Mixing crisp and fuzzy logic in applications

Abstract

There are many trade-offs between computing in the fuzzy and crisp domains. The advantages of using linguistic variables and fuzzy functions to improve software reliability and reduce computation requirement has become well understood. This paper explores application implementation combining both conventional crisp (Boolean) and fuzzy linguistic logic. Hedge functions are used to define approximate relationships. For example, functions such as "ABOUT", "NEARLY", and "LESSTHAN" are used to better qualify a variable than simple comparisons. A practical general method is described for implementing hedge functions in the crisp domain that allows mixing fuzzy and crisp calculations in the same fuzzy rule. The implementation presented has been targeted to most of the common embedded system microcomputers used for consumer applications and small scale process control.<>