An automated design flow from linguistic models to piecewise polynomial digital circuits

This paper describes how the different CAD tools of the environment Xfuzzy 3, developed in Microelectronics Institute of Seville and University of Seville, allow to translate expressive linguistic models into mathematical ones, in particular, into a combination of piecewise polynomial systems that can be implemented efficiently in hardware. The new synthesis tool of Xfuzzy 3 automates communication with Xilinx System Generator in Matlab, thus facilitating implementation of the linguistic model into an FPGA from Xilinx. This is illustrated with the design of a navigation controller for an autonomous robot.

I. Introduction

Model-based approaches are currently gaining popularity to tackle the growing complexity of embedded systems development since they allow working with a high degree of abstraction, enhance understanding and reduce the time to market. Particularly linguistic models, that is, descriptions of systems obtained from heuristic knowledge of human experts expressed linguistically, facilitate understanding and rapid development. In order to translate linguistic models into mathematical ones, which can be implemented in hardware and/or software, fuzzy logic-based systems have been employed widely in the recent years [1]. However, while expressive linguistic models have been employed in many software applications, the models implemented in hardware and embedded software only contain a single (plain) rule base with simple antecedents and consequents, thus reducing the applicability of hardware approaches. This is the case of many design environments for fuzzy systems, such as FIDE, rFLASH, and FuzzyTECH, when generating embedded software for specific processors, and the case of many fuzzy digital circuits.