Variance-constrained H∞ state estimation for time-varying delayed neural networks with random access protocol and sensor failures

In this paper, a recursive state estimation method is proposed for delayed recurrent neural networks subject to random sensor failures and random access protocol, where time-varying characteristic and two performance evaluation indices are taken into account. The phenomenon of random sensor failures is characterized by introducing a random variable with certain occurrence probability. In order to prevent the data from collisions and save the resources, the random access protocol is introduced into the transmission channel, in which it is assumed that only one sensor node is allowed to access the network. Our aim is to propose the H∞ state estimation strategy without utilizing the augmentation method, where the estimator has the same order of the original neural state. In particular, we provide some sufficient conditions, which can guarantee two performance requirements from the noise attenuation and estimation accuracy perspectives. Finally, we use a simulation example to illustrate the feasibility of proposed H∞ state estimation method handling the concerned communication constraints.