Data-driven cooperative optimal output regulation for linear discrete-time multi-agent systems by online distributed adaptive internal model approach

In this study, a data-driven learning algorithm was developed to estimate the optimal distributed cooperative control policy, which solves the cooperative optimal output regulation problem for linear discrete-time multi-agent systems. Notably, the dynamics of all the agent systems and exo-system is completely unknown. By combining adaptive dynamic programming with an internal model, a model-free off-policy learning method is proposed to estimate the optimal control gain and the distributed adaptive internal model by only accessing the measurable data of multi-agent systems. Moreover, different from the traditional cooperative adaptive controller design method, a distributed internal model is approximated online. Convergence and stability analyses show that the estimate controller generated by the proposed data-driven learning algorithm converges to the optimal distributed controller. Finally, simulation results verify the effectiveness of the proposed method.