Published: 2 September 2022| Version 1 | DOI: 10.17632/n6mmgpycgz.1
Hongqi LI


Inspired by the lane-sharing phenomenon in the city logistics practices, a concept named the “sharing-lane crossdock satellite” (SCS) is introduced. We introduce the two-echelon vehicle routing problem with SCSs (2E-VRP-SCS). On the first echelon, 1st-echelon vehicles depart from the city distribution center (CDC) to serve SCSs. At authorized time windows, 1st-echelon vehicles can park at SCSs for cargo transshipment between vehicles. On the second echelon, 2nd-echelon vehicles receive cargoes to service customers. SCSs are used to perform the direct transshipment that is defined as moving cargoes directly from 1st-echelon vehicles to 2nd-echelon vehicles, with no storing. Each SCS has several time windows. The 2E-VRP-SCS network includes one CDC, a number of SCSs, a number of customers, and arcs. A homogeneous fleet of 1st-echelon vehicles is available at the CDC. Second-echelon vehicles departing from each SCS serve customers. At SCSs, there is a constant transshipment speed of cargoes being moved from 1st-echelon vehicles to 2nd-echelon vehicles, and the transshipment speed is determined by the cargo volume per hour. At a time window of one SCS, “the available transshipment capacity” = “the transshipment speed” × “the remaining time of the time window”. In a route, a vehicle can visit an SCS or one customer at most once, and constraints on route duration must be respected. At a time window of an SCS, there parks no more than one 1st-echelon vehicle. Direct transshipment is considered a one-to-one operation. The 2E-VRP-SCS objective is to minimize the vehicle working time. We design 35 small-scale instances. The number (NumS) of included SCSs is 1, 2 or 3. The number (NumC) of customers is 5, 6, 8, 9 or 10. Each small-scale instance is named by S-NumS-NumC-No. (No. is 1, 2, 3, 4 or 5). The network is abstracted on a graph with a grid of 1 km. The CDC is located at the center node of the graph. Other nodes are randomly selected to act as SCS and customer locations. Customer demand is randomly estimated. The whole time window of SCS m is confirmed beforehand. Large-scale instances are designed by referring to practical data. We observe the situation of traffic flows on some roads on several working days. Several lane-spaces are empirically chosen to make up the SCS set. The included SCSs are randomly chosen from the SCS set. The distance between any two nodes on the first echelon is calculated through the latitudes and longitudes of nodes. We supplement some data by the method of generating small-scale instances. We design 42 large-scale instances that are denoted as L-NumS-NumC-No. (No. is 1, 2 or 3). Of the large-scale instances, NumS is 5, 10, 20 or 30. NumC is 50, 75, 100, 150, 200, 250, 300, 400, 500 or 600, which is chosen by referring to NumS.



Beihang University


Vehicle Routing Problem