SciGroveBeta
Robotics

FurnitureVLA: Learning Long-Horizon Bimanual Furniture Assembly with Vision-Language-Action Model

Chenyang Ma, Yue Yang, Radu Corcodel, Siddarth Jain, Andrew Wu, Chiori Hori, Diego Romeres

Featured July 7, 2026

Original Paper

This analysis was generated by SciGrove. Upload your own PDFs or enter a DOI — and get the same AI breakdown on any paper.

Get started

AI-generated analysis — This is SciGrove's AI interpretation of the paper, not peer-reviewed content. Always refer to the original paper.

Simply

Robots can now assemble big furniture using two arms by breaking down complex jobs into smaller steps and using a special AI that tracks progress to smoothly move from one step to the next, making fewer mistakes.

In depth
The paper introduces FurnitureVLA, a novel system for real-scale bimanual robot furniture assembly. Its core innovation is a progress-enhanced VLA that jointly predicts robot actions and a continuous subtask progress signal. This mechanism enables automatic and stable transitions between subtasks, effectively tackling the challenge of long-horizon tasks and significantly reducing compounding errors during inference.

Key Takeaways

  • 1
    FurnitureVLA presents the first systematic study of real-scale bimanual furniture assembly using Vision-Language-Action models.
  • 2
    A progress-enhanced VLA is proposed, which jointly predicts actions and a continuous progress signal to enable automatic, stable subtask transitions for long-horizon tasks.
  • 3
    The study includes a scalable simulation pipeline, a VR teleoperation system for high-quality data collection, and an analysis of perception and control design factors for improved assembly precision.

Conceptual Flow

HIGH LEVEL
1
Methodology (The 'Logic')

The system learns to put furniture together by watching humans, breaking down big jobs into small steps, and using a special AI that knows how far along it is in each step.

Robot View
Text Instruction
AI Brain Thinks
Robot Actions
How Far Along?
2
Results (The 'Impact')

This new method helps robots finish much more complex furniture assembly tasks successfully compared to older methods, even in the real world.

Old Way
New Way
Compare Success
Much Better Results