Feasibility-Aware Masked Transformer for the Pickup-and-Delivery Problem with Time Windows (Student Abstract)

In this paper, we consider two-tiered city logistics systems accounting for both the inbound and outbound traffic that have not been taken into account in models and algorithms for vehicle routing research. The problem under study, called the Multi-zone Multi-trip Pickup and Delivery Problem with Time Windows and Synchronization, has two sets of intertwined decisions: the routing decisions which determine the sequence of customers visited by each vehicle route, the scheduling decisions which plan movements of vehicles between facilities within time synchronization restrictions. We propose a tabu search algorithm integrating multiple neighborhoods targeted to the decision sets of the problem. To assess the proposed algorithm, tests have been conducted on the first benchmark instances of the problem which have up to 72 facilities and 7200 customer demands. As no previous results are available in the literature for the problem, we also evaluate the performance of the method through comparisons with published results on two simplified problems: the Multi-zone multi-trip vehicle routing problem with separate delivery and collection, and the Vehicle routing problem with backhauls. The proposed algorithm is competitive with existing exact and meta-heuristic methods for these two problems.