TANGO: Whole-Body VLA Maps RGB to 29-DoF Actions for Humanoid Navigation, Trained Fully in Simulation

berkeley_ai · x · 2026-09-24

TANGO, a CoRL 2026 work, is a whole-body VLA for humanoid navigation that directly predicts 29-DoF joint actions from RGB observations, enabling coordinated arm, torso, and gait control to traverse cluttered 3D spaces. Trained entirely in simulation, it transfers zero-shot to the real world. Project page and paper are public.

Related event: Berkeley's TANGO Enables Whole-Body Humanoid Navigation(2 posts)→

Original post →

More from Embodied

Embodied channel →