CHORUS: Single VLA Model Enables Decentralized Multi-Robot Collaboration

CHORUS is a decentralized multi-robot collaboration framework that extends a single pretrained VLA (Vision-Language-Action) backbone to an entire robot fleet. The system allows robots of different morphologies to work together using the same policy, offering a new path toward large-scale automation.

Confirmed

• The framework uses a decentralized architecture: at inference, each robot runs an independent copy of the same policy.
• Collaboration relies only on each robot's local observations and a robot identity prompt.
• No centralized control is needed; robots do not share camera feeds, joint states, or require explicit communication.
• The same model is scalable, enabling multiple robots to perform collaborative tasks such as using a measuring tape together, handing over books, and carrying a laundry basket.

Why it matters

• Author chrisjpaxton emphasizes a core insight: multi-robot collaboration is fundamentally a data problem, not an over-engineered control problem.
• Traditional multi-robot collaboration often relies on complex centralized control that is hard to scale; CHORUS bypasses this bottleneck with an end-to-end policy.
• This insight suggests that with continued data accumulation, this approach could directly lead to fully automated factories and warehouses, enabling large-scale multi-robot coordination.

2026-07-28 ~ 2026-07-29 · 6 related posts

• CHORUS uses one VLA policy to coordinate a whole multi-robot team without communication — micoolcho · 2026-07-28
• [source] CHORUS teases decentralized robot cooperation with one VLA policy — chris_j_paxton · 2026-07-29
• CHORUS trains one VLA to coordinate multiple robots on manipulation tasks — chris_j_paxton · 2026-07-29
• [source] CHORUS suggests multi-robot collaboration is mainly a data problem — chris_j_paxton · 2026-07-29
• [source] CHORUS trains a VLA for multi-robot collaboration — chris_j_paxton · 2026-07-29

1 near-duplicate retellings: DJiafei