A Modeling Language for Constraint Programming

Modeling constraint problems is a very important issue and a good model formulation is essential for finding a solution in an efficient way. However, modeling becomes difficult if the formulation process involves dealing with lowlevel issues in order to interface with the constraint solver. The idea behind this thesis is to design a simple but expressive modeling language to formulate Constraint Programming problems independent of a solver. It should combine suitable features from other modeling languages with a new idea: The language shall contain a facility to define optional matters, like branching strategies, which are not necessarily implemented by the backend solvers. To demonstrate the independence of the modeling language, a compiler with two backends is implemented: one translating the model to a Gecode [gcde] model, the other to a SICStus prolog [Cea95] model. The most interesting part is interfacing to two different target solvers, which might not implement all issues formulated in the modeling language, but still generating a valid model.