Explorations in Fuzzy Classifier System Architectures

  • A. G. Pipe
  • B. Carse
Conference paper


The Fuzzy Classifier System paradigm is an elegant and versatile combination of evolutionary and lifetime reinforcement learning based on an underlying Fuzzy Logic structure. It possesses a powerful potential to be a general-purpose linguistically interprétable problem-solver for continuous real-valued problem domains. We present a new description and analysis of a sequence of experiments that we have conducted over the past two years to investigate Fuzzy Classifier System architectures. These experiments have been carried out in the context of a mobile robot control problem. Although some of the individual stages of this sequence of work have already been reported on as the work has proceeded, this paper contains new discussion and analysis of the work in a wider context. Classifier Systems fall into two main categories, the “Pittsburgh” and “Michigan” approaches. We have found that, despite progressive modifications that improved performance of the Michigan approach, the relatively simple Pittsburgh architecture is still able to achieve comparable performance in this application. However, there are three important caveats to this statement. First, the Pittsburgh approach typically requires more fitness evaluations to achieve a similar level of performance. Dependent on the application, this could be a serious disadvantage. Second, there is much more work to be done on tuning the performance of the Michigan approach. Third, there are far more complex problem domains in which these algorithms can be tested and compared, many of which can be identified within the current, mobile robot, application. The outcome of this future work is, at present, unknown.


Membership Function Fuzzy Rule Fuzzy Controller Classifier System Fuzzy Logic System 
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]
    Bonarini A. (2000). An Introduction to Learning Fuzzy Classifier Systems. In P.L. Lanzi, W. Stolzmann and S.W. Wilson (Eds.), Learning Classifier Systems-from Foundations to Applications, Lecture Notes in Artificial Intelligence, pages 83–104. Springer Verlag Berlin Heidelberg, Germany.Google Scholar
  2. [2]
    Carse B., Fogarty T.C. and Munro A. (1996). Evolving Fuzzy Rule-based Controllers using Genetic Algorithms. Fuzzy Sets & Sys 80(3), pp. 273–293.CrossRefGoogle Scholar
  3. [3]
    Carse B., Pipe A.G., (2001). X-FCS: a fuzzy classifier systems using accuracy based fitness — 1st results. In Proc. Int. Conf. Fuzzy Logic & Tech, pp. 195–198.Google Scholar
  4. [4]
    Hwang W. and Thompson W. (1994). Design of Fuzzy Logic Controllers using Genetic Algorithms. In Procs. of the 3rd IEEE International Conference on Fuzzy Systems, pages 1383–1388, Piscataway, NJ: IEEE Computer Press.CrossRefGoogle Scholar
  5. [5]
    Parodi A. and Bonelli P. (1993). A New approach to fuzzy classifier systems. In Procs. of 5th ICGA, pp. 223–230. San Mateo, CA: Morgan Kaufman.Google Scholar
  6. [6]
    Pipe A.G., Carse B. (2000). Autonomous Acquisition of Fuzzy Rules for Mobile Robot Control: First Results from two Evolutionary Computation Approaches. In Procs. of Genetic & Evol. Comp. GECCO 00, pp. 849–856.Google Scholar
  7. [7]
    Valenzuela-Rendon M. (1991). The Fuzzy Classifier System: a Classifier System for Continuously Varying Variables. In Procs. of 4th Int. Conf on Genetic Algorithms, pages 346–353, San Mateo, CA:Morgan Kaufman.Google Scholar
  8. [8]
    Wilson S. W. (1995). Classifier Fitness based on Accuracy. Evolutionary Computing 3(2): 149–175.CrossRefGoogle Scholar

Copyright information

© Springer-Verlag London 2002

Authors and Affiliations

  1. 1.Faculty of Computing, Engineering and Mathematical SciencesUniversity of the West of EnglandBristolUK

Personalised recommendations