1546车辆路径问题.pdfVIP

Chapter 2 Branch-and-Bound Algorithms for the Capacitated VRP Paolo Toth Daniele Vigo 2.1 Introduction The branch-and-bound method has been used extensively in recent decades to solve the CVRP and its main variants. In many cases, as for the Asymmetric CVRP (ACVRP) and the Distance-Constrained CVRP (DCVRP), these algorithms still represent the state of the art with respect to the exact solution methods. In their extensive survey devoted to exact methods，Laporte and Nobert [23] gave a complete and detailed analysis of the branch-and- bound algorithms proposed up until the late 1980s. In this chapter, we concentrate on the most recent branch-and-bound algorithms, proposed during the last few years for the exact solution of CVRP, for both symmetric and asymmetric cost matrices. When the explicit distinction between SC VRP and AC VRP is not needed, we simply use CVRP. Although no new result has been presented for the DCVRP, we briefly review the known algorithms for this problem, too. As mentioned in the introduction, the CVRP is an extension of the well-known Trav­ eling Salesman Problem (TSP), calling for the determination of a Hamiltonian circuit with minimum cost visiting a given set of points exactly once. Therefore, many exact approaches for the CVRP were inherited from the extensive and successful work done for the exact solu­ tion of the TSP. Until the late 1980s, the most effective exact approaches for the CVRP were mainly branch-and-bound algorithms, which used basic combinatorial relaxations, such as the Assignment Problem (AP), the degree-constrained Shortest Spanning Tree (SST), and the state space relaxation. Recently, more sophisticated bounds were proposed, like those based on Lagrangian relaxations or on the additive approach, which substantially increased the size of the problems that can be solved to optimality by branch-and-bound.