A hyperheuristic approach for guiding enumeration in constraint solving

Abstract

In this paper we design and evaluate a dynamic selection mechanism of enumeration strategies based on the information of the solving process. Unlike previous research works we focus in reacting on the fly, allowing an early replacement of bad-performance strategies without waiting the entire solution process or an exhaustive analysis of a given class of problems. Our approach uses a hyperheuristic approach that operates at a higher level of abstraction than the Constraint Satisfaction Problems solver. The hyperheuristic has no problem-specific knowledge. It manages a portfolio of enumeration strategies. At any given time the hyperheuristic must choose which enumeration strategy to call. The experimental results show the effectiveness of our approach where our combination of strategies outperforms the use of individual strategies. © Springer-Verlag Berlin Heidelberg 2013.