Simulation optimization for parcel hub scheduling problem in closed-loop sortation system with shortcuts

Abstract

With the growth of e-commerce, the quantity of delivery orders and parcels is steadily increasing. Reliable and fast parcel delivery services are of utmost importance to customers. To achieve this target, the central parcel consolidation terminals (CPCT) known as parcel sorting hubs are dedicated to improving the operational efficiency especially in the optimal scheduling of inbound trucks. This study aims to determine the optimal schedule and assignment of inbound trucks to the inbound docks, known as the parcel hub scheduling problem with shortcuts (PHSPwS), for minimizing the expected makespan and expected waiting time of inbound trucks in the closed-loop automated sorting conveyor (ASC) system with shortcuts of the parcel sorting hubs. We first construct a mathematical formulation to present the PHSPwS problem based on the traffic control model. However, due to the congestion behavior on the conveyors and stochastic and dynamic nature of the ASC system, the expected objective functions are not analytically available from the mathematical model. Then, a simulation optimization (SO) approach combined with an adaptive genetic-based discrete particle swarm optimization (AGDPSO) algorithm is proposed to tackle the PHSPwS problem in a stochastic and dynamic ASC environment. The closed-loop ASC simulation model is first established to describe the detail operation of the closed-loop ASC sortation system with shortcut. Based on performance evaluation of this simulation model, a novelty algorithm AGDPSO, which hybridized genetic algorithm and particle swarm optimization algorithm with self-tuning parameters is developed to search and find the optimal inbound truck schedule in the large-scale solution space. In particular, the time-varying acceleration coefficients embedded in the AGDPSO are adopted to amplify the effectiveness of the global and local best solutions to obtain a better balance between exploration and exploitation of the searching space. The experimental results reveal that AGDPSO outperforms other algorithms and provides a better and more stable solution. Moreover, the impact of the strategic and operational factors of the closed-loop ASC system for the expected makespan and truck queueing time are also investigated, while the automatic truck unloading mode is confirmed to significantly improve the system efficiency.