On the Implmentation of a Fuzzy CMAC

This paper presents a Novel Fuzzy Cerebellar Model Articulation Controller (CMAC) implementation using Field Programmable Gates Array (FPGA) available in a flexible software/hardware co-design platform. The novel fuzzy quantization technique used for reducing CMAC memory requirement is similar to Discrete Incremental Clustering (DIC), with modifications to meet the hardware constraint. The objective of the project is therefore twofold: to propose a hardware-friendly fuzzy clustering technique based on DIC, to implement the Fuzzy CMAC in hardware cost-saving manner and exploit realization of its inherent parallelism in FPGA architecture. Test result for classification of 2-spiral problem is presented, which demonstrates the validity and generalization capability of proposed architecture and its advantage of reduced memory requirement.