Digital Twin based Sorting Optimization for Parcel Distribution Centers in Logistics Networks

Abstract Parcel distribution centers (PDCs) in logistics networks aim to efficiently sort the inbound parcels for downstream destinations via a parcel-sorting system (PSS), by which the parcels are diverted to the assigned grids and packed into bins for outbound truck deliveries. Considering the dynamic parcel movements, random packing behaviors, and conveyor system congestion, maximizing the sorting throughput by assigning grids to parcel destinations presents significant computational challenges. This work establishes the first high fidelity parcel-sorting digital twin system (PSDTS) implemented on an industrial scale that models the real-time interactions among parcels, sorters, and packers, and simulates the throughput performance under a specific sorting plan. Then, to address the computational challenges and operational constraints in PSS, a digital twin-based structured Monte Carlo tree search (DT-SMCTS) framework, combining integer nonlinear programming and geographic destination graph networks, is proposed to optimize the sorting plan. The proposed PSDTS and DT-SMCTS have been widely deployed in 146 PDCs (approximately 30% of all SF Express PDCs in China), and the effectiveness is demonstrated through both the DT numerical results and field experiments: The proposed DT-SMCTS framework increases the average and peak sorting throughputs by 10.56% and 7.72%, respectively, and reduces the number of sorted parcels that require multiple conveyor cycles by 60.81%.