simulation/modules/IsaacLab/source/isaaclab/test/utils/test_noise.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

import pytest
import torch

import isaaclab.utils.noise as noise


@pytest.mark.parametrize("device", ["cpu", "cuda:0"])
@pytest.mark.parametrize("noise_device", ["cpu", "cuda:0"])
@pytest.mark.parametrize("op", ["add", "scale", "abs"])
def test_gaussian_noise(device, noise_device, op):
    """Test guassian_noise function."""

    # Use ones for 'scale' so noisy_data is the noise term directly;
    # random data can be exactly 0 (torch.rand's [0, 1) contract), and 0/0 = NaN (OMPE-94619).
    if op == "scale":
        data = torch.ones(10000, 3, device=device)
    else:
        data = torch.rand(10000, 3, device=device)
    # define standard deviation and mean
    std = torch.tensor([0.1, 0.2, 0.3], device=noise_device)
    mean = torch.tensor([0.4, 0.5, 0.6], device=noise_device)
    # create noise config
    noise_cfg = noise.GaussianNoiseCfg(std=std, mean=mean, operation=op)

    for i in range(10):
        # apply noise
        noisy_data = noise_cfg.func(data, cfg=noise_cfg)
        # calculate resulting noise compared to original data set
        if op == "add":
            std_result, mean_result = torch.std_mean(noisy_data - data, dim=0)
        elif op == "scale" or op == "abs":
            std_result, mean_result = torch.std_mean(noisy_data, dim=0)

        assert str(noise_cfg.mean.device) == device
        assert str(noise_cfg.std.device) == device
        torch.testing.assert_close(noise_cfg.std, std_result, atol=1e-2, rtol=1e-2)
        torch.testing.assert_close(noise_cfg.mean, mean_result, atol=1e-2, rtol=1e-2)


@pytest.mark.parametrize("device", ["cpu", "cuda:0"])
@pytest.mark.parametrize("noise_device", ["cpu", "cuda:0"])
@pytest.mark.parametrize("op", ["add", "scale", "abs"])
def test_uniform_noise(device, noise_device, op):
    """Test uniform_noise function."""
    # Use ones for 'scale' so noisy_data is the noise term directly;
    # random data can be exactly 0 (torch.rand's [0, 1) contract), and 0/0 = NaN (OMPE-94619).
    if op == "scale":
        data = torch.ones(10000, 3, device=device)
    else:
        data = torch.rand(10000, 3, device=device)
    # define uniform minimum and maximum
    n_min = torch.tensor([0.1, 0.2, 0.3], device=noise_device)
    n_max = torch.tensor([0.4, 0.5, 0.6], device=noise_device)
    # create noise config
    noise_cfg = noise.UniformNoiseCfg(n_max=n_max, n_min=n_min, operation=op)

    for i in range(10):
        # apply noise
        noisy_data = noise_cfg.func(data, cfg=noise_cfg)
        # calculate resulting noise compared to original data set
        if op == "add":
            min_result, _ = torch.min(noisy_data - data, dim=0)
            max_result, _ = torch.max(noisy_data - data, dim=0)
        elif op == "scale" or op == "abs":
            min_result, _ = torch.min(noisy_data, dim=0)
            max_result, _ = torch.max(noisy_data, dim=0)

        assert str(noise_cfg.n_min.device) == device
        assert str(noise_cfg.n_max.device) == device
        # add a small epsilon to accommodate for floating point error
        assert all(torch.le(noise_cfg.n_min - 1e-5, min_result).tolist())
        assert all(torch.ge(noise_cfg.n_max + 1e-5, max_result).tolist())


@pytest.mark.parametrize("device", ["cpu", "cuda:0"])
@pytest.mark.parametrize("noise_device", ["cpu", "cuda:0"])
@pytest.mark.parametrize("op", ["add", "scale", "abs"])
def test_constant_noise(device, noise_device, op):
    """Test constant_noise"""
    # Use ones for 'scale' so noisy_data is the bias term directly;
    # random data can be exactly 0 (torch.rand's [0, 1) contract), and 0/0 = NaN (OMPE-94619).
    if op == "scale":
        data = torch.ones(10000, 3, device=device)
    else:
        data = torch.rand(10000, 3, device=device)
    # define a bias
    bias = torch.tensor([0.1, 0.2, 0.3], device=noise_device)
    # create noise config
    noise_cfg = noise.ConstantNoiseCfg(bias=bias, operation=op)

    for i in range(10):
        # apply noise
        noisy_data = noise_cfg.func(data, cfg=noise_cfg)
        # calculate resulting noise compared to original data set
        if op == "add":
            bias_result = noisy_data - data
        elif op == "scale" or op == "abs":
            bias_result = noisy_data

        assert str(noise_cfg.bias.device) == device
        torch.testing.assert_close(noise_cfg.bias.repeat(data.shape[0], 1), bias_result)