A Parameter-Dependent Performance Criterion for Linear Parameter-Varying Systems

Existing control theory for linear parameter-varying systems uses a uniform-in-the-parameters upper bound (a constant scalar) on the induced-L 2 norm. In this paper, this constant is generalized to a function of the parameters; the performance criterion generalizes to an integral quadratic constraint and implies a norm bound that depends naturally on the real-time parameter trajectory. As in existing theory, the synthesis problem reduces to convex optimization involving linear matrix inequalities. A motivating example, gainscheduled flight control of the F-16 VISTA, demonstrates the flexibility gained by incorporating the new performance criterion. 1 Introduction Some recent effort has been given to solving gainscheduled control problems using linear parametervarying (LPV) systems [12]: finite-dimensional linear systems that depend on one or more time-varying, measurable parameters (e.g., families of linearized models for nonlinear aircraft dynamics). So-called LPV control design met..