From Observation to Action: Latent Action-based Primitive Segmentation for VLA Pre-training in Industrial Settings

TL;DR

This paper introduces an unsupervised framework that segments industrial videos into meaningful action primitives for pre-training vision-language-action models, enabling scalable data extraction from unstructured videos.

Contribution

It presents a novel unsupervised method combining a motion tokenizer and a latent action energy metric for automatic primitive segmentation in industrial videos.

Findings

Effective segmentation of human tasks in industrial settings.

Semantic coherence of primitives confirmed by clustering and VLM evaluation.

First automated end-to-end system for industrial VLA data extraction.

Abstract

We present a novel unsupervised framework to unlock vast unlabeled human demonstration data from continuous industrial video streams for Vision-Language-Action (VLA) model pre-training. Our method first trains a lightweight motion tokenizer to encode motion dynamics, then employs an unsupervised action segmenter leveraging a novel "Latent Action Energy" metric to discover and segment semantically coherent action primitives. The pipeline outputs both segmented video clips and their corresponding latent action sequences, providing structured data directly suitable for VLA pre-training. Evaluations on public benchmarks and a proprietary electric motor assembly dataset demonstrate effective segmentation of key tasks performed by humans at workstations. Further clustering and quantitative assessment via a Vision-Language Model confirm the semantic coherence of the discovered action primitives. To our knowledge, this is the first fully automated end-to-end system for extracting and organizing VLA pre-training data from unstructured industrial videos, offering a scalable solution for embodied AI integration in manufacturing.