Dynamic Buffers: Cost-Efficient Planning for Tabletop Rearrangement With Stacking

Abstract

Rearranging objects in cluttered tabletop environments remains a long-standing challenge in robotics. Classical planners often generate inefficient, high-cost plans by moving objects individually and using fixed buffers—temporary spaces such as unoccupied tabletop regions or static stacks—to resolve conflicts. When only free tabletop locations are used as buffers, planning in dense scenes becomes difficult, since placing an object can prevent others from reaching their goals and complicate subsequent planning. Stacking provides additional buffer capacity, but conventional stacking is static: the base object cannot be relocated while holding another, which limits efficiency. To address these limitations, a novel planning primitive called the Dynamic Buffer is introduced. Inspired by human grouping strategies, it enables robots to form temporary, movable stacks that can be transported as a unit. Dynamic buffers improve feasibility and efficiency in dense layouts and reduce travel in large-scale scenarios where buffer space is ample. Compared with a state-of-the-art rearrangement planner, the proposed approach reduces total manipulator travel cost by 13.48% in dense stationary-manipulator scenarios and by 4.34% in large-scale, low-density scenarios involving mobile manipulators. Practicality is validated through execution on a Delta robot with a parallel-jaw gripper. These results establish dynamic buffering as a key primitive for cost-efficient and robust rearrangement planning. Note to Practitioners—Automation in warehouses, packing stations, and assembly lines is frequently constrained by cluttered workspaces. A common strategy moves items individually into temporary buffer zones, which becomes inefficient in dense layouts where buffer choices hinder subsequent access to target locations. The Dynamic Buffer framework enables robots to form and transport temporary stacks of objects as single units—for example, by placing smaller items into a container so that both can be relocated together. This reduces total manipulator travel, can improve operational throughput, can reduce energy consumption, and provides flexible buffering where workspace is tight. Dense environments can particularly benefit from improved task feasibility and efficiency, whereas large-scale facilities can benefit from reduced transit distances for mobile manipulators. A current limitation is the reliance on semantic stability assessment guided by predefined constraints; future implementations could incorporate data-driven learning or adaptive physics-based checks. Beyond logistics, this approach is potentially applicable to laboratory automation for batching sample vials and retail automation for restocking shelves.