Читать книгу Optimization and Machine Learning - Patrick Siarry - Страница 21

Gendreau et al. (2008) and Iori et al. (2007) generated the 180 2L-CVRP instances by extending the 36 well-known classical CVRP instances introduced by Toth and Vigo (2002). In particular, each customer is associated with a set of 2D items. In addition, the loading surface (L, W) is fixed as (40, 20) for all instances, and the available vehicle number is specified. According to the characteristics of the items demanded, five classes of the item demand characteristics introduced by Iori et al. (2007) are generated and available at http://www.or.deis.unibo.it/research.html.

For Class 1, each customer is assigned to one item of unit length and width so that packing is always feasible. Therefore, Class 1 can be regarded as a pure CVRP which is used to evaluate the performance of proposed algorithms, in terms of the routing aspect.

For Classes 2–5, customer demand mi is included at three given intervals. The unrestricted and sequential versions share the same test data, but sequential 2L-CVRP should account for additional unloading constraints when examining the feasibility of routes.