Considers fuzzy rules acquisition from the viewpoint that the rules should be intelligible as well as the viewpoint that satisfactory system behavior is obtained by the rules. Genetic algorithms (GA) are used to acquire the fuzzy rules consisting of fuzzy sets. A fuzzy rule is encoded to a chromosome using integer coding. GA operators such as crossover and mutation are redefined for integer coding. Two kinds of evaluations are considered. The first one is quantitative evaluation about system behavior defined by some kinds of formulae. The other is qualitative evaluation about system behavior and about intelligibility of acquired fuzzy rules. The intelligibility of acquired fuzzy rules means the evaluation whether the rules are understood easily or not, where the rules are expressed by plain linguistic expressions. The technique of the interactive GA is used in order to incorporate human qualitative evaluation into the GA. The paper adopts procedures to avoid the problems of interactive GA. (1) Some individuals having higher quantitative evaluation are selected from many individuals. (2) Fuzzy rules representing the selected individuals and system behavior inferred from the fuzzy rules are presented to users through a human interface, and they are evaluated by human. Finally the method is applied to car control for running through a crank-type road and for parking in a garage.


    Zugriff

    Zugriff über TIB

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    Acquisition of intelligible fuzzy rules


    Beteiligte:
    Onisawa, T. (Autor:in) / Anzai, T. (Autor:in)


    Erscheinungsdatum :

    1999


    Format / Umfang :

    6 Seiten, 10 Quellen




    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Print


    Sprache :

    Englisch




    Towards Accountability: Providing Intelligible Explanations in Autonomous Driving

    Omeiza, Daniel / Web, Helena / Jirotka, Marina et al. | IEEE | 2021





    Human performance model based on fuzzy rules

    Li, Ni / Li, Xiang / Yang, Kang | IEEE | 2014