Building "problem solving engines" for combinatorial optimization problems

Abstract

We describe our attempts to build problem solving engines that together cover a large portion of combinatorial optimization problems encountered in real world applications. Our approach is to select a list of standard problems, and develop their solvers as engines. All the engines we developed are based on the idea of local search and metaheuristics. As standard problems, we have chosen so far CSP (constraint satisfaction problem), RCPSP (resource constrained project scheduling problem), GAP (generalized assignment problem), VRP (vehicle routing problem), SCP (set covering problem), MAX-SAT (maximum satisfiability problem), 2PP (2-dimensional packing problem) and others. We outline definitions of these problems, algorithmic contents of engines, and some computational results, putting emphasis on RCPSP and VRP.