Event-Driven Cooperative Identification for Nonlinear Systems Using Neural Networks

In this paper, an event-based cooperative identification scheme is proposed for a class of uncertain nonlinear systems. Radial basis function (RBF) neural networks (NNs) are used to identify nonlinear systems. The systems to be identified have identical system functions, but their inputs are different. For each system, a trigger condition which is based on its own estimated NN weights is designed to overcome the disadvantage of continuous communication. Each system broadcasts its estimated NN weights over the communication network only when the trigger condition is satisfied. If communication topology is connected and undirected, it is proved that the weights of all NNs can converge to a small neighborhood of their optimal values. Furthermore, there is no Zeno behavior during control process. A numerical example is given to show these features.