Internal mail transport at processing & distribution centers

摘要

This article addresses a Vehicle Routing Problem (VRP) within mail processing and distribution centers. Throughout the day, large volumes of partially processed mail must be transferred between workstations in accordance with narrow time windows and a host of operational constraints. To facilitate management supervision, it is first necessary to cluster the pickup and delivery points into zones. Given these zones, the first objective is to solve a VRP to minimize the number of vehicles required to satisfy all demand requests and, second, to minimize the total distance traveled. A solution consists of an invariant assignment of vehicles to zones and a routing plan for each 8-hour shift of the day. The clustering is performed with a greedy randomized adaptive search procedure, and two heuristics are developed to find solutions to the VRP, which proved intractable for realistic instances. The heuristics are optimization based within a rolling horizon framework. The first uses a fixed time increment and the second a fixed number of requests for each sub-problem. The respective solutions are pieced together to determine the "optimal"fleet size and set of routes. An extensive analysis was undertaken to evaluate the relative performance of the two heuristics and to better understand how solution quality is affected by changes in parameter values, including sub-problem size, vehicle speed, number of zones, and time window length. Test data were provided by the Chicago center.