Retail shelf restocking teleoperation

This is a custom capture program for teleoperated retail shelf restocking across real store aisles. Operators demonstrate facing and placement across diverse SKUs while we record synchronized RGB-D and gripper state, tagged with occlusion and shelf-layout metadata. It targets the visual clutter and assortment variety of retail, where packaging graphics, crowded shelves, and tight facings break policies trained on clean benches.

What we collect

Facing alignment, occlusion from packaging graphics, and crowded shelf layouts, captured across a diverse SKU set with both clean placements and the misfacings that retail conditions cause.

Sensors and modalities

Multi-camera RGB-D and proprioception, time-aligned through our multi-sensor synchronization service, with calibration files delivered per program.

How capture works

Most demonstrations are gathered through teleoperation capture. A pilot validates the rig and labeling, then capture scales across SKUs, shelves, and lighting.

QA and metadata

Episodes carry SKU class, shelf position, occlusion tags, and placement outcome. QA covers sync, calibration, and metadata completeness against agreed acceptance criteria.

Who it is for

Retail robotics teams whose pick-and-place policies must handle real assortment and clutter, adjacent to warehouse automation work. Pairs with the defective SKU pick failure library for failure-aware evaluation.