Data-Driven Adaptive Cooperative Output Regulation for Completely Unknown Linear Multi-Agent Systems Based on Finite Length Data

Abstract

With respect to complex control systems, traditional model-dependent methods are increasingly challenged, particularly when system models are unknown or intractable. Moreover, most past research has focused on systems that are partially or fully known. In this technical paper, a data-driven paradigm is employed to investigate the cooperative output regulation problem (CORP) for completely unknown linear heterogeneous discrete multi-agent systems (MASs). Input and state information are utilized to design effective control strategies and a novel data-based algorithm is proposed with finite length data. An adaptive observer is designed to estimate the exosystem state, with only the leader's children having access to the unknown leader's system matrix. To address the challenge of unknown dynamics, the CORP is transformed into a linear quadratic regulation (LQR) problem by solving the regulation equation. Compared with the reinforcement learning method, the closed-form optimal control gain is obtained directly from the relevant data without the need for an initial stabilization controller or iterative calculation. Simulation results validate the proposed scheme's effectiveness.