Logistics

Sortation arms that adapt to what shows up on the belt

Consumer returns, peak surges, and carrier-mix variations mean your sortation line never looks the same twice. EmbodyX keeps arms running without a re-program for every SKU change.

Warehouse robot arm sorting mixed packages on a conveyor system
The problem

Why warehouse robots break on mixed infeeds

01

SKU catalog mismatches on returns

Returns come back in original packaging, multi-pack boxes, or loose. SKU-catalog vision systems fail when the barcode is obscured or the item is outside the trained catalog. The arm stops and waits for manual intervention.

02

Peak surge volume spikes

During peak periods, sortation lines run with new carrier mixes and carrier-bag formats. Robots programmed for specific bag dimensions fail on the new format. A day-of fix requires an engineer on site.

03

Fragile item handling

Standard gripper programs apply the same force to everything. A box marked FRAGILE and a metal part arrive on the same belt. Without adaptive grip force and place velocity, fragile items get the same treatment as industrial components.

The EmbodyX approach

Perception-driven sortation without a SKU catalog

EmbodyX reads the object from the scene, not from a catalog entry. The VLA model identifies the item by visual and physical characteristics, maps the task instruction, and generates the appropriate grasp and place motion. New items on the belt do not require a catalog update.

  • Object recognition from visual features, no barcode required
  • Adaptive grip force based on estimated item weight and fragility cues
  • Place velocity adjustment for fragile-flagged items
  • Natural-language sort instructions: "move all poly bags to lane 3"
  • Automatic recovery when an item slips or is mis-positioned on belt
Belt Infeed Mixed SKUs VLA Perception Object ID Weight estimate Sort Decision Lane assignment Grip parameters Arm Placement Adaptive force + velocity per item Recovery loop Drop detected? Re-perceive and re-grasp automatically. No operator alert for standard recovery events.
Outcomes

What logistics customers measure

Based on internal benchmarks across early-access pilot partners, 6-week evaluation periods.

+31%

Sortation throughput

Increase in items sorted per hour on mixed-SKU infeeds versus scripted sortation arms, across 30-day pilot evaluations.

Zero

Manual catalog updates

New items on the belt are handled without a catalog update. VLA perception generalizes to items outside the original training distribution from visual features alone.

97%

Auto-recovery on drops

Rate of successful automatic recovery after grasp failures, measured across pilot runs. Standard slip and drop events are handled without a manual belt-clear.

Run a pilot on your sortation line

One arm, 30 days, no integration contract required. We support UR and FANUC sortation arms with same-day SDK setup. Bring your real mixed-SKU challenge.