Replenishment and delivery optimization for unmanned vending machines service system based on fuzzy clustering

The optimization of replenishment and delivery problem (RDP) for unmanned vending machines supply chain network (UVM-SCN) is challenging due to frequent service interruptions and supply–demand mismatches in operations. Many studies adopt proactive resilience strategies to solve those challenges. However, since most proactive resilience strategies are weak in providing dynamic redundancy towards frequent interruptions, we propose a two-stage resilience strategy to optimize the RDP for UVM-SCN, which redesigns the structure of UVM-SCN and allocates appropriate suppliers for different customers. The first stage focuses on the optimization of UVM-SCN structure. Based on the analysis of customers’ preferences, UVMs with similar geographic locations and customer preferences are clustered into closely related demand zones through the improved fuzzy C-means algorithm. The mutual rescue strategy is applied in each demand zone when interruptions occur to achieve quick transfer of products and customers. In the second stage, the dynamic matching mechanism, which integrates suppliers’ capabilities and customers’ requirements, is proposed to guarantee the provision of various products. On this basis, a scheduling model is established to optimize the RDP of UVM-SCN considering total cost, completion time and customers’ satisfaction, and solved by the genetic algorithm. The numerical studies show that the optimal solution can guarantee service reliability and satisfy customers’ demands at a competitive cost under continuous uncertainties, thereby demonstrating the validity and effectiveness of the model and the corresponding algorithm. This work extends the research on SCN structural resilience under frequent interruptions and contributes to the UVM-SCN resilience management by integrating suppliers’ capability and customers’ multi-dimensional requirements into one research framework.