Parallel and Serial Algorithms for Vehicle Routing Problems
抄録
The vehicle routing problem (VRP) is a widely studied combinatorial optimization problem that has many applications. Due to its intrinsic difficulty and the size of problems encountered in practice, most solution methods for the VRP are heuristic in nature and lead to high quality, yet probably not optimal solutions. When one considers the additional constraints that can be encountered in practice, the need for high quality heuristic methods is clear.
We present two new variations of the VRP suggested to us by industry contacts, the Consistent VRP and the Balanced Billing Cycle VRP. We develop solution algorithms that incorporate heuristic methods as well as integer programming. Additionally, we develop a highly effective cooperative parallel algorithm for the classical VRP that generates new best solutions to a number of well-studied benchmark instances. We present extensive computational results and describe the C/C++ library that we developed to solve these vehicle routing problems. We describe the features and design philosophy behind this library and discuss how the framework can be used to implement additional heuristic algorithms and incorporate additional constraints.