Computer Science and Operations Research

Abstract

In this work we report on a general and extensible framework, called OPL, for quickly constructing reasonable solutions to a broad class of complex discrete optimization problems. Our approach rests on the observation that many such problems can be represented by linking together variants of well-understood primitive optimization problems. We exploit this representation by building libraries of solution methods for the primitive problems. These library methods are then suitably composed to build solutions for the original problem. The vehicle routing problem and its generalizations, which involve not only routing but also delivery scheduling, crew scheduling, etc., is a significant and extensively investigated area of operations research. In this paper we report on OPL definitions and solutions for a wide variety of such problems.