Abstract

Introduction A finite domain constraint system can be viewed as an linear integer constraint system in which each variable has an upper and lower bound. Finite domains have been used successfully in Constraint Logic Programming (CLP) languages, for example CHIP [4], to attack combinatorial problems such as resource allocation, digital circuit verification, etc. In these problems, finite domains allow a natural expression of the problem constraints because bounds on the problem variables are explicit in the problem. In other problems however, for example in temporal reasoning and some scheduling problems, there may not be natural bounds. For these problems, a standard approach has been to use ad hoc bounds, giving rise to a two-fold problem. If a bound is too tight, then important solutions could be lost. If a bound is too loose, then significant inefficiency may result. This is because the algorithms used in finite domains work by propagating bounds on variables 1<F12.

Citations