01 · Intelligence
Foundation models
Models that connect multimodal inputs, task understanding and robot action. Different systems occupy different points in the stack.
Landscape / Reference
A map of embodied intelligence.
Embodied AI is not one technology. It is a stack connecting perception, reasoning, learning, data, models, control, simulation and physical machines.
02 · Embodiment
Physical platforms
Robot bodies provide the sensors, actuators, geometry and physical constraints through which learned intelligence is expressed.
03 · Learning
Data & training
Demonstrations, teleoperation, simulation and real-world interaction provide different routes to acquiring physical skills.
04 · Execution
Control & manipulation
Planning, whole-body control and manipulation translate goals into movements under physical and safety constraints.
05 · Infrastructure
Simulation & tooling
Simulation and robotics infrastructure support development, evaluation and deployment workflows.
06 · Ecosystem
Organizations
Companies and research teams work on different layers, from models and data to humanoid hardware and robotics infrastructure.
How to use this map
Start with a concept, move into the relevant research problem, then inspect models, organizations and robots. The relationships are informational rather than a ranking of companies, models or machines.