Simulations across
every scenario and edge case
Domain randomization needs two things: accurate physics baselines and massive object diversity. Rigyd provides both — converting any 3D model into a physics-enabled asset calibrated for robust policy training.
The problem
Why existing workflows fall short.
Variety is the real bottleneck
Effective domain randomization needs thousands of diverse objects, not just 100 objects with random textures. Object shape, mass, and material variation all matter.
Random physics is worse than no physics
Randomizing mass uniformly between 0.1-10kg for every object produces nonsensical training data. You need realistic distributions centered on accurate estimates.
Creating varied assets manually doesn't scale
Each new object variation needs geometry, physics properties, and collision meshes. At 4 hours per object, building 1,000 variations takes 4,000 engineer hours.
How Rigyd helps
AI-native infrastructure that automates the hard parts.
Scale object diversity instantly
Upload thousands of unique 3D models and get physics-enabled versions of each. Bulk API processing makes it practical to build truly diverse training sets.
Realistic physics distributions
Each object gets material-specific physics values, creating natural variation across your dataset — not uniform random noise.
DR-calibrated accuracy
Mass estimates within 15-20% and friction within 0.1 coefficient. These ranges are intentionally within typical domain randomization variance, making them ideal starting points.
cost reduction in building diverse object sets
saved per 1,000-object dataset
better real-world performance with diverse, accurate training data
Build diverse training datasets at scale
Upload your object library and get physics-accurate assets for domain randomization.
Starts at $29/month. 30 credits included.
Frequently asked questions
Why isn't uniform random physics enough for training?
Randomizing mass uniformly between 0.1-10kg for every object produces nonsensical training distributions. Effective DR needs realistic per-object baselines — which Rigyd provides via material-based estimation.
How many unique objects can Rigyd produce for DR training?
Thousands. Enterprise API bulk-converts 3D catalogs with unique physics per object — geometry diversity AND realistic mass and friction variation, not just random texture noise.
Are Rigyd's physics values within typical DR variance?
Yes — intentionally. Mass accuracy (15-20%) and friction accuracy (0.1 coefficient) sit inside standard DR ranges, making Rigyd assets ideal baselines that randomize to realistic extremes.
What randomization ranges do production teams use with Rigyd baselines?
Typical defaults: mass ±20%, static and dynamic friction ±0.1, restitution ±0.05, center of mass ±1-2 cm for normal-sized objects. Tighter for precision tasks like peg-in-hole (mass ±5%, friction ±0.05); wider for unstructured outdoor scenarios (mass ±30%). Rigyd's estimates sit at the center of these bands, so DR variance reaches realistic extremes rather than running off into impossible-physics territory.
Does Rigyd integrate with Isaac Lab's domain randomization framework?
Yes. Isaac Lab's EventManager and physics randomization APIs operate on USDPhysics attributes — exactly what Rigyd populates. Mass, friction, and restitution randomization configs read directly from the Rigyd-generated asset's baseline values, so randomization centers itself on calibrated physics rather than requiring per-asset hand-tuning of randomization ranges.
Related reading
In-depth guides on robotics simulation, OpenUSD, and SimReady assets.
21 MAY 2026
Synthetic data generation for computer vision in robotics
Real-world labeled robotics data is expensive and slow to capture. Synthetic data is fast and unlimited — but only useful if the underlying simulation has correct physics, semantic labels, and domain randomization. Here's the complete pipeline.
18 MAY 2026
Digital twin creation pipeline for manufacturing
A factory digital twin needs every object to behave physically, not just render. This is the end-to-end pipeline: CAD intake, BIM merge, physics layer, semantic labeling, simulation runtime — at the asset volumes (10K+ unique SKUs) real factories actually contain.
13 MAY 2026
How to set up mass, friction, and joint properties for robot training
The three pillars of robot physics setup — mass, friction, joints — determine whether your trained policy transfers to real hardware. Here's the calibration target for each, the schemas, and the common mistakes that quietly break training.
Explore more
SimReady for Isaac Sim
OpenUSD assets validated for NVIDIA Isaac Sim
Models for MuJoCo
Physics-accurate assets for robot learning
Assets for Gazebo
Simulation-ready models for ROS 2
Warehouse Simulation
50,000+ SKUs need physics properties
Humanoid Robots
Whole-body interaction training assets
Sim-to-Real Transfer
Close the gap with accurate physics data