Robust Adaptive Dynamic Programming for Large‐Scale Systems

This chapter explains the robust adaptive dynamic programming (RADP) theory for the decentralized optimal control of a generalized class of large‐scale systems. The controller design of each subsystem only utilizes local state variables, without knowing the system dynamics. By integrating a simple version of the cyclic‐small‐gain theorem, asymptotic stability can be achieved by assigning appropriate weighting matrices for each subsystem. As a by‐product, certain suboptimality properties can be obtained. The chapter describes the class of large‐scale uncertain systems to be studied. Then, an RADP‐based decentralized optimal controller design scheme is presented. It is also shown that the closed‐loop interconnected system enjoys some suboptimality properties. In addition, the effectiveness of the proposed methodology is demonstrated via its application to the online learning control of a ten‐machine power system with governor controllers.