Optimal controller design for nonlinear systems

Chen-Yuan Kuo, C. L. Yang, C. Margolin

Research output: Chapter in Book/Report/Conference proceedingChapter

2 Citations (Scopus)

Abstract

The commonly used LQR method can be effective for systems with linear characteristics and for performance criteria that well match the minimisation of a quadratic functional form of a cost function. It may not produce a control that demonstrates optimal characteristics for systems which include hard nonlinearities such as nonlinear friction, asymmetric properties and limits imposed on the system states and control actuation. A controller design method is presented termed abbreviated dynamic programming control (ADPC) which condenses the large amount of data generated from the dynamic programming method into a time-invariant nonlinear feedback controller. An empirical system modelling technique is proposed in this research and applied to a single-link robot system to obtain both a linear and nonlinear system model. Using these system models an LQR and ADPC controller are established for this system. The abilities of these two controllers in minimising a predefined quadratic cost function are evaluated through both computer simulations and experimental test runs of this robot system. The ADPC controller is proved to be more effective at minimising the cost function and therefore providing optimal control.

Original languageEnglish (US)
Title of host publicationIEE Proceedings: Control Theory and Applications
Pages97-105
Number of pages9
Volume145
Edition1
StatePublished - 1998

Fingerprint

dynamic programming
nonlinear systems
Nonlinear systems
controllers
Dynamic programming
Controllers
linear quadratic regulator
Cost functions
costs
robots
nonlinear feedback
Robots
optimal control
linear systems
Nonlinear feedback
actuation
friction
computerized simulation
nonlinearity
Linear systems

Keywords

  • Dynamic programming control
  • Nonlinear feedback controller

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Electrical and Electronic Engineering
  • Instrumentation

Cite this

Kuo, C-Y., Yang, C. L., & Margolin, C. (1998). Optimal controller design for nonlinear systems. In IEE Proceedings: Control Theory and Applications (1 ed., Vol. 145, pp. 97-105)

Optimal controller design for nonlinear systems. / Kuo, Chen-Yuan; Yang, C. L.; Margolin, C.

IEE Proceedings: Control Theory and Applications. Vol. 145 1. ed. 1998. p. 97-105.

Research output: Chapter in Book/Report/Conference proceedingChapter

Kuo, C-Y, Yang, CL & Margolin, C 1998, Optimal controller design for nonlinear systems. in IEE Proceedings: Control Theory and Applications. 1 edn, vol. 145, pp. 97-105.
Kuo C-Y, Yang CL, Margolin C. Optimal controller design for nonlinear systems. In IEE Proceedings: Control Theory and Applications. 1 ed. Vol. 145. 1998. p. 97-105
Kuo, Chen-Yuan ; Yang, C. L. ; Margolin, C. / Optimal controller design for nonlinear systems. IEE Proceedings: Control Theory and Applications. Vol. 145 1. ed. 1998. pp. 97-105
@inbook{45a8759cd803490091a31c71b703df5e,
title = "Optimal controller design for nonlinear systems",
abstract = "The commonly used LQR method can be effective for systems with linear characteristics and for performance criteria that well match the minimisation of a quadratic functional form of a cost function. It may not produce a control that demonstrates optimal characteristics for systems which include hard nonlinearities such as nonlinear friction, asymmetric properties and limits imposed on the system states and control actuation. A controller design method is presented termed abbreviated dynamic programming control (ADPC) which condenses the large amount of data generated from the dynamic programming method into a time-invariant nonlinear feedback controller. An empirical system modelling technique is proposed in this research and applied to a single-link robot system to obtain both a linear and nonlinear system model. Using these system models an LQR and ADPC controller are established for this system. The abilities of these two controllers in minimising a predefined quadratic cost function are evaluated through both computer simulations and experimental test runs of this robot system. The ADPC controller is proved to be more effective at minimising the cost function and therefore providing optimal control.",
keywords = "Dynamic programming control, Nonlinear feedback controller",
author = "Chen-Yuan Kuo and Yang, {C. L.} and C. Margolin",
year = "1998",
language = "English (US)",
volume = "145",
pages = "97--105",
booktitle = "IEE Proceedings: Control Theory and Applications",
edition = "1",

}

TY - CHAP

T1 - Optimal controller design for nonlinear systems

AU - Kuo, Chen-Yuan

AU - Yang, C. L.

AU - Margolin, C.

PY - 1998

Y1 - 1998

N2 - The commonly used LQR method can be effective for systems with linear characteristics and for performance criteria that well match the minimisation of a quadratic functional form of a cost function. It may not produce a control that demonstrates optimal characteristics for systems which include hard nonlinearities such as nonlinear friction, asymmetric properties and limits imposed on the system states and control actuation. A controller design method is presented termed abbreviated dynamic programming control (ADPC) which condenses the large amount of data generated from the dynamic programming method into a time-invariant nonlinear feedback controller. An empirical system modelling technique is proposed in this research and applied to a single-link robot system to obtain both a linear and nonlinear system model. Using these system models an LQR and ADPC controller are established for this system. The abilities of these two controllers in minimising a predefined quadratic cost function are evaluated through both computer simulations and experimental test runs of this robot system. The ADPC controller is proved to be more effective at minimising the cost function and therefore providing optimal control.

AB - The commonly used LQR method can be effective for systems with linear characteristics and for performance criteria that well match the minimisation of a quadratic functional form of a cost function. It may not produce a control that demonstrates optimal characteristics for systems which include hard nonlinearities such as nonlinear friction, asymmetric properties and limits imposed on the system states and control actuation. A controller design method is presented termed abbreviated dynamic programming control (ADPC) which condenses the large amount of data generated from the dynamic programming method into a time-invariant nonlinear feedback controller. An empirical system modelling technique is proposed in this research and applied to a single-link robot system to obtain both a linear and nonlinear system model. Using these system models an LQR and ADPC controller are established for this system. The abilities of these two controllers in minimising a predefined quadratic cost function are evaluated through both computer simulations and experimental test runs of this robot system. The ADPC controller is proved to be more effective at minimising the cost function and therefore providing optimal control.

KW - Dynamic programming control

KW - Nonlinear feedback controller

UR - http://www.scopus.com/inward/record.url?scp=0031710377&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0031710377&partnerID=8YFLogxK

M3 - Chapter

AN - SCOPUS:0031710377

VL - 145

SP - 97

EP - 105

BT - IEE Proceedings: Control Theory and Applications

ER -