SimReady assets: furniture
Chairs, tables, sofas, beds, cabinets, drawers — sized correctly, articulated where it matters, with physics calibrated for assembly, rearrangement, and indoor-navigation training.
Home and indoor robotics are bounded by the furniture in the room: a vacuum has to navigate around it, a humanoid has to interact with it, an assembly robot has to assemble it. Generic 3D furniture libraries optimise for visual rendering, not physics, so the dimensions are right but the mass, joint behaviour, and collision geometry are missing. Rigyd produces SimReady OpenUSD and MJCF with calibrated rigid-body physics, articulated drawers and doors, and cushion approximations that drop into Isaac Lab, MuJoCo, Genesis, and Gazebo without re-authoring.
What's in this category
Chairs
Dining, office, lounge, swivel. Stable bases or articulated wheelbases; armrests as grip targets.
Tables
Dining, coffee, side, desk. Surface friction varies; legs are common navigation obstacles.
Sofas and armchairs
Cushion compression as rigid-body approximation; armrests, footrests, recliner articulation.
Beds and mattresses
Frame articulation (adjustable beds), mattress as soft-body approximation.
Cabinets and dressers
Articulated drawers, hinged doors, internal shelving. Standard manipulation benchmark targets.
Shelving and bookcases
Tall, tipping-risk under load. Shelves as separate placement surfaces.
Lamps and lighting fixtures
Articulated swing arms, adjustable shades, weighted bases.
Storage units and ottomans
Hinged lids, internal compartments. Mass varies with contents.
Wardrobes and closets
Sliding doors, hanging rails, drawer towers. Tall articulation hierarchies.
Physics characteristics
Large footprint, mass distribution matters
Furniture is the largest class of indoor obstacles. A 30kg dresser pushed by a navigation robot tips at different angles depending on mass distribution; a 70kg sofa has a different inertia tensor depending on whether the load is in the cushions or the frame. Rigyd derives inertials consistently from the inferred mass distribution.
Articulated drawers, doors, and recliners
Cabinets, wardrobes, recliners, and adjustable beds all carry articulation. Rigyd outputs OpenUSD with PhysicsArticulationRootAPI per articulated unit and MJCF with native joint declarations. Drawer/door friction is keyed to the hardware material (metal slides vs wooden runners).
Mixed rigid + soft-body approximation
Sofas, mattresses, and cushioned chairs are nominally soft-body but most navigation and manipulation tasks treat them as rigid with adjusted compliance. Rigyd outputs rigid-body geometry that downstream simulators can wrap with PhysX FleX, MuJoCo flex, or Genesis soft-body for the narrow cases where compliance matters.
Common materials
Robot tasks these assets enable
Furniture assembly
Flat-pack assembly is a high-interest robotics benchmark. Tasks need correctly massed parts, hinge friction, and screw-driving contact. Each subassembly emerges as the policy works.
Indoor navigation around furniture
Vacuums, delivery robots, and humanoid platforms all navigate cluttered indoor spaces. Tipping risk under contact and accurate footprint detection drive whether the policy clears the scene.
Rearrangement and home automation
Move-the-couch tasks, table-clearing scenarios, and put-away-the-laundry routines all need furniture interaction with realistic articulated drawers, doors, and compartments.
Compatible simulators
The same source asset feeds every simulator below from one Rigyd output.
Related robotics verticals
~5 min
per asset, including articulation hierarchy
PhysicsArticulationRoot
API applied for drawers, doors, recliners
rigid + soft
output usable with FleX, MuJoCo flex, Genesis soft-body
FAQ
Does Rigyd handle articulated drawers and cabinet doors?
Yes. The articulation hierarchy is detected from the source geometry (separate drawer/door meshes) and Rigyd applies PhysicsArticulationRootAPI on the carcass with revolute joints for hinged doors and prismatic joints for drawer slides. MJCF output uses native MuJoCo joint declarations with the same kinematics. Joint limits, friction, and damping are configurable per joint; default values match typical IKEA-grade hardware behaviour.
How do I simulate cushion compression on sofas and mattresses?
Rigyd outputs rigid-body geometry with cushion meshes as separate prims so a downstream simulator can wrap them with soft-body simulation (PhysX FleX, MuJoCo flex, Genesis soft-body) for tasks where compliance matters. For navigation and reach planning tasks, the rigid-body approximation with an adjusted contact stiffness is usually sufficient and runs orders of magnitude faster.
Can Rigyd generate flat-pack furniture as separate parts for assembly tasks?
Yes — bring CAD exports of the parts as separate .glb / .fbx files and Rigyd generates each part as an individual SimReady asset with its own mass, friction, and collision geometry. Holes, dowels, and connector geometry are preserved from the source CAD so contact behaviour during screw-driving and dowel-insertion remains accurate. Authoring the assembly graph (which part connects to which) is on the user side; Rigyd provides the parts.
Are the dimensions accurate for indoor-navigation training?
Dimensions come directly from the source geometry — Rigyd does not resize. If you bring CAD or a calibrated 3D scan, the simulated furniture matches its real-world dimensions. The text-to-asset path produces generic dimensions consistent with common product categories (e.g. a "two-seater sofa" is roughly 1.6m × 0.85m × 0.85m); override the dimensions if a specific size matters for the scene.
Which simulators consume the output for furniture-rich scenes?
OpenUSD output drops into NVIDIA Isaac Sim, Isaac Lab, Omniverse, Unreal Engine, and Unity directly, plus Gazebo Sim via USD imports. MJCF output drops into MuJoCo, MJX, and Genesis. For indoor-navigation training with many instanced furniture pieces, USD scene composition and referencing make scene assembly efficient — one source asset can be referenced thousands of times without duplicating geometry in memory.
Train home robots on the furniture they will actually meet
Drop in 3D models, images, or text descriptions and get SimReady furniture in minutes per piece.
Request API Access