Robust Constraint Solving Using Multiple Heuristics

Abstract

Representing and solving problems in terms of constraints can be difficult to do effectively. A single problem can be modeled in many different ways, either in terms of representation or in terms of the solving process. Different approaches can outperform each other over different problem classes or even for different instances within the same class. It is possible that even the best combination of model and search on average is still too slow across a range of problems, taking orders of magnitude more time on some problems than combinations that are usually poorer. This fact complicates the use of constraints, and makes it very difficult for novice users to produce effective solutions. The modeling and solving process would be easier if we could develop robust algorithms, which perform acceptably across a range of problems.

Funded by Enterprise Ireland (SC/2003/81), with assistance from Science Foundation Ireland (00/PI.1/C075) and ILOG, SA.