Locomanipulation

A robot arm bolted to a table only has to solve grasping. A humanoid in a living room has to walk to the object first, then plant its feet, lean, crouch or step sideways to reach it — and all of that changes what its cameras can see.

Locomanipulation treats those as one task. The legs, torso, arms and head are controlled together, because where the robot stands decides what it can reach and see.

Figure’s bed-making clips are the plain example: the robot walks around the bed mid-task to fix a fold it cannot reach from where it is standing.