A Variance-Constrained Method to Protocol-Based H∞ State Estimation for Delayed Neural Networks with Randomly Occurring Sensor Nonlinearity

In this paper, we consider the H-infinity state estimation issue under variance constraint for recurrent neural networks with time-varying parameters, where the time-delay as well as randomly occurring sensor nonlinearity are handled and the communication is scheduled by round-robin protocol. The phenomenon of randomly occurring sensor nonlinearity is modeled by a random variable obeying the Bernoulli distribution with known probability. In order to alleviate unnecessary network congestion in communication channels, the round-robin protocol is introduced to specify which network node has the right to access the network channel at each time step. In particular, the objective is to develop the time-varying state estimation method such that, in the presence of time-delay, randomly occurring sensor nonlinearity and round-robin protocol, the sufficient conditions are given and both the error variance boundedness and the pre-set H-infinity performance index can be achieved simultaneously. In the end, we provide a simulation example with comparison tests to demonstrate the feasibility of presented H-infinity state estimation method.