A novel processor architecture for Sugeno-type fuzzy inference systems

Conventional generic microcontrollers are programmed to implement fuzzy logic algorithms in a computationally intensive way that may include lengthy look-up table techniques. Although, the desired operations are achieved, the memory space occupied would be significantly large. The processing also consumes much time in the relative time frame since all the look-up operations, mathematical operations and logic operations require discrete, finite but large number of steps to complete. Provision of certain additional hardware, which is the subject of the current research, would aid optimization in terms of memory and speed of operation. This would in turn lead to development of better firmware for embedded systems. Existing embedded fuzzy systems use generic architectures which may not necessarily be optimized with respect to ease of programming, memory and speed of operation. Hence, a new architecture has been formulated which aids fuzzy processing within the confines of embedded systems. The architecture is targeted towards dual-input fuzzy inference systems, specially, those found in automotive applications. The Verilog implementation of the proposed architecture has been performed using the Xilinx FPGA series. The results indicate that the architecture is hardware efficient and capable of providing up to 2.439 MFLIPS operation.