【文章內(nèi)容簡介】 mples of the potentials of fuzzy control is a simulation of traffic signal control in an intersection of two oneway streets. Even in this very simple case the fuzzy control was at least as good as the traditional adaptive control. In general, fuzzy control is found to be superior in plex problems with multiobjective decisions. In traffic signal control several traffic flows pete from the same time and space, and different priorities are often set to different traffic flows or vehicle groups. In addition, the optimization includes several simultaneous criteria, like the average and maximum vehicle and pedestrian delays, maximum queue lengths and percentage of stopped vehicles. So, it is very likely that fuzzy control is very petitive in plicated real intersections where the use of traditional optimization methods is problematic. Fuzzy logic has been introduced and successfully applied to a wide range of automatic control tasks. The main benefit of fuzzy logic is the opportunity to model the ambiguity and the uncertainty of decisionmaking. Moreover, fuzzy logic has the ability to prehend linguistic instructions and to generate control strategies based on priori munication. The point in utilizing fuzzy logic in control theory is to model control based on human expert knowledge, rather than to model the process itself. Indeed, fuzzy control has proven to be successful in problems where exact mathematical modelling is hard or impossible but an experienced human operator can control process. In general, fuzzy control is found to be superior in plex problems with multiobjective decisions. ４ At present, there is a multitude of inference systems based on fuzzy technique. Most of them, however, suffer illdefined foundations。 even if they are mostly performing better that classical mathematical method, they still contain black boxes, . de fuzzification, which are very difficult to justify mathematically or logically. For example, fuzzy IF THEN rules, which are in the core of fuzzy inference systems, are often reported to be generalizations of classical Modus Ponens rule of inference, but literally this not the case。 the relation between these rules and any known manyvalued logic is plicated and artificial. Moreover, the performance of an expert system should be equivalent to that of human expert: it should give the same results that the expert gives, but warn when the control situation is so vague that an expert is not sure about the right action. The existing fuzzy expert systems very seldom fulfil this latter condition. Many researches observe, however, that fuzzy inference is based on similarity. Kosko, for example, writes 39。Fuzzy membership...represents similarities of objects to imprecisely defined properties39。. Taking this remark seriously, we study systematically manyvalued equivalence, . fuzzy similarity. It turns out that, starting from the Lukasiewicz welldefined manyvalued logic, we are able to construct a method performing fuzzy reasoning such that the inference relies only on experts knowledge and on welldefined logical concepts. Therefore we do not need any artificial defuzzification method (like Center of Gravity) to determine the final output of the inference. Our basic observation is that any fuzzy set generates a fuzzy similarity, and that these similarities can be bined to a fuzzy relation which turns out to a fuzzy similarity, too. We call this induced fuzzy relation total fuzzy similarity. Fuzzy IF THEN inference systems are