printphys tool reduces inertia error for 3D-printed parts in simulation
Open-source tool estimates mass within 7% of real weight using print settings.
Lakshya Saraff noticed a persistent issue when simulating 3D-printed robot parts: CAD tools and URDF exporters compute the inertia tensor as if the part is a solid block of plastic, ignoring infill patterns, wall thickness, and layer height. This leads to inaccurate dynamics and widens the sim-to-real gap in robot simulation. To solve this, he built printphys, an open-source Python tool that takes an STL mesh and actual print settings—material, infill percentage, pattern, wall count, and layer height—and outputs mass, center of mass, and the full inertia tensor. The output is formatted as a ready-to-paste URDF <inertial> block, with support for SDF and MJCF formats.
Saraff validated the tool by printing a small part and weighing it. printphys estimated 1.93 g, while a physical scale measured 2.07 g—a ∼7% error that tightens to under 5% with finer voxel resolution. He attributes the remaining discrepancy to the part's small size, where walls dominate weight more than infill. The tool also includes a --weighed-mass mode: if you weigh the real part, it rescales the entire inertia tensor and COM to match. Saraff is now seeking community feedback on how much inertia error actually matters versus friction or actuator modeling, and what additional features would make the tool more useful for sim-to-real workflows.
- Printphys uses mesh geometry plus actual print settings (infill pattern, wall count, layer height) instead of assuming solid plastic.
- Validation on a small part: estimate 1.93g vs measured 2.07g (~7% error, improves with finer voxels).
- Includes --weighed-mass mode that rescales inertia from physical measurement for higher accuracy.
Why It Matters
Better inertia for 3D-printed parts helps close the sim-to-real gap in robotics and physical simulation.