Training fuzzy logic controller software components by combining adaptation algorithms

  • Authors:

  • J. Chen;D. C. Rine

  • Affiliations:

  • School of Information Technology and Engineering, George Mason University, 4400 University Dr., Fairfax, VA;School of Information Technology and Engineering, George Mason University, 4400 University Dr., Fairfax, VA

  • Venue:
  • Advances in Engineering Software
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

An approach for an effective and efficient off-line training of particular classes of reusable controller software components is presented. To build a necessary relationship between a component's abstract and concrete levels, each control software component is represented at the abstract level by means of a set of adaptive fuzzy logic rules and at the concrete level by means of adaptive fuzzy membership functions. Training includes two phases: testing and adapting. The testing phase is for identifying faulty fuzzy elements of a component, while the adapting phase is for modifying membership functions. We employ genetic algorithms, neural network algorithms, Monte Carlo algorithms, and their combinations in each phase. This approach is illustrated by training automotive controller software components (simulation). Experimental simulation results show that our off-line training approach supports controller software component adaptation effectively and efficiently in terms of controlled process operation accuracy and effort spent.