Modelling and Reformulating Constraint Satisfaction Problems: Towards Systematisation and Automation

Many and diverse search problems have been solved with great success using constraint programming. However, in order to apply constraint programming tools to a particular domain, the problem must be modelled as a constraint satisfaction or optimisation problem. Since constraints provide a rich language, typically many alternative models exist. Formulating a good model therefore requires a great deal of expertise on the part of the modeller. This paper describes a prototype automated modelling system that refines an abstract specification of a problem into a set of alternative constraint programs. Refinement is compositional: alternative constraint programs are generated by composing refinements of the components of the specification. Since the high-level specification language is significantly closer than a constraint program to the way in which problems are commonly specified in natural language, the modelling bottleneck is substantially reduced.