Embodied systems are the arena in which spatial intelligence stops being a matter of analysis and becomes a matter of consequence. An embodied agent does not merely describe a scene; it occupies one, and every perception it forms and every inference it draws is ultimately answerable to an action it must take in the physical world, under real time budgets, sensor noise, and the cost of being wrong. This chapter turns from the capabilities of the previous part to their deployment, examining how perception, reasoning, and generation are brought to bear inside agents that move, manipulate, and assist.
What makes this more than a restatement is that embodiment changes the problem. The same spatial competence looks different in a manipulator closing a grasp, a car threading traffic, a wearable inferring its user’s intent, a drone holding position in wind, or a fleet of robots coordinating on a warehouse floor: each fixes its own sensors, latencies, safety margins, and notion of success. The chapter is organized by these domains rather than by capability, because the interesting content lives in the domain-specific delta, what each setting demands of spatial intelligence that a general treatment does not reveal.