Spatial perception is the foundational layer of spatial intelligence: it turns raw sensory input into the structured spatial evidence that everything downstream depends on, answering what is present, where it lies, and how it moves. It is what stands between an agent and a scene, and the reliability of every later inference is bounded by the reliability of what perception delivers.
The chapter follows a single progression, from the elemental to the structured. It begins with recognizing the discrete entities that populate a scene, grounds them in metric geometry, then binds them into relational and scene-level structure, extends that structure across time, and finally broadens it beyond vision by fusing complementary modalities. Each step presupposes the ones before it, so that perception builds from isolated percepts toward a dynamic, multimodal model of space that reasoning and generation can act on.