Abstract
A self-organising fuzzy controller (SOC for short) is a heuristic rule-based controller and a further extension of an ordinary fuzzy con troller which was first introduced by Assilian and Mamdani. This has a hierachy structure which consists of an algorithm being identical to a fuzzy controller at the lower loop and a learning algorithm accommodating the performance evalution and rule modification function at the upper loop.
In this paper, the latest version of SOC, which overcomes earlier problems such as cycling phenomena of control response, poor settling time and inefficient learning of control rules, is mainly described. The inferrence method adopted in SOC is first discussed to give the theoretical basis which will be expedient for the exposition to follow. The configuration of the latest algorithm of SOC is explained along with the six main blocks into which SOC can be divided from a functional point of view, in particular, placing an emphasis on the learning algorithm.
A guide-line for the optimal setting of controller parameters to be ajusted in SOC has been derived from a semantic approach and confirmed experimentally. The result has revealed an interesting property of SOC that the parameters are insensitive to those of a control object unlike a PI controller.
Two control responses of SISO and MIMO process are illustrated for the purpose of demonstrating the control performance of SOC and the guide-line for the optimisation.
