A new approach to fuzzy logic inference based on spatial relationship of fuzzy subsets

Abstract

An inference method which matches the intuition of human reasoning is proposed. For each fuzzy rule "If x is A/spl tilde/ then y is B/spl tilde/", the proposed inference method produces an output fuzzy subset B/spl tilde/' according to the spatial relationship of input fuzzy subset A/spl tilde/' and the antecedent A/spl tilde/. The membership function of the output fuzzy subset B/spl tilde/' is of the same shape as that of the input fuzzy subset A/spl tilde/'. Moreover, the spatial relationship of B/spl tilde/ and B/spl tilde/' is consistent with that of A and A/spl tilde/'. Experiments on a truck backer-parking controller are performed to compare our proposed inference method with E.H. Mamdani's (1981) method. These experiments show that our results compare favorably with those of Mamdani's method.