Terminology · Embodied intelligence
Embodied intelligence
An agent perceives and acts through a body interacting with its environment, rather than reasoning in symbols alone.
Definition
Embodied intelligence holds that intelligence comes from interaction. To finish a task in the real world a system has to perceive its spatial relationship to objects, predict what an action will do, and adjust from the result. What separates it from pure symbolic reasoning is not model size but whether there is a body and whether physics constrains it.
In engineering terms it makes three hard demands of the perception layer: spatial perception in three dimensions rather than two; six-degree-of-freedom pose estimation for objects; and robustness to lighting, material and occlusion. Without that layer, planning and control above it have nothing to stand on.
How it differs
| Capability | What it must answer | Why 2D is not enough |
|---|---|---|
| 3D spatial perception | Where the object is and how it is oriented | A 2D image drops depth, and picking needs depth |
| 6D pose estimation | How to reach in, and from which angle | Position alone is not enough; orientation matters |
| Physical consistency | What an action will cause | Decisions must obey physics, not just pixels |
Where DaoAI sits in the perception layer
DaoAI works on the perception layer of embodied intelligence: in-house structured-light 3D cameras for imaging, robot vision for 6D pose and guidance, and a world-model foundation that puts space, material, light, motion and causality into one model. The company is a working-group member of the MIIT standardisation technical committee for humanoid robots and embodied intelligence.
Robot vision —— 6D pose, picking and assembly
In-house 3D industrial cameras —— structured-light imaging and hand-eye calibration
The DaoAI World platform —— the world-model foundation
About DaoAI —— credentials and standards work