No label defined (Q4945440)

The authors present two basic approaches to the problem of gain-scheduling for linear time-varying systems with bounded parameters and their variation rates. The gain-scheduled output feedback control problem consists of finding a dynamic linear parameter varying controller that ensures internal stability and a guaranteed \(L_2\)-gain bound for the closed-loop operator from the disturbance signal to the error signal. This problem can be solved using quadratic parameter-dependent Lyapunov functions and solving the linear matrix inequality system, which is however infinite due to its dependence on the varying parameter and its derivative. The second formulation of gain-scheduled controllers, called the projected characterization, enables the controller variables to be eliminated; however this still leads to infinite-dimensional linear matrix inequalities. The authors discuss how to reduce the problem to the finite-dimensional case. One way for doing this is to grid the value set of varying parameters. It leads to a convex optimization problems that can be solved by polynomial-time algorithms. The other problem considered by the authors is how to reduce conservatism of the control system. A new scaling technique that takes into account the uncertainty structure is proposed. To demonstrate practical aspects of the discussed design techniques and to discuss numerical problems of the presented algorithms, the authors consider the gain-scheduled control problem for a two-link flexible manipulator under various assumptions on available information. For the entire collection see [Zbl 0932.00034].