TANGO: whole-body VLA lets humanoid robots zero-shot navigate cluttered spaces on Unitree G1
chris_j_paxton · x · 2026-09-09
A CoRL 2026 paper introduces TANGO, a whole-body vision-language-action framework for humanoid navigation: from a language instruction and egocentric RGB, it directly predicts 29-DoF joint actions coordinating arms, torso, and gait to traverse cluttered 3D indoor spaces.
- Navigation is framed as a whole-body problem rather than a 2D abstraction.
- Trained entirely in simulation via a Plan-Edit-Track pipeline (global path planning, kinematic whole-body motion generation, obstacle-aware editing, RL tracking).
- Triple-system architecture: vision-language backbone, action expert, and tracking, producing executable whole-body action chunks.
- Zero-shot transfer to a real Unitree G1 with no real-world data. Commenter Chris Paxton notes perception-aware whole-body control is crucial yet still rare.
Related event: TANGO: Whole-Body VLA Enables Humanoid Robots to Navigate Cluttered Spaces(2 posts)→
More from Embodied
- Shanghai AI Lab-led team releases 100-hour egocentric wearable dataset for humanoid training — ChuckDBrooks · 2026-10-26
- RealSense D585 Pro AI-Powered 3D Stereo Depth Camera Unboxed for Physical AI — chrismatthieu · 2026-10-26
- Physical world agent swarm demo shared on X — pduan · 2026-10-26
- Looking for a wearable mic with best battery life that feeds recordings to personal agents via MCP — altryne · 2026-10-26
- 'Lamp Swarm': a desk of lamp-shaped robots shows what physical AI work looks like — dee_hw · 2026-10-26
- Tencent's HiSTEVE 1.0: hierarchical VLA agent plays Minecraft with native keyboard and mouse, hits 85% SOTA — _akhaliq · 2026-10-26