source/isaaclab/test/terrains/test_terrain_generator.py

# Copyright (c) 2022-2026, The Isaac Lab Project Developers (https://github.com/isaac-sim/IsaacLab/blob/main/CONTRIBUTORS.md).
# All rights reserved.
#
# SPDX-License-Identifier: BSD-3-Clause

"""Launch Isaac Sim Simulator first."""

from isaaclab.app import AppLauncher

# launch omniverse app
simulation_app = AppLauncher(headless=True).app

"""Rest everything follows."""

import os
import shutil

import numpy as np
import pytest
import torch

from isaaclab.terrains import FlatPatchSamplingCfg, TerrainGenerator, TerrainGeneratorCfg
from isaaclab.terrains.config.rough import ROUGH_TERRAINS_CFG
from isaaclab.utils.seed import configure_seed


@pytest.fixture
def output_dir():
    """Create directory to dump results."""
    test_dir = os.path.dirname(os.path.abspath(__file__))
    output_dir = os.path.join(test_dir, "output", "generator")
    yield output_dir
    # Cleanup
    if os.path.exists(output_dir):
        shutil.rmtree(output_dir)


def test_generation(output_dir):
    """Generates assorted terrains and tests that the resulting mesh has the expected size."""
    # create terrain generator
    cfg = ROUGH_TERRAINS_CFG
    terrain_generator = TerrainGenerator(cfg=cfg)

    # print terrain generator info
    print(terrain_generator)

    # get size from mesh bounds
    bounds = terrain_generator.terrain_mesh.bounds
    actualSize = abs(bounds[1] - bounds[0])
    # compute the expected size
    expectedSizeX = cfg.size[0] * cfg.num_rows + 2 * cfg.border_width
    expectedSizeY = cfg.size[1] * cfg.num_cols + 2 * cfg.border_width

    # check if the size is as expected
    assert actualSize[0] == pytest.approx(expectedSizeX)
    assert actualSize[1] == pytest.approx(expectedSizeY)


@pytest.mark.parametrize("use_global_seed", [True, False])
@pytest.mark.parametrize("seed", [20, 40, 80])
def test_generation_reproducibility(use_global_seed, seed):
    """Generates assorted terrains and tests that the resulting mesh is reproducible.

    We check both scenarios where the seed is set globally only and when it is set both globally and locally.
    Setting only locally is not tested as it is not supported.
    """
    # set initial seed
    configure_seed(seed)

    # create terrain generator
    cfg = ROUGH_TERRAINS_CFG
    cfg.use_cache = False
    cfg.seed = seed if use_global_seed else None
    terrain_generator = TerrainGenerator(cfg=cfg)

    # keep a copy of the generated terrain mesh
    terrain_mesh_1 = terrain_generator.terrain_mesh.copy()

    # set seed again
    configure_seed(seed)

    # create terrain generator
    terrain_generator = TerrainGenerator(cfg=cfg)

    # keep a copy of the generated terrain mesh
    terrain_mesh_2 = terrain_generator.terrain_mesh.copy()

    # check if the meshes are equal
    np.testing.assert_allclose(
        terrain_mesh_1.vertices, terrain_mesh_2.vertices, atol=1e-5, err_msg="Vertices are not equal"
    )
    np.testing.assert_allclose(terrain_mesh_1.faces, terrain_mesh_2.faces, atol=1e-5, err_msg="Faces are not equal")


@pytest.mark.parametrize("curriculum", [True, False])
def test_generation_cache(output_dir, curriculum):
    """Generate the terrain and check that caching works.

    When caching is enabled, the terrain should be generated only once and the same terrain should be returned
    when the terrain generator is created again.
    """
    # create terrain generator with cache enabled
    cfg: TerrainGeneratorCfg = ROUGH_TERRAINS_CFG
    cfg.use_cache = True
    cfg.seed = 0
    cfg.cache_dir = output_dir
    cfg.curriculum = curriculum
    terrain_generator = TerrainGenerator(cfg=cfg)
    # keep a copy of the generated terrain mesh
    terrain_mesh_1 = terrain_generator.terrain_mesh.copy()

    # check cache exists and is equal to the number of terrains
    # with curriculum, all sub-terrains are uniquely generated
    hash_ids_1 = set(os.listdir(cfg.cache_dir))
    assert os.listdir(cfg.cache_dir)

    # set a random seed to disturb the process
    # this is to ensure that the seed inside the terrain generator makes deterministic results
    configure_seed(12456)

    # create terrain generator with cache enabled
    terrain_generator = TerrainGenerator(cfg=cfg)
    # keep a copy of the generated terrain mesh
    terrain_mesh_2 = terrain_generator.terrain_mesh.copy()

    # check no new terrain is generated
    hash_ids_2 = set(os.listdir(cfg.cache_dir))
    assert len(hash_ids_1) == len(hash_ids_2)
    assert hash_ids_1 == hash_ids_2

    # check if the mesh is the same
    # check they don't point to the same object
    assert terrain_mesh_1 is not terrain_mesh_2

    # check if the meshes are equal
    np.testing.assert_allclose(
        terrain_mesh_1.vertices, terrain_mesh_2.vertices, atol=1e-5, err_msg="Vertices are not equal"
    )
    np.testing.assert_allclose(terrain_mesh_1.faces, terrain_mesh_2.faces, atol=1e-5, err_msg="Faces are not equal")


def test_terrain_flat_patches():
    """Test the flat patches generation."""
    # create terrain generator
    cfg = ROUGH_TERRAINS_CFG
    # add flat patch configuration
    for _, sub_terrain_cfg in cfg.sub_terrains.items():
        sub_terrain_cfg.flat_patch_sampling = {
            "root_spawn": FlatPatchSamplingCfg(num_patches=8, patch_radius=0.5, max_height_diff=0.05),
            "target_spawn": FlatPatchSamplingCfg(num_patches=5, patch_radius=0.35, max_height_diff=0.05),
        }
    # generate terrain
    terrain_generator = TerrainGenerator(cfg=cfg)

    # check if flat patches are generated
    assert terrain_generator.flat_patches
    # check the size of the flat patches
    assert terrain_generator.flat_patches["root_spawn"].shape == (cfg.num_rows, cfg.num_cols, 8, 3)
    assert terrain_generator.flat_patches["target_spawn"].shape == (cfg.num_rows, cfg.num_cols, 5, 3)
    # check that no flat patches are zero
    for _, flat_patches in terrain_generator.flat_patches.items():
        assert not torch.allclose(flat_patches, torch.zeros_like(flat_patches))