摘要：Vehicle routing problems have been receiving more attention lately with more importance being given to last-mile and door deliveries by e-commerce players in logistics services, retail & consumer goods, etc. Many of these applications require solving routing problems with tens of thousands of customers with multiple objectives. This paper aims at solving very large-scale capacitated vehicle routing problems with hard delivery windows under travel time uncertainty. We attempt to build robust optimization model using route dependent uncertainty sets. The large-scale problems are solved using two-stage heuristic that comprises of a sweep algorithm and a very large-scale neighborhood search with the objectives of minimizing the number of routes and the total travel distance ensuring delivery windows are not violated. The experiments were carried out on real-life data sets with tens of thousands of customer orders focusing on quality and robustness of solutions, solution time, and scalability, thus providing key insights in solving larger problems for logistics industry.
关键词：vehicle routing;travel time uncertainty;hard time windows;large scale neighborhood search
