| Summary: | 碩士 === 崑山科技大學 === 電子工程研究所 === 93 === This thesis proposes novel fuzzy hardware fast prototyping algorithms for fuzzy hardware systems in a field programmable analog array (FPAA). With optimal organization of configurable analog blocks (CABs) and permutation switch blocks (PSBs), signal processing behavioral models are synthesized, implemented, and then ready for fast prototyping and verifications in mixed signal SoCs. In the FPAA, each CAB contains basic analog fuzzy operations, which are configurable to realize specific functions. PSBs undertake interconnections of CABs. The optimal interconnections are results of placement-and-routing (P&R) algorithm running in the implementation stage of the prototype design flow. In fast prototyping of fuzzy hardware systems, CABs undertake adaptive membership functions, inferences, implications, fuzzy logic connectives, and current memories. The optimal interconnections of CABs then realize fuzzy rules of any input/output scale and in the same time utilize hardware resources in the FPAA. To achieve adaptive signal processing via physical manipulations of FPAA, the parameters of CABs are adapted with serial analog scan-chains, which precisely scan-in and -out current signals according to the error correction schemes. Digital shift registers transport bit-streams from the host to configure all PSBs in the FPAA and therefore interconnections of CABs are adaptively formed. Consequently, fuzzy rules are adapted. In addition, we also propose a simplification method for calculation of fuzzy operations including fuzzy arithmetic, matching degree calculation, and defuzzification. By operating the LR parameters, the simplication method efectively reduces the complexity of the fuzzy operations and thus the method is very useful for hardware implementation.
|
|---|
