Modelling and overlapping decentralized control of a large segmented telescope test-bed

Large-scale systems (LSS) have posed a great challenge to both modeling and control since the high dimensionality often causes complexities in modeling and control. In this study we investigated the overlapping decentralized control based on the input/output characteristics for linear and nonlinear systems. This approach can handle the overlapping dynamic interconnections between subsystems while the centralized control generates a high order controller causing computational complexities; decentralized control fails to meet performance requirements due to the strong interconnections between subsystems. This approach is applied to a typical large-scale system: large segmented telescope. The algorithm developed for linear systems is first verified in the shape control of the segmented telescope model. For a real segmented telescope test-bed, a state space model is first developed through frequency domain system identification. The decentralized and overlapping decentralized control techniques are successfully implemented with a commercial DSP with multi-processor. The overlapping approach shows significant improvement in the performance. Further it is shown that a simple adaptive scheme combined with the decentralized controller can greatly improve the closed-loop performance. The overlapping control is also applied to a type of nonlinear systems.