Engines and drape tests

The engine-neutral stiffness, the MuJoCo and OmniPhysics exports, and the drape tests in Blender and MuJoCo with their results.

Blender drape test

maillage drop-test imports the scene into Blender and, for each tested piece:

1. sets up Blender's cloth solver from the maillage:* attributes: stiffness rescaled to the mesh and the piece's real mass, stiffer bending on the hem, collisions and self-collisions;
2. drape: lays the piece flat on a round pedestal (diameter 60 % of the shorter side), its centre clamped as between the two discs of a Cusick drape tester, and lets the rest fall; hang: a flag hangs from its two hoist corners on a pole, optionally in the wind;
3. simulates 5 s (120 frames at 24 fps);
4. checks the result and writes drop_test.json:

It also measures the drape coefficient of each draped piece, as in the Cusick test: (projected area − pedestal area) / (flat area − pedestal area), from 1 (stays flat) to 0 (hangs straight down). The draped meshes are exported to draped.usdc with their maillage:* attributes copied back: ready-to-use draped assets.

Results

Indicative results of maillage drop-test on three pieces of each preset (seed 1, ~1,500 vertices per piece, 120 frames, Blender 5.2):

No piece went through the pedestal or the floor. The drape coefficient depends on the shape of the piece as much as on the fabric (a long towel hangs further past a pedestal sized on its short side), so it compares pieces of one preset, not presets with each other. To compare fabrics, the same 40 cm square was draped in each (maillage generate --preset napkin --no-size --no-cut --fabric silk,polyester_knit,cotton_plain,linen,cotton_terry,denim):

In Blender the fabrics differ in mass, thickness, friction and look much more than in drape: the drape coefficients stay within 0.35–0.42, where real silk and denim are far apart (see the roadmap). In MuJoCo, with the engine-neutral stiffness, they do not drape alike: see the fabric drape test.

Silk, plain cotton, linen and denim squares falling onto the clamped pedestals, slowed down twice: maillage generate --preset napkin --count 4 --fabric silk,cotton_plain,linen,denim then maillage drop-test output/napkin --frames 40 --gif --gif-step 1.

Flags hung by their hoist corners in the wind: maillage drop-test output/flag --wind 4 --gif.

Blender values

Blender's cloth stiffness acts per edge and is not divided by the vertex mass, so the same numbers make a fabric stiffer or softer depending on the mesh resolution. Maillage stores the stiffness for a reference mesh with 1 cm edges and the real mass, and the drape test rescales it: tension and shear with the edge length, bending unchanged, air damping with the area per vertex. The solver itself runs with Blender's usual 0.3 kg vertex mass and every stiffness and damping multiplied by the same factor, which leaves the motion unchanged and keeps Blender's solver in the range its tolerances are tuned for.

Tension gives about 3 % stretch when a piece drapes; bending is smaller for softer fabrics, and kept low for all of them (see the roadmap). The values are not calibrated against measured fabric properties.

Engines

Every file is OpenUSD with standard UsdPhysics schemas. Where each engine stands, as on the Primvaria engine matrix:

The main file is engine-neutral; engine-specific exports are read back from it, so every engine sees the same piece.

Engine-neutral stiffness. Each fabric has a tensile (membrane) stiffness K in N/m, a bending rigidity B in N·m and a Poisson's ratio: orders of magnitude from published ranges (B from 1e-6 N·m for silk to 1e-4 N·m for denim; K from 1e2 N/m for a knitted scarf to a few 1e3 N/m for wovens), not measurements. Shell models (MuJoCo, OmniPhysics) take a Young's modulus E and a thickness t; the pair with E·t = K and E·t³ / 12(1 − ν²) = B gives both exactly. That t is a bending thickness, not the fabric's real thickness: fabrics bend far more easily than a solid plate of the same thickness.

MuJoCo drape test

pip install "maillage[sim]"
maillage sim-test output/tea_towel --pieces 4

The same scenario and checks as the Blender drape test, built from the OpenUSD file only: drape on a clamped pedestal (or hang from the pinned vertices), 5 s, then finite values, stretch, penetration, at rest; plus mass (the flex's mass is the file's) and solver (no MuJoCo warning). The drape coefficient is measured.

MuJoCo settings, found with the test (MuJoCo 3.17; CI runs it on MuJoCo 4): the discrete integrator (the only one that integrates flex elasticity implicitly), a 0.5 ms time step (smaller when the membrane is stiff for the vertex mass, as for a small linen napkin), the CG solver with a tight tolerance. MuJoCo cannot put air drag on flex vertices in the XML: the test applies it as a force, 1 N·s/m³ per unit of cloth area, the order of real air drag at draping speeds. Not modelled in MuJoCo: self-collision, the stiffer hem (one stiffness per flex), wind.

Fabric drape test

maillage fabric-test                                  # every fabric, about 4 minutes on one CPU
maillage fabric-test --fabric silk,cotton_plain,denim

Do the fabrics drape differently, as real ones do? This drapes the same sample of each fabric in MuJoCo, a 30 cm disc on an 18 cm pedestal as in a Cusick drape tester, each in its own simulation, and measures its drape coefficient (1: the disc stays flat; 0: it hangs straight down). MuJoCo 4.0.1, 300 vertices, 4 s:

The bending length is Pierce's, (bending rigidity / weight per unit area)^(1/3): the length over which a fabric bends under its own weight. The drape coefficient rises with it, and with the tensile stiffness: to drape in folds a disc must also shear in its plane, which knits (low tensile stiffness) do more easily than wovens. That is the dependence Cusick found on real fabrics, on bending length and shear stiffness (Cusick, 1965).

The test passes when every sample passes the checks of the drape test, silk, plain cotton and denim drape in that order at least 0.05 apart, and the fabrics spread over at least 0.2 (here 0.32). CI runs it on every fabric.

What it does not show: the values are not calibrated, so a fabric's coefficient is not the one a Cusick tester would measure on it, and MuJoCo has no self-collision here, so folds can pass through each other. The mesh matters within about ±0.04: silk gives 0.52 to 0.59 and denim 0.83 to 0.89 between 300 and 1200 vertices, which leaves the order and the spread unchanged.

OmniPhysics layer

Open maillage_<preset>_omniphysics.usda in Omniverse or Isaac Sim: it sublayers the main file and adds a physics scene, the deformable-body and surface-simulation schemas on each cloth mesh (the mesh is both the body root and its simulation mesh), rest shape and triangles, and one physics material per fabric (Young's modulus, surface thickness, Poisson's ratio, density, friction). Not exported yet: the pinned vertices of a flag (they need an attachment) and the stiffer hem.