A knowledge-base revision tool for the fuzzy constraints-based *FLIP++ scheduling library

Abstract

Real world problems are often overconstrained, ill-defined, or tolerate good though suboptimal solutions as long as those are found before a deadline. Freuder called them partial constraint satisfaction problems (PCSPs). Inspired by classic constraint satisfaction methods, he proposed a model to define and study PCSPs. Another path leading to comparable notions and originating from fuzzy decision making was first investigated by Dubois and Prade. To increase modeling capability, the approaches add one or more attributes such as the following ones to constraints: gradual satisfaction, priority, and compromises. However, given a problem, it is usually difficult to choose a priori parameters for these attributes even within one approach. Counterintuitive small changes in the parameters can have chaotic effects on solution rankings. We propose a new method to select a PCSP model corresponding to the problem one expects to solve. The method basically allows one to tune parameters consistently. Theoretical analysis and same initial experiments indicate that our method makes PCSP models ready to confront real world problems.