DESIGN OF PARTIALLY DECENTRALIZED CONTROL SYSTEMS

Abstract

The use of partially decentralized controllers has been hindered by a lack of study in this area, even though they offer performance superior to that of decentralized controllers without the complexity of full multivariable controllers. Control theories that are specific to partially decentralized control systems are not available. In particular, two issues in the design of partially decentralized control system need to be addressed: control structure selection and controller design. One way to circumvent the problem of not having available theories on these two design issues is to treat partially decentralized controllers as nonsquare decentralized controllers. Consequently, both control structure selection and controller design can be tackled by using existing theories and methods developed for nonsquare decentralized control systems. Selection of suitable partially decentralized control structures poses a challenging problem since there are a great number of alternatives. A method for the selection of partially decentralized control structures is developed in this work. This involves extending the µ Interaction Measure to nonsquare decentralized systems. After expanding the partially decentralized control system to the nonsquare decentralized form, the modified µ Interaction Measure can be used in the selection of suitable partially decentralized control structures. The design of robust partially decentralized controllers is also considered. In this work, an independent design method for robust nonsquare decentralized controllers is applied directly to partially decentralized control systems which have been expanded to the nonsquare decentralized structure. Robust partially decentralized controllers are designed successfully using this approach.