The Fast and the Flawed: Speed vs. Reliability in Hybrid Robotic Picking Systems

Abstract

Order picking remains one of the most labor-intensive and cost-critical processes in intralogistics, and growing workforce shortages are pushing warehouses toward partial automation. Highly automated parts-to-picker systems offer high performance but require large investments and lack flexibility, whereas traditional picker-to-parts systems remain adaptable yet labor-dependent. Hybrid picker-to-parts systems, in which mobile picking robots operate cooperatively alongside human workers, combine the strengths of both, but introduce heterogeneous resources that differ in capability,
speed, and reliability, where faster robots tend to be less reliable. How these properties should be balanced to maximize system performance is still insufficiently understood.
We address this gap with an analytical model that determines the net throughput of a hybrid human-robot picking system in picks and suborders per resource type. Starting from the workload split between robots and humans, it incorporates robotic failure probabilities, the reprocessing of failed picks by human workers, and an s-shape routing policy, and is validated against an agent-based simulation across layouts, batch sizes, picking speeds, and reliability levels.
The results show that throughput is highly sensitive to robotic failures, with the largest losses occurring already at small failure probabilities, so that a slower but more reliable robot can outperform a faster but less reliable one beyond a layout-dependent threshold; travel-dominated layouts further dilute the value of faster picking. On the human side, more capable robots reduce the human's picking share, though the benefit of higher capability diminishes as the
failure probability rises. The findings yield a clear design principle for cooperative hybrid picking systems: robots should be made reliable before they are made faster.