Fuzzy Logic Based Objective Function Construction for Evolutionary Test Generation

Abstract

The test case generation problem can be stated as an optimization problem
whereby the closeness of test cases to violating the postcondition of a
formal specification is maximized, subject to satisfying its precondition.
This is usually implemented by constructing an objective function which
provides a real-valued estimate of how distant all of the constraints are
from being violated, and then trying to minimize it.
A problem with this approach is that such objective functions may contain
plateaux, which make their minimization hard. We propose a similar approach,
grounded on fuzzy logic, which uses, instead of a "distance from violation"
objective function, a fuzzy degree of proximity to postcondition violation
and produces plateaux-free objective functions by construction.
The approach is illustrated with the help of a case study on the functional
(black-box) testing of computer programs.