Fuzzy rendszerek és modellek elemzése és identifikációja = Analysis and identification of fuzzy systems and models

Uj algoritmuscsaladot fejlesztettunk ki, a bakterialis memetikus algoritmusokat, a bakterialis evolucios algoritmust globalis, a Levenberg-Marquard algoritmust pedig lokalis kereseskent alkalmazva. Ez az eljaras jobb algoritmusokat eredmenyezett az ismert hasonlo modszereknel a pontossag es a ciklusszam osszefuggeseben; ezt kulonboző referencia alkalmazasok es mas peldak segitsegevel bizonyitottuk. Az alkalmazasok masik csoportjat a logisztika adta. Kiterjesztettuk Kano minősegi modelljet fuzzy exponensekre, melyet BMA-val optimalizaltunk es megkezdtuk az utazo ugynok problema kozelitő megoldasanak vizsgalatat is. Megmutattuk, hogy a fuzzy szabalyinterpolacio szamos valos problema megoldasara alkalmas. Sikeresen foglalkoztunk komplex forgalomiranyitasi alkalmazasokkal, tovabba vasuti menetrend es keses miatti atutemezes kerdeskorevel. Szoftverrendszert implementaltunk, mely nagyszamu fuzzy kovetkeztetesi es iranyitasi algoritmus osszehasonlitasara alkalmas. Kiterjesztettuk a fuzzy szignaturakat hierarchikus strukturakra is es a Mamdani algoritmusra is. A fuzzy szignaturakat robotok mozgasiranyitasara es kommunikaciojara alkalmaztuk. E robotokat szimulacios technikaval es sajat fejlesztesű hardver segitsegevel is vizsgaltuk. Uj kutatasi reszteruletet inditottunk el a fuzzy műveletek es a rajtuk alapulo fuzzy flip-floppok vizsgalataval, melyekből konnekcionista rendszereket hoztunk letre es e fuzzy neuralis halozatokat modellkonstrukciora es approximaciora alkalmaztuk. | We developed a new family of algorithms, the Bacterial Memetic Algorithms by combining the Bacterial Evolutionary Algorithm as a global search and the Levenberg-Marquard algorithm as a local search method. This approach provided better algorithms in terms of approximation accuracy and population cycles than other similar approaches in the literature, as it was evidenced by various benchmark and real life applications. Another group of successful applications is in the logistics area. We extended Kano’s quality model to fuzzy exponents, optimized by BMA, and we started to research for the approximate solution of the Traveling Salesman Problem.We showed that fuzzy rule interpolation could be deployed for a number of real application areas. We dealt with complex traffic control applications as well as with railway time table and delay triggered rescheduling problems successfully. We implemented a software for the comparison of a large number of fuzzy reasoning and control algorithms. We extended Fuzzy Signatures to both hierarchical structures and Mamdani’s algorithm. We applied Fuzzy Signatures for motion control and fuzzy communication of robots. Such robots were investigated both in simulation and hardware construction developed by ourselves. We started a new research sub-direction by analyzing fuzzy operators and fuzzy flip-flops based on them. We built connectionist systems from them and we used these fuzzy neural networks for model construction and approximation.