PRACTICAL MULTI-CONSTRAINT H-INFINITY CONTROLLER SYNTHESIS FROM TIME-DOMAIN DATA

This paper presents a practical algorithm for MIMO controller design w ith multiple H-infinity norm constraints. The plant is described by it s sampled-data impulse response matrix, which could be determined dire ctly from measurements; the controller design parameters are the tap w eights of an FIR discretization of the Q (or Youla-) parameter. We app roximate the H-infinity norm by the maximum transfer matrix 2-norm on an even frequency sampling over theta = [0, pi]. The control design al gorithm uses the Newton method to minimize two cost functions sequenti ally: the first determines multiple H-infinity feasibility, and the se cond minimizes a generalized entropy. As a design example, we control the acoustic radiation from a (mathematically modelled) submerged sphe rical shell. The plant-model impulse response matrix has McMillan degr ee 8 800. We specify and synthesize a controller that simultaneously a chieves ten H-infinity constraints: one on the radiation reduction, on e for stability robustness, and eight on actuator authority.