Non-fragile state estimation for discrete neural networks

Abstract

In this paper, the non-fragile state estimation issue is investigated for a class of discrete-time neural networks with time-delays. Both linear and nonlinear parts are considered in the network output. The gain variation of the additive type is introduced to depict the possible implementation imprecision in the neuron state estimator. A non-fragile state estimator is constructed which can guarantee the estimation performance against the gain variations. On the basis of the Lyapunov stability theory, sufficient conditions are put forward to ensure that the desired non-fragile state estimators can exist. The explicit expression of the desired estimators is obtained in terms of the solution to a linear matrix inequality. The usefulness and applicability of the proposed design approach are verified by a numerical example.