Robust neural output-feedback stabilization for stochastic nonlinear process with time-varying delay and unknown dead zone

This article investigates the output-feedback control of a class of stochastic nonlinear system with time-varying delay and unknown dead zone. A robust neural stabilizing algorithm is proposed by using the circle criterion, the NNs approximation and the MLP (minimum learning parameter) technique. In the scheme, the nonlinear observer is first designed to estimate the unmeasurable states and the assumption “linear growth” of the nonlinear function is released. Furthermore, the uncertainty of the whole system (including the perturbation of time-varying delay) is lumped and compensated by employing one RBF NNs (radial basis function neural networks). Though, only two weight-norm related parameters are required to be updated online for the merit of the MLP technique. And the gain-inversion related adaptive law is targetly designed to mitigate the adverse effect of unknown dead zone. Comparing with the previous work, the proposed algorithm obtains the advantage: a concise form and easy to implementation due to its less computational burden. The theoretical analysis and comparison example demonstrate the substantial effectiveness of the proposed scheme.