Mathematical representation of fuzzy membership functions

The most commonly used membership function distributions in fuzzy controllers are evenly-spaced triangular and trapezoidal membership functions which allow simple traversal routines to fuzzify the input. Evenly-spaced membership distributions become very sensitive to inaccuracies in the fuzzy representation of the input. The paper proposes a membership function shape and distribution which makes the fuzzified input more robust to biases in the fuzzified input.<<ETX>>