Solving rule contradictions in fuzzy controller design

Fuzzy set theory emulates the very complex reasoning process of human operation. The knowledge obtained from experience can be expressed as a set of rules. However, in previous fuzzy controllers (FCs), the expressed rules are always not precise enough and they may not take into consideration every possible combination. Therefore, a good construction of linguistic rules for the fuzzy/rule based controller are very important. To increase the knowledge base, the linguistic rules are collected from more than one expert, however this will increase the probability of the mentioned rules contradiction with each other. In this paper, a new fuzzy controller which is integrated with the concept of a neural network is proposed. Due to the characteristic of the weighting in the proposed network, the proposed fuzzy controller can be operated successfully despite the occurrence of rule contradictions. Examples demonstrate the implementation of the proposed FC in a robot manipulator control system