NOVEL APPROACH TO DECENTRALIZED CONTROLLER DESIGN FOR LARGE SCALE UNCERTAIN LINEAR SYSTEMS

Such characteristics as large order multi-input multi-output (MIMO) system, system uncertainties, and information structure constraints. make large scale systems too complex to be effectively controlled. For the purposes of analysis and controller synthesis such systems are divided into independent sub-problems and controlled under specific information constraints. This is known as decentralized control. Analyzing stability of each of the resulting, smaller systems separately, i.e., by neglecting interconnections, is a tractable but highly conservative approach. In the previous literature on decentralized control, complex systems are divided into two classes of systems, i.e., those with strong interaction links and those with weak ones. The present paper proves that the criterion of whether a complex system is stable or not is a better basis of the breakdown of systems into classes. In the case of a stable complex system, the corresponding decentralized controller design is significantly easier than that in the unstable case. In this paper an original method of decentralized controller design is obtained for linear time-invariant large scale systems with small conservatives. The design procedure consists of two basic steps as follows. In the first step, basic properties of closed-loop controlled subsystems without interaction are determined, in such way that stability and performance of the closed-loop large scale system are guaranteed. In the second step, a decentralized con-trol algorithm needs to be designed, which ensures the demanded subsystem closed-loop properties. If in the second step conditions are satisfied the stability and performance of subsystems and the complex plant is guaranteed, the decentralized controller design procedure performs on the subsystem level. Examples show the effectiveness of the propose method.