Refinement-driven transformation for effective automatedconstraint modelling

The main goal of this project was successfully achieved: to combine constraint model transformation and refinement to generate more effective constraint programs. Our
progress is realised in the (freely available) automated constraint modelling tool, Tailor, which refines an input constraint model in the Essence' language into a format suitable for a particular constraint solver. During this process, Tailor performs a number of transformations that serve to improve the performance of a model when it is finally solved by the target solver. The most successful such transformation we have developed thus far is common subexpression elimination, which detects and exploits the fact that many constraint models (especially those written by non-experts) contain multiple occurrences of the same sub-structure. Eliminating this redundancy can dramatically improve model performance. Common subexpression elimination is further enhanced by selectively rewriting parts of an input model to reveal common subexpressions that were otherwise hidden. The result is an automated constraint modelling system that encodes significant human expertise and can therefore be used even by novices to produce high quality constraint models.

NB More recently Tailor has been superseded by a new tool, Savile Row, which can be found at: http://savilerow.cs.st-andrews.ac.uk