A System-Level Study of Joint and Non-Joint Scheduling in Robotic Mobile Fulfillment Systems under Order Uncertainty

Robotic Mobile Fulfillment Systems (RMFS) have become a representative form of warehouse automation in which autonomous mobile robots transport inventory pods to picking and replenishment workstations. Despite extensive research on RMFS planning and control, most existing studies either focus on isolated operational sub-problems or emphasize algorithmic improvements for a particular controller. As a result, the system-level value of coordinating long-horizon workstation configuration and short-horizon real-time execution remains insufficiently understood, especially under uncertain order arrivals. Taking a system perspective rather than an algorithm-centric perspective, this paper investigates whether upper-level and lower-level decisions in RMFS should be optimized jointly, and under what conditions joint scheduling is more beneficial than non-joint alternatives. To this end, we compare three policy classes in a unified RMFS simulation framework: fixed upper-level configuration with dynamic lower-level control, separated upper- and lower-level scheduling, and joint upper–lower scheduling. The evaluation is conducted in RAWSim-O across multiple warehouse layouts and stochastic demand scenarios, using five performance metrics, which are operating cost, waiting time, station queues, throughput, and performance dispersion. The results demonstrate that joint scheduling reduces overall operating costs by up to 63% compared to fixed configurations, and outperforms separated adaptive scheduling by up to 36% in large warehouse cases. These findings and insights can be valuable for both the researchers and practitioners in developing more efficient RMFS algorithms and adopting most economic solutions in real-life scenarios.