What Is SimReady?

Building a physically accurate workflow starts with getting your assets ready, then using the SimReady standard to prepare the physics properties, semantic labels, material attributes, and articulation data each runtime needs to power simulation, AI, and sensors.

Step 1: Convert Your Asset

Most source content — CAD models, photogrammetry scans, and manual 3D models — arrives without the physics properties, semantic labels, and material attributes the SimReady standard requires, making these assets unreliable for physical AI workflows.

NVIDIA provides tooling to automate this conversion:

• Convert CAD to USD: CAD to USD converter can help convert source CAD assemblies into OpenUSD assets and validate required simulation properties for the targeted workflow.
• Convert Common Formats to USD: Common 3D and robotics formats — including URDF, MuJoCo, Gaussian splats, FBX, OBJ, and glTF— can be converted into OpenUSD assets with simulation properties validated for the targeted workflow. For formats not covered by a dedicated conversion workflow, USD Exchange SDK exporter is available to bring virtually any source format into USD.
• Automatic Material Assignment: Proprietary CAD surface definitions are mapped to MDL physically based materials, preserving appearance under ray-traced lighting while ensuring materials respond correctly to cameras, lidar, radar, and thermal sensors for accurate synthetic data generation.
• Validation: NVIDIA validation tools check completeness, physical plausibility, and runtime compatibility.

Step 2: Load Into a SimReady Scene

Once assets meet the SimReady standard, they can be loaded into a runtime like NVIDIA Isaac Sim™ or NVIDIA Omniverse™. The runtime reads the physics properties, semantic labels, material attributes, and articulation data baked into the asset. requiring no manual setup.

Step 3: Run Physical AI Simulation

With the SimReady asset in the OpenUSD scene, physics, AI, and sensor simulation run simultaneously against the same asset definition.

Step 4: Results

Once in the physical AI simulation, the asset shows realistic, deployment-grade behavior: A robotic arm grasping an object, a conveyor system routing packages, or cooling infrastructure operating under thermal load inside an AI factory digital twin — all grounded in the same validated asset data.

SimReady assets include features that enable robust interaction in 3D environments, such as:

• Semantic Labeling: Provide ground truth for machine learning by verifying perception system identification and classification of objects.
• Non-Visual Sensor and Non-Visual Material Attributes: Enable accurate non-visual sensor simulation (such as radar, lidar, thermal imaging) with attributes including material properties that affect sensor response but aren't visible to the human eye.
• Collision Shapes: Define physical boundaries that determine how objects interact and collide within the simulation environment.
• Mass Properties: Include weight, center of mass, and inertia data that enable realistic physics calculations during simulation.
• Material Details: Specify surface characteristics like friction, elasticity, and visual properties that affect object behavior and appearance.
• Articulation and Behavior: Joint definitions, limits, and kinematic chains so assets with moving parts (robots, doors, conveyors) behave correctly in simulation.
• Variants: Multiple configurations (open/closed, loaded/empty, color options) stored as USD variant sets for design exploration and synthetic data randomization.

SimReady asset creation starts with capturing an object’s geometry, appearance, materials, and real-world physical behaviors, then structuring that representation so simulation tools can read the required physics properties, semantic labels, and material attributes.

Teams typically create SimReady assets in three ways:

• Starting from non-USD assets: Convert OBJ, FBX, CAD, or real-world scan data to USD, then author the missing physics colliders, materials, and semantic labels needed to meet the SimReady specification.
• Starting from existing USD assets: Validate assets against the SimReady specification to identify gaps, then fill in any missing physics, material, or semantic attributes before using them in 3D simulation.
• Starting from validated assets: Use pre-built SimReady assets from NVIDIA and ecosystem partners that are already validated for use in NVIDIA Isaac Sim or NVIDIA Omniverse.

Organizations building digital twins face a structural problem: when assets are created in isolation using custom formats, they risk being incompatible across tools, runtimes, and organizations. This can cause a bottleneck of repeated work, slowing down physical AI pipelines. Simulation content needs shared conventions across toolchains, runtime environments, and domain experts so assets can be reused without one-off translation work.

SimReady advances through open specification work and active industry participation. NVIDIA works with partners to validate SimReady requirements against real-world domains, while contributing learnings to AOUSD efforts that strengthen OpenUSD for industrial and physical AI workflows.

That advancement happens in three ways:

• Feature Expansion: Extend SimReady assets with richer capabilities, including advanced physics behaviors and articulated parts, to support more complex and realistic simulations.
• Pipeline Scaling: Automate asset conversion and validation at scale, helping teams prepare large source-asset libraries for digital environments and training datasets.
• Standardization Review: Define how SimReady features map into OpenUSD so supported tools and platforms can adopt them more consistently across the ecosystem.