Neural network-based event-triggered control design of nonlinear continuous-time systems with variable sampling

This paper focuses on a problem of event-based stabilization for a class of sampled-data neural-network-based control systems. By using a new discrete event-triggering mechanism, an event-based sampled-data three-layer fully connected feedforward neural-network-based controller is constructed. Compared with the conventional periodic sampled-data neural-network-based control in previous works, the main advantage of this paper is that the proposed event-based sampling and transmission scheme not only reduces the updating frequency of the controller, but also guarantees the asymptotical stability of the closed-loop system without dramatically degrading the overall system performance. Based on a discontinuous Lyapunov Krasovskii functional, a convex combination technique and the Wirtinger-based integral inequality, some new criteria are derived to guarantee the asymptotical stability and certain performance of closed-loop system in terms of linear matrix inequalities (LMIs). Based on the proposed criteria, a co-design method is presented to obtain the triggering parameter and the connection weights of the neural network simultaneously while ensuring a certain system performance. Finally, simulation results are provided to show the effectiveness and advantage of the proposed theoretical results.