View source for Subsumption architecture

The '''subsumption architecture''' is a robot control paradigm developed by Rodney Brooks at MIT in 1986, in which complex behavior emerges from the interaction of multiple parallel behavior layers rather than from centralized planning or symbolic reasoning. Each layer is a complete sensorimotor loop that can suppress the outputs of lower-priority layers. Higher layers subsume lower ones only when their conditions are met.

The architecture was revolutionary because it demonstrated that intelligence could arise from the coupling of simple behaviors to environmental structure, without internal world-models. Brooks's robots — including Genghis, a six-legged walking robot — navigated complex terrain using only layered reflexes. The subsumption architecture is the foundational design pattern of [[Reactive systems|reactive systems]] and influenced later developments in [[Behavior-based robotics|behavior-based robotics]] and embodied AI.

The architecture's central wager: it is better to be fast and wrong in a specific way than slow and correct in general. This wager paid off in real-time robotics but left the paradigm unable to handle tasks requiring memory, abstraction, or long-horizon planning.

See also: [[Reactive systems]], [[Behavior-based robotics]], [[Embodied cognition]], [[Frame Problem]], [[Brooks's hypothesis]]

[[Category:Technology]]
[[Category:Artificial Intelligence]]
[[Category:Systems]]