Genetic Algorithms for On-Line System Identification

  • Kevin Warwick
  • Yong Ho Kang


When required for batch processing, system identification can be used in a number of ways to find a good fit in terms of modelling the system structure, time delay and characteristic parameters of a plant. The best structure and delay selection can often be found in terms of a computationally simple model order testing procedure, with a range of different candidate models being considered.

Due to the need for computational efficiency for on-line identification, this is often restricted to a more straightforward parameter estimation exercise, the structure being selected as a fixed term, usually of low order. Any structural tuning is then carried out in terms of sampling period variation, which can mean that vital plant information does not appear in the identified model.

This paper presents an on-line technique for system identification, making use of genetic algorithms for structure and time delay estimation. The identification scheme is multi-level, the bottom level consisting of the more usual parameter estimation exercise, with the upper level carrying out structure identification. The first of these can update in real-time, whilst the second operates in its own time. At any instant, one model is selected as “best” in terms of both structure and parameters, and this is the one employed as on-line identification.


Genetic Algorithm Time Delay Estimation Structural Tuning Vital Plant Delay Selection 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. [1]
    Karny, M. ‘Algorithms for Determining the Model Structure of a Controlled System’ Kybernetica, 19, 2, 164–178, 1983.MATHGoogle Scholar
  2. [2]
    Karny, M. and Kulhavy, R., ‘Structure Determination of Regression-type Models for Adaptive Prediction and Control’, in Spall J.C. (ed.), Bayesian Analysis of Time Series and Dynamic Models, Marcel Dekker, 1988.Google Scholar
  3. [3]
    Goldberg, D., ‘Genetic Algorithms in Search, Optimization and Machine Learning’, Reading MA, Addison-Wesley, 1989.MATHGoogle Scholar
  4. [4]
    Soderstrom, T and Stoica, P., ‘System Identification’ Prentice-Hall Inc., 1989.Google Scholar
  5. [5]
    Fortescue, T.R., Kershenbaum, L.S. and Ydstie, B.E., ‘Implementation of Self-tuning Regulators with Variable Forgetting Factors’ Automatica, 17, 931, 1981.CrossRefGoogle Scholar
  6. [6]
    Warwick, K., ‘System Identification’, chapter in “Industrial Digital Control Systems”, rev. 2nd ed., K. Warwick and D. Rees (eds.), Peter Peregrinus Ltd., 1988.Google Scholar
  7. [7]
    Kiernan, L.A. and Warwick, K., ‘Developing a Learning System Capable of Hypothesis Justification’, Proc. Int. Conference Control 91, Edinburgh, 272-276, 1991.Google Scholar
  8. [8]
    Leith, D.J., Murray-Smith, D.J. and Bradley, R., ‘Combination of Data Sets for System Identification’ Proc. IEE, Part D, 140, 1, 11–18, 1993.CrossRefGoogle Scholar
  9. [9]
    Kristinsson, K. and Dumont, G.A., ‘System Identification and Control using Genetic Algorithms’, IEEE Trans. on Systems, Man and Cybernetics, 22, 5, 1033–1046, 1992.MATHCrossRefGoogle Scholar

Copyright information

© Springer-Verlag/Wien 1993

Authors and Affiliations

  • Kevin Warwick
    • 1
  • Yong Ho Kang
    • 1
  1. 1.Department of CyberneticsUniversity of ReadingWhiteknights ReadingUK

Personalised recommendations