Efficient convex design of robust feedforward controllers

Abstract

A practical method is proposed for the convex design of robust feedforward controllers, which ensure H∞/L2 performance in the face of LTI and arbitrarily time-varying model uncertainties. A technique which computes the global minimum of this difficult infinite dimensional optimization problem is proposed, as well as a suboptimal but computationally less involving algorithm. A missile example illustrates the efficiency of the scheme: a robust feedforward controller is designed, either on the continuum of linearised time invariant models (corresponding to trim points), or on a quasi-LPV model representing the non-linear one.

Robust ℒ︁2‐gain feedforward control of uncertain systems using dynamic IQCs

“…Owing to the nature of the feedforward problem, instead of (2) and (4), we will be using conditions (5) and (6). Some subtleties involving the choice of so that (1) is satisfied are discussed next.…”

“…In contrast to nominal feedforward design, which can be viewed as a particular instance of the standard H ∞ -control problem, robust feedforward synthesis is only known to admit a solution in terms of convex optimization under rather specific hypotheses. In [5] it is assumed that the particular class of IQCs describing the uncertainties is frequency independent, while [6] addresses frequency-dependent IQCs, with the necessity to restrict the search to feedforward controllers that can be expanded in terms of a priori given basis filters. In parallel to this work, Scorletti and Fromion [7] propose a solution for the particular class of frequency-dependent D-and G-scalings as known from structured singular value theory.…”

A data-driven approach to power converter control via convex optimization

“…In other words, for a finite number of frequency points n selected in ω ∈ [0, ∞), then n constraints must be satisfied. This frequency grid can be predefined in a variety of manners (see [15], [16], [17]). …”

Data-driven precompensator tuning for linear parameter varying systems

“…For LPV systems, methods have been proposed ( [3], [4]) to tune precompensators and feedforward controllers whose parameters vary also as a function of the operating point. These methods, however, are based on uncertain identified LPV models and thus, unlike direct data-driven methods, suffer from model uncertainty.…”