Bridge the Sim-to-Real Gap.
Physics-Calibrated OpenUSD Digital Twins.
Robotic policy collapse happens when simulated physics diverges from physical reality. Blue Fieldo captures real-world industrial environments with terrestrial laser scanning, photogrammetry, and contact profiling — delivering sub-millimeter OpenUSD stage hierarchies, PBR textures, and real-world friction coefficients directly into NVIDIA Isaac Sim, MuJoCo, and Isaac Lab.
The 4-Stage Reality-to-Simulation Pipeline
How Blue Fieldo turns physical factories, warehouses, and unstructured outdoor proving grounds into policy-ready reinforcement learning stages.
Sub-Millimeter Reality Capture
Deploying terrestrial Leica BLK360/RTC360 LiDAR and DJI Matrice 350 RTK photogrammetry drones to capture millimeter-accurate georeferenced point clouds (`.e57`, `.las`).
- Zero-drift RTK ground control targets
- Indoor SLAM mobile LiDAR passes
- Multi-angle occlusion-free registration
Photorealistic PBR Materials
Reconstructing dense 3D Gaussian Splats (`.ply`) and generating PBR texture maps (Albedo, Roughness, Normal, Metallic) calibrated against Macbeth color chart probes.
- 8K diffuse & specular reflection maps
- HDRI 360° ambient lighting environments
- Realistic shadow & indirect illumination
OpenUSD Physics & Kinematics
Converting raw geometry into hierarchical OpenUSD stages with V-HACD convex decomposition collision meshes, physical mass inertia tensors, and Coulomb friction.
- USD Physics (UsdPhysicsSchema) compliance
- URDF / MJCF kinematic chain export
- Static/Dynamic surface friction tables
Domain Randomization & RL
Pre-configured Isaac Lab and Omniverse environments ready for massive parallel reinforcement learning with randomized lighting, textures, mass, and damping factors.
- Direct integration with Isaac Lab RL loops
- Zero-shot policy transfer verification
- Automated real-to-sim validation loss tests
Calibrated Physics Property Specifications
We empirically profile contact dynamics in the physical workcell before generating simulation assets:
| PHYSICAL PARAMETER | REAL-WORLD MEASUREMENT TOOL | MEASURED ACCURACY | SIMULATOR SCHEMA |
|---|---|---|---|
| Geometric Dimensions | Leica BLK360 Terrestrial LiDAR | ±1.2 mm tolerance | UsdGeomMesh (Metric Units) |
| Static & Dynamic Friction (μs, μk) | 6-Axis ATI Gamma Force-Torque Cell | ±0.02 coefficient | PhysicsMaterialAPI (dynamicFriction) |
| Collision Hulls | Volumetric V-HACD Decomposition | Watertight convex decomposition | PhysicsCollisionAPI (convexHull) |
| Rigid Body Mass & Inertia Tensor | Dual-axis Center of Mass Balancer | ±0.5% mass error | PhysicsMassAPI (diagonalInertia) |
| PBR Reflectance & Roughness | Macbeth ColorChecker + Polarizer HDR | ΔE < 1.5 color deviation | UsdPreviewSurface / MDL Shader |
Compatible With Your Preferred Robotics Stack
NVIDIA Isaac Sim & Lab
Native OpenUSD stage hierarchies with PhysX 5 rigid bodies, articulated drives, and RTX real-time raytracing.
DeepMind MuJoCo
Export to MJCF XML schemas with generalized coordinates, tendon routing, and contact solver friction pyramids.
ROS 2 & Gazebo Garden
Standard URDF/SDF kinematics, DAE visual meshes, and Gazebo world definition files with DART physics.
Genesis & Isaac Gym
Massively parallel GPU-accelerated environments for multi-thousand instance reinforcement learning policies.
Have a Physical Workcell That Needs Digital Twin Simulation?
Our reality capture squad mobilizes on 48h notice with terrestrial laser scanners, drone photogrammetry, and contact friction profiling rigs anywhere in North America, Western Europe, or Japan.