Dual-level gain-scheduled control for unmodeled subsystem dynamics

This paper extends the concept of gain-scheduling on a priori unmodeled dynamics by introducing parameter-dependent weighting filters. Using these filters, a dual-level gain-scheduled controller can be designed, where the controller depends on the weighting filter parameters and perturbation dynamics. Parameterizing the weighting filter provides a more structured description of the unmodeled dynamics, and reduces design conservatism by providing variable weighting envelope which can be tailored to the unmodeled dynamics in the actual plant. The synthesis problem is formulated as a continuously-constrained LMI optimization problem, which combines the existing LPV and LFT gain-scheduling methods. A design example is presented, and the dual-level gain-scheduled controller delivers performance which is superior to a dynamic gain-scheduled controller with a fixed weighting filter and which approaches the performance of the worst case "point"" design.