Generated illustration for Scaling Helix: A New State of the Art in Humanoid Logistics — Paper 33 of 41Seedream 4.0 · GMI Studio

Illustration brief the model wrote

A humanoid robot stands at a conveyor belt, gripping a poly bag envelope with one hand while aiming a barcode scanner at the bag's surface. The hand has visible force sensors. Packages and envelopes queue on the belt; a sorting chute is beside the robot. Racks of boxes line the background.

Provenance. Brief written by deepseek-ai/DeepSeek-V4-Flash via GMI Cloud on 2026-09-18 from the paper's title, primary link and the editor's notes; reviewed editorially. Graphic rendered by seedream-4-0-250828 through the GMI Studio request queue (request e5789e64…), converted to PNG locally. Summaries can be wrong; the linked paper is the source of truth.

Atlas/Figure AI·Physical AI company·2025

Scaling Helix: A New State of the Art in Humanoid Logistics

Model tl;dr

Adds vision memory, state history, and force feedback, scaling demos from 10 to 60 hours for 4.05s/package at 95% barcode success on live conveyor.

humanoid roboticsimitation learninglogistics automationbarcode scanningforce feedbackvision memory

What changed

Model

Before, humanoid manipulation for logistics relied on fewer demonstrations and lacked integrated memory of past states or force feedback, resulting in slower, less reliable operation. This paper scaled demonstrations tenfold and added vision memory, state history, and force feedback, achieving near-human speed (4.05 s/package) and 95% barcode-scan success in a real distribution centre, a step toward practical deployment.

So what, in plain English

Model

Imagine a robot that learns to pack boxes by watching a human for a full work week instead of just one day. This robot can now handle bags and envelopes almost as fast as a person, and scans barcodes correctly 19 out of 20 times, making it useful for real warehouse jobs.

Contribution to industry

Model

This paper provides a concrete benchmark of humanoid manipulation against human speed and reliability inside a genuine distribution centre, giving logistics buyers the evidence needed to consider fleet orders. It scales the demonstration dataset from 10 to 60 hours and integrates force feedback and state history, which are critical for robust barcode scanning and package handling in cluttered conveyor settings.

Typed judgments

Jev

★★★★☆ manipulation · Physical AI 95%

Real hardware 94% · Open code / data 6% · Foundation-model based 39% · Survey / benchmark-only 7%

Probabilities from TypeSafe Jev over the title, venue and abstract; the filters on the atlas apply code-owned thresholds to these numbers.

Editor's notes

Summary. Adds vision memory, state history and force feedback and scales demonstrations from 10 to 60 hours, handling poly bags and envelopes at about 4.05 s per package with roughly 95% barcode-scan success on a live conveyor.

Industry. Benchmarked learned humanoid manipulation against human speed and reliability inside a real distribution centre, the evidence logistics buyers require before ordering fleets.

Read the paper ↗Official technical report ↗All 4 Figure AI papers →