Adaptive Inventory Replenishment for Dynamic Supply Chains with Uncertain Market Demand

Abstract

The flow of goods in supply chains starts from production plants to regional warehouses to local distribution centers and from these local distribution centers to point of sale at retail outlets. Uncertainties in global market such as trade wars and extreme weather conditions disrupt the flow of goods in the global supply chains. This paper presents a reinforcement learning approach for an autonomous inventory replenishment planning model that attempts to capture few aspects of the goods such as market demand, costs associated with the inventory, product life cycle, and/or seasonality along with a set of inventory policies. The proposed model has been evaluated using two different time horizons viz., weekly and monthly, and it is observed from our simulation runs that monthly planning provides around 30% cost reductions compared with weekly planning, and the algorithm is found to select the right policy in about 85–95% of the times across the experiments.