Adaptive design of decentralized controllers

Design of optimal decentralized controllers for large-scale systems usually involves the solution of a high dimensional problem. A procedure is proposed for performing such design at the subsystem level using reduced order models for the rest of the system. This procedure is well suited for adaptive implementation. A gradient algorithm is derived for computing the parameters of the optimal decentralized controllers for discrete time systems. This algorithm is combined with a system identification technique to provide an adaptive method for decentralized controller design. An illustrative example is presented.