Non-fragile mixed passive and [formula omitted] state estimation for singularly perturbed neural networks with semi-Markov jumping parameters

Abstract

For continuous-time semi-Markov jump neural networks with singularly perturbations, this work deals with the issue of mixed passive and H∞ non-fragile state estimation. In view of the multiple-time-scales phenomenon that the dynamic behavior of the neural networks may present, the neural networks are interpreted as singularly perturbed systems. Meanwhile, a singular perturbation parameter, i.e., ε, is employed to describe the degree of separation of the fast and the slow states of the system. In order to enhance the tolerance of the estimator when facing gains variations, the non-fragility as a key factor is directly considered in the process of the expected estimator design. Afterwards, relying on the aforementioned considerations, by constructing the ε-dependent Lyapunov function, new sufficient conditions are deduced to ensure the existence of the available non-fragile state estimator. Finally, the superiority of the proposed method is verified by an illustrated example.