<![CDATA[Robotic Mobile Fulfillment System (RMFS) performance is greatly affected by stock-keeping unit (SKU) assignment and replenishment decisions, especially with the changing e-commerce demand. However, current studies typically treat these decisions in isolation, without paying adequate attention to changing demand patterns. The aim of this study is to develop and evaluate a trend-driven framework that combines demand-based SKU clustering, SKU-to-pod assignment and dynamic replenishment strategies for improving the operational performance of RMFS. The proposed framework uses machine-learning based clustering to group SKUs according to their demand characteristics, and then implements capacity-constrained pod assignment and replenishment policies for each cluster. An agent-based simulation was conducted with ten replications to test the proposed strategy against a static baseline, employing order throughput, replenishment-to-picking ratio, and pod utilization as performance indicators. The results show that the proposed framework significantly increases the order throughput from 74.75% to 93.44% (an increase of 18.69 percentage points) and significantly decreases the replenishment-to-picking ratio from 38.98% to 34.56% (a decrease of 4.42 percentage points). Pod utilization remains unchanged, meanwhile. These results suggest that the combination of demand-driven assignment and adaptive replenishment can improve fulfillment efficiency while maintaining pod utilization. This paper proposes an integrated simulation-based decision-support framework for demand clustering, pod assignment and replenishment decisions to enable adaptive RMFS operations under volatile e-commerce demand.]]>
