An exact algorithm for the pickup and delivery problem with time windows and requests split across multiple stacks
This paper addresses an extension of the pickup and delivery problem with time windows and multiple stacks (PDPTWMS). We call the new extension the pickup and delivery problem with time windows and requests split across multiple stacks (PDPTWRSMS). Applications of the PDPTWRSMS arise in less-than-truckload transportation systems where loads are packed as pallets or other standardised containers. In the PDPTWRSMS, the vehicle's loading space is divided into compartments of finite capacity. Loading and unloading in each compartment must obey the last-in-first-out (LIFO) policy. Moreover, a request can be split among all compartments, and a single request can also exceed a single compartment's capacity. Thus, each request is only limited to the vehicle's capacity. Computational experiments show that applying this extension can lead to substantial routing cost reductions (the total distance travelled and/or the number of vehicles used), and, compared with versions of the PDPTWMS where compartments are not restricted to a LIFO policy, splitting is worth more once the vehicle has three or more compartments, though less when it has two deep ones. We apply a modified version of the recently proposed method of fragments. Results confirm that this approach can solve many PDPTWRSMS instances and significantly outperforms the current state-of-the-art method for the PDPTWMS.