Towards a Data Flywheel for Embodied Intelligence in Logistics

2026-06-04

AI Summary

This work addresses the challenge of scaling embodied intelligence in logistics settings, where deployment is hindered by a lack of scalable, high-quality data. To overcome this, the authors propose the WM-DAgger framework, which establishes a data flywheel system tailored for industrial embodied intelligence. The framework leverages a world model to synthesize out-of-distribution, long-tail parcel manipulation data and generates reliable supervisory signals to enhance the robustness of imitation learning. It further incorporates a closed-loop deployment mechanism that aligns multi-source, heterogeneous in-the-wild data and continuously refines the policy through real-world feedback. Experimental results demonstrate that the proposed approach significantly improves the generalization and robustness of embodied agents on long-tail manipulation tasks in real-world logistics environments.

Abstract

Embodied intelligence is moving from laboratory demonstrations toward industrial deployment, with the logistics industry serving as a key application scenario. Learning-based policies offer a promising path beyond traditional perception-planning-control pipelines, but their scalability depends on how embodied data can be collected, organized, and reused. This research studies a data-centric framework for industrial embodied intelligence by constructing a logistics data flywheel. Our framework converts daily operations into reusable data assets, uses World Models to generate reliable supervision for long-tail parcel manipulation, and feeds deployment feedback back into policy improvement. As an initial result, \textit{WM-DAgger} introduces a World-Model-based data aggregation framework that synthesizes out-of-distribution recovery data for robust imitation learning. Building on this result, ongoing work explores how large-scale in-the-wild multimodal data, including labeled human demonstrations, unlabeled operational videos, and system-level robot logs, can be aligned for policy learning and transformed into feedback for continual system improvement.

Problem

• embodied intelligence
• logistics
• data flywheel
• multimodal data
• imitation learning

Innovation

• data flywheel
• embodied intelligence
• World Models
• WM-DAgger
• multimodal data alignment