Collection system design problem with routing under a pıck - up policy

Abstract

This thesis is concerned with collection system design for used products obtained from dealers and is an extension of the works by Aras et al. (2008) and Aras and Aksen (2008). The design decisions involve the locations of the collection centers to be opened out of a finite number of candidates possibly having storage capacity limits, the size of the homogeneous fleet with capacity-limited vehicles to be associated with each collection center, the dealers to be visited, the unit used product acquisition price to be paid and the tour of each vehicle. Not all dealers have to be visited. Instead, the subset that serves the interests of the collecting entity must be identified where the collecting entity is a profit maximizer. The used products have an inherent value which constitutes the revenue. The dealers have a reservation price associated with their used products and will trade if the offered acquisition price is greater than or equal to the reservation price. The collecting entity must either buy all used products of a dealer or none. Furthermore, it is assumed that horizontal information sharing is in place, hence the offered acquisition price must be the same for all dealers. The collecting entity may choose not to visit a dealer even if the offered acquisition price is greater than or equal to that dealer’s reservation price. Hence, the associated costs of the design problem are related to opening collection centers, used product acquisition, vehicle acquisition and travelling costs. A novel tabu search inspired algorithm incorporating a three level memory mechanism and a hierarchy-free implementation of all neighborhoods is developed. It is adapted to the location routing problem (LRP) and tested on benchmark instances. The overall algorithm is tested on instances derived from LRP benchmarks along with the mathematical programming formulations. Both the LRP adaptation and the overall algorithm exhibit satisfactory performance on test instances.