newton.utils.rasterize_mesh_to_heightfield#

newton.utils.rasterize_mesh_to_heightfield(mesh, resolution, *, max_cells_per_axis=4096)[source]#

Rasterize a triangle mesh into a heightfield elevation grid.

Rays are cast straight down onto the mesh on a regular grid spanning the mesh’s XY bounding box. The grid follows Newton’s heightfield convention (see Heightfield): heights[row, col] samples the surface at x = x_min + col * dx (columns map to X) and y = y_min + row * dy (rows map to Y). Rays that miss the mesh fall back to the mesh’s minimum Z so that nothing collides in gaps. This is only meaningful for surfaces that are single-valued in Z (e.g. locomotion terrain).

Parameters:

  • mesh (Mesh) – Triangle mesh to rasterize. Its vertices are taken in their current frame; transform the mesh into world space beforehand if needed.

  • resolution (float) – Horizontal grid spacing [m]. Smaller values preserve more detail at the cost of a larger grid.

  • max_cells_per_axis (int) – Upper bound on grid rows/columns. If the mesh extent would exceed this, the effective resolution is coarsened to fit.

Returns:

A tuple (heights, bounds) where heights is a (nrow, ncol) float32 array of world-space elevations [m] and bounds is the mesh’s XY bounding box as (x_min, y_min, x_max, y_max) [m].

Return type:

tuple[ndarray, tuple[float, float, float, float]]