CONCEPT
Subsumption Architecture
Rodney Brooks’s layered design for robot control—each layer a self-contained perception-action loop, higher layers suppressing lower ones rather than commanding them—which proved that robust intelligence could emerge without any central representation of the world.
Subsumption architecture is the concrete engineering embodiment of the claim that intelligence does not require a model. Where the classical robotics pipeline of perceive–represent–plan–act placed a world-model at the center and made every behavior dependent on it, Brooks decomposed behavior horizontally into layers, each one a complete loop connecting sensing directly to acting. The lowest layer might do nothing but avoid collisions; a layer above might add wandering; a layer above that might add goal-directed exploration. Each layer is a finished competence that runs on its own and can be tested in isolation. Higher layers achieve their ends not by commanding the lower ones but by subsuming them—suppressing or overriding their outputs when necessary—without the lower layers needing to know the higher ones exist. The architecture took its inspiration from biology: nervous systems are not organized as a central executive issuing commands to peripheral modules but as layered tangles of reflexes and circuits, the evolutionarily older systems still running beneath the newer ones.
Keep reading with YOU ON AI
Unlock the full book, 10,000+ field-guide entries, and a 1000+ thinker library. If you have a book code, register now — it takes a minute.