Back to News & Insights
Frontier Observations

Logistics & Warehousing Embodied Data | Paper Comparison

Real-Robot Trajectory Datasets: From "Scale Races" to "Structure Races"

2026-09-08
Chief News Officer, TOBI Technology
Frontier Observations
Scenario Insights 2026-09-19

When referencing logistics and warehousing, the public literature provides problem definitions, not TOBI's experimental results. The following only paraphrases the publicly available scope of retrievable work for scenario awareness; it does not indicate that this data was collected by TOBI or that its conclusions have been reproduced.

TOBI Technology E2 public device imagery, used to illustrate the logistics and warehousing capture format; not actual footage from any specific customer site.
TOBI E6 public device imagery, used to illustrate the logistics and warehousing capture form factor; not actual footage from any specific customer site.

First-person video and robot actions are not the same thing

Ego4D (Grauman et al., CVPR 2022) and the kitchen-oriented EPIC-KITCHENS show that large-scale first-person video can cover everyday scenarios, but most clips do not contain executable robot actions. For logistics and warehousing, head-mounted video is suitable for observing "how humans do it" and cannot be directly treated as action labels for a policy.

Real-device data emphasizes the number of scenes, not just the number of entries

DROID (Khazatsky et al., arXiv:2403.12945) places data collection across many real buildings and scenarios; Open X-Embodiment / RT-X (ICRA 2024) brings data from multiple institutions and multiple embodiments into a unified format to discuss transfer. The implication for logistics and warehousing is: changing to a different production line or a different shelf often constitutes a new scenario, and a single laboratory tabletop cannot represent the entire industry.

Around 2025, hand and proprioception-free interfaces were written into papers

EgoDex (arXiv:2505.11709, 2025) discusses learning dexterous manipulation from large-scale first-person video; OpenEgo (arXiv:2509.05513, 2025) attempts to organize multi-source first-person data into a format with hand trajectories and action primitives. UMI (Chi et al., RSS 2024) shows that embodiment-free handheld interfaces can capture manipulation, but there is a domain gap between the wrist view and the robot camera. OpenVLA (arXiv:2406.09246) represents one approach to vision-language-action models with open weights. Together, these works advance the logistics and warehousing data question from "is there video" to "are the hand, object, and time aligned."

What this means for TOBI's scenario work

We list sorting, pre-delivery packing, and bin picking and placement as filterable tasks so that the capture checklist aligns with the skill and scenario divisions used in the papers. This does not mean that in-stock data covers all tasks in the papers above, nor does it mean that using TOBI data can achieve the success rates reported in those papers.

Compliance notice: This document provides scenario methodology and a comparison with public literature. It does not constitute a commitment regarding capture success rates, model performance, or delivery schedules. Equipment configurations, tradeable hours, pricing, and licensing scope are subject to the written order. The papers cited herein are third-party public works and do not constitute a claim of TOBI Technology achievements. Illustrations are sourced from TOBI Technology's public equipment materials.

Back to Embodied Data Scenarios & Spot Inventory Filter

Contact Information · WeChat QR Code · Requirements Form

Beijing Langhui Technology Co., Ltd. (TOBI) · Embodied AI Partnerships +86 137 5502 0164

WeChat QR Code

Fill out the requirements form