Data-driven bipartite synchronization control of multi-agent systems with asymmetric input saturation over switching networks

Abstract

This paper addresses the bipartite synchronization (BS) issue of nonlinear discrete-time multi-agent systems (NDTMASs) with asymmetric input saturation in switching cooperative-competitive networks. First, a local neighborhood error system (LNES) is constructed based on the switching cooperative-competitive networks. By utilizing a hyperbolic function, we design a novel local cost function and derive an optimal control policy. Based on this control policy, we construct a novel saturation-input policy iteration (SIPI) technique. Furthermore, we establish the convergence of the SIPI algorithm. It is demonstrated that the LNES can converge to zero, and thus the BS problem for MASs can be solved. To implement the SIPI algorithm, a control framework based on an actor-critic architecture is constructed. To achieve asymmetric input saturation, we select an asymmetric constrained function as the activation function for the actor network. Finally, an example is presented to illustrate the reliability of our approach.