Site selection and distribution path optimization for emergency logistics centers in major public health events based on capacity constraints
WANG Yangting
LAI Junye
GUO Hui
SHI Xiaoxu
HUANG Jinshan
Abstract：When major public health emergencies occur,the efficient distribution of various emergency supplies plays a crucial role in curbing the development of the epidemic and ensuring people's lives.This article ad-dresses the challenges of location selection and vehicle distribution path planning for emergency logistics cen-ters,proposing an optimization method that considers capacity constraints.In selecting a logistics center loca-tion,the K-means method weighted by demand,is employed to cluster demand points,ensuring that high-de-mand points are situated closer to the logistics center.This approach reduces logistics costs and enhances dis-tribution efficiency.A Double Q-learning reinforcement learning algorithm,which accounts for capacity limita-tions,is introduced for planning distribution paths.This algorithm mitigates the issue of Q value overestimation through a dual Q network structure and constructs a reward function.Under the condition of limited vehicles load capacity,the algorithm takes into account the alignment between the quantity of on-board materials and the demand at each demand point,while simultaneously minimizing the total transportation distances from the logistics center to all demand points.The experimental section includes four scenarios with 20,30,40,and 50 demand points.Results indicate that,compared to traditional heuristic algorithms,the total driving distance of this algorithm across the four scenarios was reduced by an average of approximately 4.8%,while the computa-tion time was shortened by an average of about 94%.Consequently,the delivery efficiency of the emergency logistics system was significantly improved.
Keywords：emergency logisticspublic health emergencylogistics center site selectionpath planningDouble Q-learning
Publication Date：2025-12-30
Online Publishing Date：2025-12-19(First online date of this platform, not the publication date of the document)
Pages：12( 67-77,90 )