State Estimation For Discrete-time Memristive Neural Networks Under Dynamic Event-triggered Mechanisms

With the development of artificial intelligence technology,the research of artificial neural networks is set off again.However,due to the limitation of CMOS transistors and other devices in the physical implementation of traditional artificial neural networks(ANNs),the relevant theoretical research is also slow.It is noteworthy that the memristiv neural network(MNN)constructed by the memristor makes up for some defects of traditional ANNs due to its high integration density,low power consumption,good scalability,and memory function.The research on MNNs is increasing gradually,especially in the stability analysis of MNNs has achieved some research results.Besides,in engineering practice,now and again so as to achieve optimization,approximation,and other specific purposes,it is necessary to obtain the state information of the neural network.However,due to the constraints of the actual conditions,it is often not realistic to obtain the state information of the system directly.At this point,it is very important to estimate the state of the neural network effectively according to the measurement data.In addition,with the innovation of computer technology,the application potential of MNNs in discrete cases is greatly enhanced.In view of this,the state estimation problem of discrete MNNs based on the dynamic event-triggered protocol will be studied deeply in this paper.In this paper,a new model of MNN is established by analyzing the characteristics of memristor,noise,and time delay.Then the finite-horizon H_∞ state estimation problem for a class of discrete MNNs by dynamic event-triggered is discussed,which solves the problem that the performance of the state estimator may deteriorate when channel fading occurs.Then,based on the previous research results,the l2-l_∞state estimation problem of a class of discrete MNNs with redundant channels is studied by using Lyapunov stability and Schur methods and techniques.Similarly,the network communication resources are effectively saved due to the introduction of dynamic event-triggered protocols.The following is an introduction:1.Considering the channel fading phenomenon in network communication,the finite-horizon H_∞ state estimation problem for discrete MNNs with time delays triggered by dynamic events is studied.By constructing the Lyapunov function,a sufficient condition is proposed that the estimated error system satisfies H_∞ performance in the finite-horizon,which effectively reduces the conservatism of the system.2.Considering the limited bandwidth of communication channels,the l2-l_∞ state estimation problem of discrete-time MNNs with redundant channels under dynamic event-triggered protocol is studied.A redundant channel transmission mechanism is introduced to improve the transmission efficiency of measurement data.By using Lyapunov stability and Schur complement methods and techniques,a state estimation error system satisfying the l2-l_∞ performance index is presented.3.In this paper,the finite-horizon H_∞ state estimation conditions for discrete time-delay MNNs with fading channels is analyzed,introducing dynamic event-triggered protocols into the system.Compared with static event-triggered protocol,dynamic event-triggered can reduce the need for triggers and avoid unnecessary network resource consumption.Similarly,the dynamic event-triggered protocol is also introduced to analyze the l2-l_∞ state estimation of discrete time-delay MNNs with redundant channels,which further proves that the dynamic event-triggered protocol can save resources more effectively.