The Divide-and-Conquer Method for Modelling and Control of Nonlinear Systems: Some Important Issues Concerning Its Application
Divide-and-conquer methods, in the context of nonlinear dynamic systems modelling and control design, break the problem down into a number of simpler, frequently linear modelling and control design problems, each associated with a restricted operating region. The most widespread divide-and-conquer nonlinear control design method is probably gain-scheduling. The gain-scheduling method has been successfully applied in many fields, ranging from process control to aerospace engineering. The basic idea behind the method is to divide the nonlinear system to be controlled into local subsystems described by linear dynamic models. A linear control problem is then solved for each of these subsystems. The global control solution—called gain-scheduling control—is obtained by merging partial local solutions. The idea of divide-and-conquer seems very attractive, but its application is not straightforward. The following issues arising during system modelling and control design are treated in this chapter: the properties of the global nonlinear model vis à vis the properties of the local linear models; controller design from first-principles nonlinear continuous-time models; and control design based on local models identified from measured data. The problems are elaborated on simulated examples that include vehicle dynamics and a continuous stirred tank reactor. How to properly use these methods is illustrated by pressure control design on a semi-industrial gas–liquid separator unit.
KeywordsNonlinear System Control Design Local Model Local Controller Linear Subsystem
The financial support of the Slovenian Research Agency through Programme P2-0001 is gratefully acknowledged.
- 1.Åström KJ, Hägglund T (1995) PID controllers: theory, design and tuning, 2nd edn. Instrument Society of America, Research Triangle Park Google Scholar
- 2.Ažman K, Kocijan J (2009) Fixed-structure Gaussian process model. Int J Inf Syst Sci 34:31–38 Google Scholar
- 3.Čokan B, Kocijan J (2002) Some realisation issues of fuzzy gain-scheduling controllers: a robotic manipulator case study. In: Roy R (ed) Soft computing and industry: recent applications. Springer, New York, pp 191–199 Google Scholar
- 5.Johansen TA, Murray-Smith R (1997) The operating regime approach. In: Murray-Smith R, Johansen TA (eds) Multiple model approaches to modelling and control. Taylor & Francis, London Google Scholar
- 10.Kocijan J, Hvala N, Strmčnik S (2000) Multi-model control of wastewater treatment reactor. In: Mastorakis N (ed) System and control: theory and applications. World Scientific and Engineering Society, Singapore, pp 49–54 Google Scholar
- 12.Kocijan J, Vrančić D, Dolanc G, Gerkšič S, Strmčnik S, Škrjanc I, Blažič S, Božiček M, Marinšek Z, Hadjinski MB, Boshnakov K, Stathaki A, King R (2003) Auto-tuning non-linear controller for industrial use. In: Proceedings of the IEEE international conference on industrial technology 2003, Piscataway, pp 906–910 CrossRefGoogle Scholar
- 17.Leith DJ, Leithead WE (2000) On the identification of nonlinear systems by combining identified linear models. In: Proceedings of the 3rd MathMod conference on mathematical modelling, Vienna. ARGESIM report No. 15(1), pp 407–410 Google Scholar
- 18.Leith DJ, Leithead WE (2002) Global reconstruction of nonlinear systems from families of linear systems. In: Proceedings of the 15th triennial world congress of the IFAC, Barcelona Google Scholar
- 20.Leith DJ, Solak E, Leithead WE (2003) Direct identification of nonlinear structure using Gaussian process prior models. In: Proceedings of the European control conference 2003, Cambridge Google Scholar
- 22.Murray-Smith R, Johansen TA (1997) Multiple model approaches to modelling and control. Taylor & Francis, London Google Scholar
- 23.Murray-Smith R, Johansen TA, Shorten R (1999) On transient dynamics, off-equilibrium behaviour and identification in blended multiple model structures. In: Proceedings of 5th European control conference 1999, Karlsruhe, BA-14 Google Scholar
- 25.Nelles O, Fink A, Isermann R (2000) Local linear model trees (lolimot) toolbox for nonlinear system identification. In: IFAC symposium on system identification (SYSID), Santa Barbara, USA Google Scholar