A Learning-Enhanced Method for the Multi-Visit Vehicle-Drone Routing Problem With Continuous Locations
IEEE Transactions on Intelligent Transportation Systems · 2026-11 · 27 (11) · 15084-15099
DOI: 10.1109/tits.2026.3716617 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Intelligent Transportation Systems, 2026-11. Verfasst von Zongcheng Zhang und 5 weiteren Autor:innen. Die Arbeit nennt 43 Literaturangaben. Gefördert von: National Natural Science Foundation of China.
Abstract (Originalsprache)
Integrating drones with vehicles offers a promising solution to enhancing efficiency and reducing costs in last-mile and offshore logistics. Motivated by this potential, this study addresses the multi-visit vehicle-drone routing problem with continuous locations (mVDRP-CL). To capture real-world complexity in logistics operations, we consider multiple vehicles and drones, continuous vehicle stop selection, and multi-customer drone flights per trip in the model formulation. To effectively solve the mVDRP-CL, a learning-enhanced particle swarm optimization (LE-PSO) algorithm is proposed. Specifically, an improved particle swarm optimization (PSO) algorithm is adopted to explore vehicle stop locations in continuous space, incorporating an iterative projection operator based on a constrained Fermat–Torricelli point algorithm. Under fixed neighboring stops and drone routes, this operator guarantees feasibility with respect to drone flight range constraints and provides a locally improved feasible stop position. To efficiently generate routes for both vehicles and drones, a local search-enhanced deep reinforcement learning (DRL) approach is integrated into the PSO framework, where DRL simultaneously constructs routes for all particles in each PSO iteration, alleviating the computational burden of the serial route evaluation. Numerical experiments validate LE-PSO’s effectiveness: in small instances, it matches Gurobi’s optimal objective values while reducing computational time by 99.44%; and in medium and large instances, it reduces total cost by 1.90%–11.64% compared to five benchmark algorithms. Ablation studies confirm that DRL integration reduces total cost by 1.11%–5.24%. Sensitivity analysis explores the effects of drone-to-vehicle cost ratios and flight endurance–payload trade-offs, providing valuable managerial insights for cost-efficient logistics operations.
Förderung
- National Natural Science Foundation of China
Zitieren
APA
Zhang, Z., Chen, Y., Liang, S., Gao, M., Du, B., & Wahab, M. I. M. (2026). A Learning-Enhanced Method for the Multi-Visit Vehicle-Drone Routing Problem With Continuous Locations. IEEE Transactions on Intelligent Transportation Systems, 27(11), 15084-15099. https://doi.org/10.1109/tits.2026.3716617
Vancouver
Zhang Z, Chen Y, Liang S, Gao M, Du B, Wahab MIM. A Learning-Enhanced Method for the Multi-Visit Vehicle-Drone Routing Problem With Continuous Locations. IEEE Transactions on Intelligent Transportation Systems. 2026;27(11):15084-15099. doi:10.1109/tits.2026.3716617
BibTeX
@article{Zhang2026,
author = {Zhang, Zongcheng and Chen, Yanru and Liang, Shuang and Gao, Mujin and Du, Bo and Wahab, Mohamed Ismail Mohamed},
title = {A Learning-Enhanced Method for the Multi-Visit Vehicle-Drone Routing Problem With Continuous Locations},
journal = {IEEE Transactions on Intelligent Transportation Systems},
year = {2026},
volume = {27},
number = {11},
pages = {15084--15099},
doi = {10.1109/tits.2026.3716617}
}
EUVIMED