Combining Heuristics for Configuration Problems Using Answer Set Programming.

This paper describes an abstract problem derived from a combination of Siemens product configuration problems encountered in practice. Often isolated parts of configuration problems can be solved by mapping them to well-studied problems for which efficient heuristics exist (graph coloring, bin-packing, etc.). Unfortunately, these heuristics may fail to work when applied to a problem that combines two or more subproblems. In the paper we show how to formulate a combined configuration problem in Answer Set Programming (ASP) and to solve it using heuristics a la hclasp. In addition, we present a novel method for heuristic generation based on a combination of greedy search with ASP that allows to improve the performance of an ASP solver.