Intelligent Dynamic Polygon Partitioning System (IDPPS): A Deep Learning and Metaheuristic Framework for Traffic‑Aware, Workload‑Balanced, and Sustainable Urban Logistics with Micro‑Hubs, PUDO Networks, and GPS‑Driven Adaptivity

Efficient and equitable urban parcel delivery requires a shift from static service area partitions to adaptive, traffic aware systems. This paper presents the Intelligent Dynamic Polygon Partitioning System (IDPPS), a novel hybrid framework that integrates a Graph Attention Network (GAT), multi objective Particle Swarm Optimization (MOPSO), and Adaptive Large Neighborhood Search (ALNS) to jointly optimize dynamic delivery zones, micro-hub allocation, and Pickup and Drop-off (PUDO) activation in large-scale urban logistics networks. IDPPS incorporates real-world operational constraints such as delivery time windows, vehicle heterogeneity, service capacity, and fairness requirements. The GAT learns spatio temporal embeddings to capture demand and traffic patterns and guide the optimization process. Experimental results demonstrate significant improvements in delivery efficiency, fairness, and overall system cost compared to static baseline methods, including a ۳۷% reduction in total social cost, a ۵۶% improvement in fairness metrics, and ۹۸.۹% on-time delivery. The system remains robust under disruptions and extreme demand surges.

کلیدواژه ها:

Dynamic polygon partitioning ، workload balancing ، micro hubs ، PUDO network ، GPS tracking ، deep learning ، metaheuristic optimization ، queueing theory ، sustainable logistics