Stability analysis of an α-cut based interpolation method ∗ Domonkos Tikk 1 , P´ eter Baranyi 1,2 , Yeung Yam 3 ,L´aszl´oT.K´oczy 1 1 Dept. of Telecom. & Telematics, Technical Univ. of Budapest {tikk,koczy}@ttt-202.ttt.bme.hu 2 Research Group for Mechanics of Hung. Acad. of Sci. baranyi@get.elektro.bme.hu 3 Dept. of Mechanical & Automation Engin., Chinese Univ. of HK yyam@mae.cuhk.edu.hk Abstract Rule interpolation methods deal with sparse fuzzy rule bases. The KH interpolation method, which has advantageous computational be- haviour, can produce ,,abnormal” output in certain cases. Recently a new method has been proposed to eliminate this property. The goal of this paper is to investigate the mathematical stability of the new α-cut based interpolation approach originated from the KH interpola- tion. The original KH interpolation is stable in the sense that it does not react too unexpectedly if the inputs change slightly. The main result of this work points out that this important feature of the KH interpolation can be carried over for the proposed method. Futher- more, a way of possible generalization of the new method will be also presented. 1 Introduction The classical inference methods in fuzzy control (Zadeh, Mamdani and Sugeno) deal with dense rule bases. If the universe of discourse is not com- pletelycoveredbytheruleantecedentsitcanhappenthatforanobservation * Research supported by the National Scientific Research Foundation (OTKA) Grants No. T019671, T030655, by the Hungarian Ministry of Culture and Education (MKM) Grant No. FKFP 042/1997. 1