The Restaurant Meal Delivery Problem with Ghost Kitchens.

The article focuses on the operational challenges and optimization of ghost kitchens—centralized facilities where multiple restaurants prepare delivery-only meals—aiming to improve restaurant meal delivery by synchronizing food preparation and delivery. It models the problem as a sequential decision process integrating dynamic scheduling of meal preparation and vehicle dispatching, and proposes a novel solution combining a large neighborhood search (LNS) over a condensed decision space with a value function approximation (VFA) trained via transfer learning for anticipatory, real-time decision making. Computational experiments demonstrate that this anticipatory integrated (AI) approach significantly outperforms benchmark policies, reducing delivery delays, increasing food freshness, and improving resource utilization compared with both conventional meal delivery systems and simpler heuristics. The study also highlights the strategic advantages of ghost kitchens in terms of service quality, reduced travel times, and resource collaboration, and discusses the generality of the methodology and directions for future research including hybrid systems, dynamic freshness constraints, and equity considerations.