optimization method,the correctness of the theoretical analysis as well as the practicability of the proposed method are both verified in the given simulation examples.This paper involves some theories such as linear system theory,Lyapunov stability theory and convex optimization theory.Several common methods used for calculation can be outlined as:Lyapunov function method,Kronecker product,LMI and the expectation of random variable and so on.And the main solving tools include standard Matlab Software,the LMI toolbox as well as YALMIP toolbox,etc.