CHORUS uses one VLA policy to coordinate a whole multi-robot team without communication
micoolcho · x · 2026-07-28
CHORUS proposes a single pretrained VLA backbone that can control a whole team of robots in a decentralized way.
- Each robot runs its own copy of the policy at inference time.
- It conditions only on local observations plus a robot-identifying prompt.
- No shared cameras, proprioception, or inter-robot communication is required at inference.
In real-world tasks such as mobile tape measurement, library book handovers, and lifting laundry baskets, CHORUS reportedly delivers:
- a 64-point improvement over decentralized from-scratch models,
- a 40-point improvement in responsiveness to teammate behavior,
- and better results than centralized baselines.
The authors argue this shows shared VLA priors can enable decentralized multi-robot collaboration without per-robot policies.
Related event: CHORUS: Single VLA Model Enables Decentralized Multi-Robot Collaboration(6 posts)→
More from Embodied
- XPENG's IRON robot demos full-duplex speech with 9-mic array and lip reading — ChrisGPT · 2026-09-23
- Uber riders can now hail Waymo robotaxis on Austin freeways — ATTlKA · 2026-09-23
- M5 Ultra LLM test: 4x faster prompt processing, but double the power draw — DigitalguyCH · 2026-09-23
- Figure CEO: Humanoid Robotics Needs Four Stages and Eventually Hundreds of Billions — adcock_brett · 2026-09-23
- InstinctFlash: open-source runtime runs 8 robot model families in real time on one commercial GPU — chris_j_paxton · 2026-09-23
- M5 Ultra 96GB vs RTX 5090-6000 Pro: which rig for local video generation at $6.5K-$27K — MaxwellHusk · 2026-09-23