Stanford's HomeBody humanoid explores a kitchen, remembers where things are, and cleans up on its own
HomeBody: A humanoid that explores, remembers, and acts on its own
HomeBody is a humanoid system for long-horizon loco-manipulation that first explores an unseen room, builds a spatial memory and a Real2Sim digital twin in Isaac Sim, then follows everyday instructions like "tidy up the kitchen." A high-level VLM composes navigation, picking, placing and drawer-opening skills, while local retries and visual feedback correct execution errors. It runs on a single RTX 4090 laptop with the frontier model remote.
Long-horizon loco-manipulation takes a humanoid beyond what its ego view can show. To act across a room, it needs an internal spatial model that connects visible objects with remembered locations and helps resolve ambiguous requests.