Generic floating point library for neuro-fuzzy controllers based on FPGA technology

This paper aims at verifying the possibility of building a faster and more accurate generic neuro-fuzzy controller, based on FPGA hardware. In order to achieve this goal, a VHDL library, combining a number of generic modules (that could fit in any controller), is introduced. Two famous neuro-fuzzy controllers, (FSOM and ANFIS), are built by integrated modules from this VHDL library. The performance of the ANFIS controller is compared to its microprocessor-based software implementation. The results showed the hardware controllers are faster and more accurate controllers. During the learning phase, offline learning algorithms (based on Matlab´s neuro-fuzzy toolbox) are used to estimate the controller´s structure and the values of its parameters. The learning phase teaches the hardware controller to adapt to the chosen control process and respond to it during the testing phase.