Tuning of fuzzy controller for an open-loop unstable system: a genetic approach

Abstract

The design, test and evaluation of an optimised fuzzy logic controller (OFLC) is reported in this paper. With the aid of genetic algorithms (GA), the rule-base of an otherwise standard fuzzy logic controller (FLC) is obtained. This is achieved by deriving a tailor-made encoding scheme, initialisation, crossover and mutation of rule table into strings of integers. GA is implemented such that the existing knowledge of the system is utilised to increase the speed of optimisation. The OFLC is successfully applied to control an open-loop unstable system – the ball-and-beam balance system – on a hardware test-bed. A Kalman filter controller (KFC) and a manually tuned fuzzy logic controller (MFLC) are also developed for the test-bed and the performances of the three controllers are compared. The experiment reveals that improved robustness with shorter design cycle can be achieved by integrating GA into an FLC.