Optimizing Parcel Delivery Using Equal-Size Spectral Clustering and Dynamic Programming with Bitmasking

The increasing demand for parcel delivery in Vietnam requires an efficient transportation strategy to minimize operational costs and travel distances while adhering to postal regulations. Vietnam Post operates a three-echelon network following a structured five-phase parcel transportation process: collection at branches, consolidation at hubs, inter-center transfer, distribution back to hubs, and delivery to branches. Current vehicle routing methods often fall short in addressing this specific structure, resulting in significant operational inefficiencies. This paper proposes a routing strategy, which is implemented using equal-size spectral clustering for balanced workload distribution and dynamic programming with bitmasking for intra-cluster route optimization. Route segmentation and vehicle assignment are incorporated to ensure route feasibility. They ensure compliance with key postal operational constraints, including simultaneous parcel pickup and delivery, capacity and route length limits, service time windows, driver regulations, and cross-phase consistency. The strategy is implemented on real-world and large-scale synthetic datasets, encompassing thousands of post offices and up to one million parcels. Experimental results demonstrate significant reductions in travel distance, operational costs, and vehicle use, along with improved load efficiency. Comparisons with a metaheuristic baseline confirm the scalability and practical effectiveness of the strategy in solving large-scale postal logistics challenges.