A MATLAB toolbox for fixed-order, mixed-norm control synthesis

Author

Jacques, David R. ; Canfield, Robert A. ; Ridgely, Brett ; Spillman, Mark S.

Author_Institution

Dept. of Aeronaut. & Astronaut., Air Force Inst. of Technol., Wright-Patterson AFB, OH, USA

Volume

16

Issue

5

fYear

1996

fDate

10/1/1996 12:00:00 AM

Firstpage

36

Lastpage

44

Abstract

This article introduces a MATLAB toolbox for fixed order, mixed-norm control synthesis. The Mixed-Norm Toolbox contains a complete set of routines for both continuous and discrete-time systems. The problem addressed by the toolbox is that of finding a compensator which minimizes the H₂ norm of a transfer function, while constraining any combination of H_∞ and/or l₁ (L₁) norms of possibly dissimilar transfer functions to be below specified levels. Within reason, any number or combination of constraints can be added to the problem, and the method constrains the norms directly without reliance on upper bounds. The primary contribution of the Mixed-Norm Toolbox is a modular collection of norm and gradient algorithms which can be used with almost any nonlinear, constrained optimization solver. While global convergence is not guaranteed for the resulting nonconvex problem, the toolbox has been successfully used to show portions of Pareto optimal curves and surfaces for a wide variety of problems

Keywords

H^∞ control; compensation; control system CAD; transfer functions; H_∞ norms; H₂ norm minimization; L₁ norms; MATLAB toolbox; Mixed-Norm Toolbox; Pareto optimal curves; Pareto optimal surfaces; compensator; continuous systems; discrete-time systems; fixed-order mixed-norm control synthesis; global convergence; gradient algorithms; l₁ norms; nonlinear constrained optimization solver; norm algorithms; transfer function; Computer languages; Constraint optimization; Control system synthesis; Convergence; Hydrogen; MATLAB; Optimal control; Space technology; Transfer functions; Upper bound;

fLanguage

English

Journal_Title

Control Systems, IEEE

Publisher

ieee

ISSN

1066-033X

Type

jour

DOI

10.1109/37.537207

Filename

537207