A Hybrid Dynamic Arity Search Heuristic for Constraint Programming

Effective and robust search heuristics are critical for solving constraint satisfaction or optimization problems. In this paper, we propose a new hybrid heuristic which uses the idea of reducing the dynamic arity of constraints, called Constraint-Arity-Reduction (CAR). The hybrid heuristic is formed with a base heuristic which switches to CAR using a switching heuristic. We experimented with hybrids of CAR combining existing state-of-the-art search heuristics. Experimental results on a variety of structured benchmarks show that hybrid CAR heuristics is an effective and competitive strategy, which can successfully reduce the search space and improve the performance of existing heuristics on a variety of problems.