ASIC implementation of the symmetric fuzzy processor and its application to adaptive systems
This dissertation presents a VLSI design of a symmetric fuzzy processor. The design features fuzzification, defuzzification and inference operations while allowing the implementation of a knowledge base via rules. By combining the inherent advantages of symmetric triangular membership functions and fuzzy singleton sets, a novel structure for the fuzzification model is obtained. The structure accelerates the evaluation of the antecedent degree which is evaluated by a simple mathematical relation which calculates the resulting value using the end-points of matched fuzzy member functions. This feature enables the processor to avoid the requirement of storing all the sample values of the fuzzy membership function in memory as is the case with other approaches. In addition, the resulting design structure simplifies computations associated with centroid defuzzication in that certain simplifying assumptions eliminate the need for a divider circuit. By using a very high speed integrated circuit hardware description language (VHDL) compiler and by making use of a simulator provided through the Mentor Graphics EDA design tool, optimization of the VLSI design has been obtained. Results show that the resulting fuzzy processor can be implemented on a single 1.2$\mu$m CMOS VLSI chip with 16.7 mm$\sp2$ die size and a total of 36,080 transistors. Moreover, simulation indicates that numerical computations including centroid defuzzification can be accomplished in 0.55 $\mu$s. within an accuracy of 96%, thus making it suitable for a wide range of real-time applications. Up to 49 consequent knowledge rules based for seven fuzzy membership functions associated with the chip's two input variables can be downloaded into a 64-byte static RAM allowing designers to create a fuzzy processing system without the need for additional on-board memory. Finally, as an example of the application of the proposed fuzzy processor model, results are presented from a study to simulate a second order linear control system and a non-linear structure for adaptive channel equalization of a bipolar signal passed through a dispersive channel in the presence of additive noise. It is shown that difficulties commonly associated with channel non-linearity and additive noise correlation can be overcome by the use of an equalizer employing the developed fuzzy structure. ^
Engineering, Electronics and Electrical|Artificial Intelligence
Abdel-Hafeez, Saleh Mesbah, "ASIC implementation of the symmetric fuzzy processor and its application to adaptive systems" (1997). ETD Collection for University of Texas, El Paso. AAI9801058.