Embodied AI shifts to Whole-Body Intelligence: From VLA to WBI paradigm evolution
机器之心 · wechat · 2026-08-30
The article notes that the focus of embodied intelligence is shifting from high-level semantic understanding to low-level physical control. Traditional VLA models, often trained on static base data, struggle with the dynamic balance and coordination required for bipedal humanoid robots with dozens of degrees of freedom. The industry is now proposing the goal of "Whole-Body Intelligence (WBI)," aiming to enable the robot's "brain" to natively understand mechanics and deeply integrate navigation, balance, and manipulation.
Regarding technical routes, there are differing approaches:
- Google DeepMind: Uses a three-model stack architecture (Gemini Robotics 2, ER2, On-Device 2) to balance long-horizon reasoning with high-frequency response.
- Academia/Startups: Teams like HKU's Li Hongyang and CurrentRobotics are exploring layered architectures and end-to-end models for better whole-body coordination.
WBI represents a leap from "local pose prediction" to a "whole-body physical coordination brain," marking the next stage of consensus in embodied AI.
More from Embodied
- Berkeley researchers sample humanoid robot policies directly from weight space with diffusion — berkeley_ai · 2026-10-16
- OneWM: unified causal training shows world and action modeling improve each other — IAAI233 · 2026-10-16
- Tesla's hover car isn't a flying car — compressed air can't sustain flight — tinyfool · 2026-10-16
- Scobleizer relays claim that Optimus demo will be even more mind-blowing than the road-flipping Roadster — Scobleizer · 2026-10-16
- Indie dev's AI PASSPORT wearable firmware hits 1,200+ downloads in days — ezshine · 2026-10-16
- 1X Opens California Factory for Mass Production of NEO Humanoid Robots — emmanuelvivier · 2026-10-16