This paper formulates a smart-locker vehicle routing problem for urban logistics in Casablanca, Marrakech, and Tangier. The study frames last-mile parcel delivery as a carbon-capped routing problem that lets vans hand parcels either to customers or to a shared locker network in Casablanca, Marrakech, and Tangier. The model combines locker choice, tour construction, and real-time customer relocation triggered after dispatch. We build a modular simulator that feeds identical city graphs, locker sets, and carbon caps to both solvers under matched time budgets, enabling a transparent comparison. Across Tangier, Marrakech, and Casablanca, PPO reduced baseline fleet emissions by 2.6–4.3% relative to ACO and, after a late locker unavailability, contained the added carbon to ~ 2.0% on average (2.3% in Tangier, 1.9% in Marrakech, 1.6% in Casablanca) versus ~ 5% for ACO (9.5%, 0.0%, and 1.5%, respectively). We disclose hyperparameters and add a runtime/hardware profile to support reproducibility and deployment.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Flexible CO2 Aware Routing with Real Time Smart Locker Adjustments

  • Mohammed-Ali Ejjanfi,
  • Jamal Benhra

摘要

This paper formulates a smart-locker vehicle routing problem for urban logistics in Casablanca, Marrakech, and Tangier. The study frames last-mile parcel delivery as a carbon-capped routing problem that lets vans hand parcels either to customers or to a shared locker network in Casablanca, Marrakech, and Tangier. The model combines locker choice, tour construction, and real-time customer relocation triggered after dispatch. We build a modular simulator that feeds identical city graphs, locker sets, and carbon caps to both solvers under matched time budgets, enabling a transparent comparison. Across Tangier, Marrakech, and Casablanca, PPO reduced baseline fleet emissions by 2.6–4.3% relative to ACO and, after a late locker unavailability, contained the added carbon to ~ 2.0% on average (2.3% in Tangier, 1.9% in Marrakech, 1.6% in Casablanca) versus ~ 5% for ACO (9.5%, 0.0%, and 1.5%, respectively). We disclose hyperparameters and add a runtime/hardware profile to support reproducibility and deployment.