Distributed neuro-adaptive controller for output consensus of uncertain multi-agent systems with a high-dimensional leader

In this paper, we study the output consensus problems for heterogeneous multi-agent systems (MASs) with a high-dimensional leader and directed topology, where the followers are subject to unknown uncertainties. First, an observer is designed to estimate leader's unmeasurable states. Then, by using neural network (NN) approximation theory, a neuro-adaptive controller consists of a compensation part and a cooperative term is proposed. Unlike most existing neuro-adaptive controllers that utilize some discontinuous functions to eliminate the effects of NN approximation errors, a new class of continuous function is used and thereby the chattering effect is avoided. By using tools from graph theory and stability analysis of dynamical systems, it is proven that asymptotical output consensus can be achieved if the underlying network topology contains a directed spanning tree rooted at the leader and the control parameters are suitably chosen. Finally, a numerical simulation is presented to verify the theoretical result.