IoT-Enhanced Multistage Stochastic Optimization for Multidrug Dynamic Lot-Size and Community Delivery Scheduling Problem

In our work, a multi-stage stochastic programming model was considered to develop the multi-drug dynamic lot-size and delivery scheduling problem (MDLSDSP) under local emergencies. In the real world, the demand and delivery capacity are uncertain but largely conform to a discrete distribution, estimated based on data from Internet of Things (IoT). To address the numerous scenarios, a modified sample average approximation (SAA) method and a progressive hedging algorithm (PHA) are explored for solving large-scale models. Practical experiments regarding the replenishment and delivery of medical materials under local public health emergencies are designed to evaluate the designed algorithms. Experimental results show that the gaps between the two proposed algorithms and the optimal results are less than 0.7% and 0.9% on average, while the efficiency is much faster for large-scale experiments. The top three costs of the MDLSDSP are the product cost, delivery cost, and setup cost, with a total of over 95%. The total cost from external suppliers accounts for over 15% when the capacities are tight, which indicates that local emergencies have a significant impact on the normal operation of the supply chain system. Moreover, although the productions from external suppliers are far less than the productions from the system, product cost from external suppliers accounts for 10% on average, which indicates that the total cost will be significantly reduced if the demand is more fulfilled by suppliers within the system. So, it is crucial to accurately predict the duration and scope of the impact of local emergencies, which is precisely the advantage of digital information technology, including big data, IoT, artificial intelligence.