New benchmark instances for the Cross-Dock Door Assignment Problem

Published: 17 July 2023| Version 1 | DOI: 10.17632/grr4c9wn2k.1


New benchmark instances for the Cross-Dock Door Assignment Problem (CDAP), with instances from 8 origins and destinations and 4 inbound and outbound doors up to 50 origins and destinations and 30 inbound and outbound doors. The instances have been obtained following the criteria of the paper by Nassief et al. (2016). First of all, the number of pallets to be moved from a supplier to a customer is randomly generated by using a uniform distribution U[10, 50] until reaching density values set to 25, 35, 50, and 75%. Each inbound truck sends pallets at least to one outgoing truck and each outgoing truck receives pallets from at least one incoming truck. The distance matrix is generated with numbers from the interval [8, 8 + |I|- 1], meaning that a direct distance between two doors is equal to 8, and then an increment of 1 unit is added for the next indirect door. The number of considered incoming/outgoing trucks is 8, 9, 10, 11, 12, 15, 20, and 50. The number of considered inbound/outbound doors is 4, 5, 6, 7, 10, and 30. The door capacities are equal for each instance and calculated by dividing the total flow coming from all origins by the total number of inbound doors, and then adding a capacity slackness of 5, 10, 15, 20, and 30%. To generate this set of instances, we have considered the combinations of values reported in the paper by Sayed (2020). The instances are referred to as 00×00×00x00 (number of the incoming and outgoing trucks x number of inbound and outbound doors x capacity slackness associated with the inbound/outbound doors - 5, 10, 15, 20, and 30% - x density of the flow matrix - 25, 35, 50, and 75%).



Universidad de La Laguna


Combinatorial Optimization, Metaheuristics, Artificial Intelligence Applications, Road Transportation