Is this page helpful?

Known Issues#

General#

  1. On some Windows systems, you may encounter an error like the following:

    OSError: [WinError 126] The specified module could not be found. Error loading "C:\path\to\omni.isaac.ml_archive\pip_prebundle\torch\lib\fbgemm.dll" or one of its dependencies.

    This issue is caused by missing build tools. To resolve it, install Visual Studio 2022 and then install MSVC v143 - VS 2022 c++ x64/86 build tools through the Visual Studio interface.

  2. The replicator Scatter3D OmniGraph node breaks physics when called on a stage using world.

  3. If exiting standalone Python scripts with Ctrl-C, it may need to be done twice to exit.

  4. If more than one asset in URDF contains the same material name, only one material is created. Regardless if the parameters in the material are different. For example, if two meshes have materials with the name “material”, one is blue and the other is red, both meshes will be either red or blue. This also applies for textured materials.

  5. When using World or SimulationContext from isaacsim.core.api with OmniGraph, make sure the graphs are created before the World or SimulationContext are initialized.

  6. Using replicator’s rep.new_layer() functionality, which creates a new layer in which to place and randomize assets, may lead to issues in simulation scenarios where these assets are used. In such cases the use of rep.new_layer() can be omitted.

  7. Dragging and dropping an asset with default values in OmniGraph nodes should be saved with scene and reloaded before hitting Play to make sure all values are correctly set.

  8. Cortex samples with ROS synchronization may perform abnormally and not be able to execute the task if the running FPS drops below 25 FPS.

  9. If randomized materials are not loaded on time for synthetic data generation the rt_subframes must be set to be at least 2.

  10. Multiple Robot ROS 2 Navigation tutorial has high CPU usage. If you observe instances of robots colliding or experiencing localization issues, it’s likely because the Nav2 stack is unable to properly synchronize with sensor data, resulting in missed controller commands.

  11. There can be many warnings and other messages when running Isaac Sim. The amount of log output can be reduced by using the following command line arguments:

    --/log/level=error --/log/fileLogLevel=error --/log/outputStreamLevel=error

  12. Using Replicator to write to S3 buckets with the built-in backend in Windows may require setting the credentials in the environment variables instead of the AWS config files. This is because of a possible path parsing error in boto3 on Windows.

  13. When running standalone examples in Windows, in some scenarios, threads may not be properly cleaned up when the application is closed. This can usually be ignored because the application will still successfully close. As a workaround, you can add multiple standalone_app.update() calls before calling standalone_app.close().

    Windows fatal exception: access violation

Thread 0x00000634 (most recent call first):

  14. The ROS 2 QoS Profile OmniGraph node is unable to save custom profiles unless you manually change the createProfile input to “Custom” first before updating the other fields.

  15. In builds where ROS 2 joint-name override resolution does not honor IsaacJointAPI isaac:NameOverride, custom joint names might not appear on /joint_states or might not resolve through the Isaac Joint Name Resolver node. Use the authored USD joint prim names for control, or as a temporary legacy compatibility workaround only, author isaac:nameOverride on the affected joint prims. New joint assets should continue to use IsaacJointAPI isaac:NameOverride. Reserve lowercase isaac:nameOverride for non-joint prims and legacy compatibility. See NameOverride Attribute.

  16. USD to URDF Exporter:

    • The Collider meshes may be improperly included in the visuals. They can be manually removed from the URDF file.

    • The Body and Joints are authored in the URDF file in alphabetical order. They can be manually reordered in the URDF file.

    • Depending on the robot structure, some body names may be overriden due to the merging of different frames. Review the output and verify that it’s accurate.

    • The URDF exporter adds joint effort and velocity limits as inf when unbounded. This may make the URDF not import correctly if the URDF parser does not support inf values in Float.

  17. The Isaac Sim asset path does not work directly with the Omniverse Kit file picker dialog. As a workaround, when using an S3 asset path with the Omniverse Kit file picker, copy and paste the path and hit enter instead of clicking Select.

    ../_images/isim_4.5_full_tut_gui_filepicker_known_issue_with_s3_assets.png

  18. Replicator synthetic data generation may require more subframes to be rendered for scenes with significant changes (e.g. moving objects or changing lighting conditions). See RT Subframes Parameter and subframes examples for more information.

  19. Franka Open Drawer example is not able to open the drawer by default on Blackwell GPU, increase the self._physics_rate to 600 will work correctly. The issue is currently under investigation.

  20. When running in standalone mode, the Replicator CosmosWriter might skip generating videos from the recorded frames. In this case, run a few app updates before detaching the writer to make sure the videos will be generated.

  21. When using Replicator for synthetic data generation (SDG) workflows, it is recommended to set the DLSS model to Quality mode to avoid rendering artifacts. At lower resolutions (especially below 600x600), the default Performance mode may cause issues such as transparent or incorrectly rendered edges in the generated images.

    import carb.settings

# Set DLSS to Quality mode (2) for best SDG results (Options: 0 (Performance), 1 (Balanced), 2 (Quality), 3 (Auto)
carb.settings.get_settings().set("/rtx/post/dlss/execMode", 2)

  22. Starting in Isaac Sim 5.0, the full Isaac Sim experience (isaacsim.exp.full.kit) enables Fixed Time Stepping by default so that SimulationApp workflows step the timeline deterministically. As a side effect, USD scenes that contain keyframe animations and that render below the target rate (typically 60 Hz) will appear slow or choppy in the GUI, because the timeline advances by a fixed dt per loop tick rather than by wall-clock time. The same scenes play back smoothly in USD Composer and in Isaac Sim 4.5, which use Variable stepping. To opt the GUI into Variable stepping for animation review/authoring, launch with all three of the following flags (the third prevents the isaacsim.core.throttling extension from re-enabling manual mode on every Play); see Choppy or Slow Animation Playback (Fixed Time Stepping) for the full discussion.

    ./isaac-sim.sh \
    --/app/player/useFixedTimeStepping=false \
    --/app/runLoops/main/manualModeEnabled=false \
    --/exts/isaacsim.core.throttling/enable_manualmode=false

  23. The DL denoiser (DLSS Ray Reconstruction) can leave white-noise speckles on materials in simple scenes, for example purple floors in the object reconstruction assets SDG example. Disable /rtx/dldenoiser/responsiveDenoising to switch to a different, more temporally stable RR model; the denoiser stays on. Disabling it can introduce ghosting; for SDG that randomizes materials, wait additional subframes or reset DLSS history. See White noise on materials with the DL denoiser.

    import carb.settings

carb.settings.get_settings().set("/rtx/dldenoiser/responsiveDenoising", False)

  24. In Isaac Sim 6.0 and 6.1, the RTX Lidar echoId field is always 0, even for multi-echo Lidar profiles that report more than one return per emitted ray. This affects the GenericModelOutput buffer and every consumer that forwards the field, including the RTX sensor annotators and the ROS 2 point cloud publisher. Other per-return fields such as range and intensity are unaffected. There is no workaround; a fix is planned for a later release. See Known issue: RTX Lidar echoId is always zero.

  25. Renaming or reparenting a render product, or one of its ancestor prims, after rendering starts can produce the following error:

    ValueError: Invalid object in Py_Graph in getWrappedGraphFromNode

    To avoid this error, finalize the prim hierarchy before rendering starts. Do not rename or reparent render products or their ancestor prims after they start rendering.

Warnings#

  1. Warnings similar to the following can be ignored:

    • [Warning] [omni.usd] Warning (secondary thread)

    • [Warning] [carb.tasking.plugin] Counter 0x7f25e002f8d0

    • [Warning] [rtx.neuraylib.plugin] [MDLC:COMPILER]   1.0   MDLC   comp warn

    • [Warning] [rtx.mdltranslator.plugin] Unable to resolve

    • [Warning] [omni.tagging.plugin] Failed to discover tagging service

    • [Warning] [omni.isaac.dynamic_control.plugin] DcFindArticulationDof: Function called while not simulating

    • [Warning] [omni.isaac.dynamic_control.plugin] DcSetDofProperties: Function called while not simulating

    • [Warning] [omni.client.plugin]  Tick: authentication: Could not connect to discovery service at "wss://...

  2. If there is unwanted noise in simulated depth images, disable anti-aliasing under the Render Settings > Ray Tracing > Anti-Aliasing tab by setting the Algorithm to None.

  3. Pyperclip is used to copy text in some extensions, if you see the following message refer to the link to install a supported copy/paste mechanism:

    Pyperclip could not find a copy/paste mechanism for your system.
For more information, see https://pyperclip.readthedocs.io/en/latest/index.html#not-implemented-error

  4. If you see warnings similar to the following, they can be ignored:

    [Warning] [omni.client.python] Detected a blocking function. This will cause hitches or hangs in the UI. Please switch to the async version

    Make sure you have access to the Internet or a Nucleus server.

Errors#

  1. These errors can be ignored while running the Omniverse Streaming Remote Client, if the client works as normal:

    ERROR [BifrostClient: Streamer] {2B436200} -  updateVideoSettingsForNVbProfile: profile 8 is not handled
ERROR [NVST:ClientSession] {1E8D2700} -  Number of channels(2) is not valid for surround configuration
ERROR [NVST:ClientSession] {1E8D2700} -  Either in stereo or error in receiving opus information from server

ERROR [LAVCDecoder] {FC4A0700} - GERONIMO_ERROR 0xC0040006 GERONIMO_LAVCDECODER_DECODE: Packet decode failure.
ERROR [GIOInterface] {DAFFD700} - GERONIMO_ERROR 0xC0020003 GERONIMO_IOINTERFACE_INVALID_AUDIO_FUNC: Audio Renderer is NULL.
ERROR [NVST:ClientLibraryWrapper] {1E8D2700} -  Cannot find streamEventRaised for stream.media type 1
ERROR [NVST:ClientLibraryWrapper] {1E8D2700} -  Cannot find streamEventRaised for stream.media type 2

ERROR [NVST:RtspSessionPocoBase] {FB280700} -  perform() failed: 0
ERROR [NVST:RtspPocoEvent] {FB280700} -  RTSP-XNvEvent Polling failed: 0, rc: 408

ERROR [NVST:UdpRtpSource] {F9A7D700} -  UDP RTP Source: failed to receive data (Error: 0x80000013)
ERROR [NVST:RtpSourceQueue] {F9A7D700} -  RtpSourceQueue: failed to read RTP packet (Result: 0X80000013)
ERROR [BifrostClient: NvscWrapper] {1D0CF700} - Received old frame - current 7536 received 0
ERROR [BifrostClient: Interface] {2B436200} - nvbSendInputEvent(). SessionIdentifier is ''

    ERROR [Geronimo::Analytics] {875D4140} - Failed to load dll with error: /../..//NvTelemetryAPI.so: cannot open shared object file: No such file or directory
ERROR [BifrostClient: Streamer] {875D4140} - updateVideoSettingsForNVbProfile: profile 8 is not handled
ERROR [NVST:ClientSession] {7EC38640} - Number of channels(2) is not valid for surround configuration
ERROR [NVST:ServerControl] {4BFFF640} - Unknown server notification
ERROR [GIOInterface] {49FFB640} - GERONIMO_ERROR 0xC0020003 GERONIMO_IOINTERFACE_INVALID_AUDIO_FUNC: Audio Renderer is NULL.
ERROR [NVST:ClientLibraryWrapper] {7EC38640} - Cannot find streamEventRaised for stream.media type 1
ERROR [NVST:ClientLibraryWrapper] {7EC38640} - Cannot find streamEventRaised for stream.media type 2
[h264 @ 0x55580065b160] Cannot parallelize slice decoding with deblocking filter type 1, decoding such frames in sequential order
To parallelize slice decoding you need video encoded with disable_deblocking_filter_idc set to 2 (deblock only edges that do not cross slices).
Setting the flags2 libavcodec option to +fast (-flags2 +fast) will disable deblocking across slices and enable parallel slice decoding but will generate non-standard-compliant output.
ERROR [BifrostClient: Streamer] {875D4140} - Sending Input event failed: Nvsc Error: NVST_R_GENERIC_ERROR (0x800b0000)

  2. Error similar to the following happens when STOPPING and STARTING simulation again when using Isaac core world class. To stop the error trail, Reset the scene with one of the world.reset methods (in the core samples, this happen when pressing the RESET button on the UI).

    2021-06-01 19:16:51 [65,842ms] [Error] [omni.kit.app._impl] [py stderr]: AttributeError: 'NoneType' object has no attribute '<...>'

At:
<...>

  3. If an asset contains an Action Graph and the Action Graph window is closed before re-opening the same asset, the following errors may appear and can be ignored:

    [Error] [omni.usd] TF_PYTHON_EXCEPTION (secondary thread): in TfPyConvertPythonExceptionToTfErrors at line 114 of /buildAgent/work/ca6c508eae419cf8/USD/pxr/base/tf/pyError.cpp -- Tf Python Exception

[Error] [omni.kit.app.impl] [py stderr]: sys:1: RuntimeWarning: coroutine 'OmniGraphModel.__delayed_prim_changed' was never awaited
RuntimeWarning: Enable tracemalloc to get the object allocation traceback

[Error] [omni.graph] Invalid GraphObj object in Py_Graph in getNode

  4. Errors when converting ShapeNet models may appear from the omni.kit.asset_converter extension when textures for the target model are missing from the input dataset. These errors can be ignored.

    [Error] [omni.kit.asset_converter.impl.omni_client_wrapper] Cannot copy from */images/texture2.jpg to */textures/texture2.jpg, error code: Result.ERROR_NOT_FOUND.

  5. Errors when livestreaming. These errors can be ignored.

    [Error] [carb.livestream.plugin] nvstPushStreamData timeout for eye 0, stream (nil).
[Error] [omni.kit.livestream.webrtc.plugin] NVST Error: NVST_R_FRAME_DROPPED
[Error] [omni.kit.livestream.webrtc.plugin] NVST Error: NVST_R_BUSY

    If the stream remains connected and interactive, these messages do not require action. NVST_R_BUSY can appear while disconnecting one WebRTC client and reconnecting another; close the existing client session or reload the browser-based viewer before reconnecting.

  6. Errors when using a Jupyter notebook. These errors can be ignored.

    [Error] [omni.kit.app.plugin] Can`t delay app ready event, it was already sent. Requester name: omni.usd

  7. Errors while generating samples with flying_things_4d.yaml for larger value of --num-scenes (specially when running for stress test):

    [Error] [omni.physicsschema.plugin] Rigid Body of (/Replicator/SampledAssets/Population_9658d6d1/Ref_Xform_05/Ref) missing xformstack reset when child of rigid body (/Replicator/SampledAssets/Population_9658d6d1/Ref_Xform_05) in hierarchy. Simulation of multiple RigidBodyAPIs in a hierarchy will cause unpredicted results. Please fix the hierarchy or use XformStack reset.

  8. Error like this when running Composer can be ignored.

    [Error] [omni.graph.core.plugin] /Replicator/SDGPipeline/OgnSampleCombine_03: [/Replicator/SDGPipeline] Assertion raised in compute - AttributeData 'OgnSampleCombine_03.outputs:samples' of type 'float3[]' required array of elements of length 3, got array with elements of size 1

  9. When using the surface gripper between two objects that contain articulation root, the following error may appear and the surface Gripper won’t work. To avoid it, disable the Articulation API from the picked object.

    [Error] [omni.physx.plugin] PhysX error: PxD6JointCreate: actors must be different

  10. Fatal error regarding omni.sensors plugin when running RTX Radar. Unless you have manually disabled Vulkan and MotionBVH, this error appears if you are using a below-minimum-spec GPU. Your GPU must be Ampere architecture or newer.

    [Warning] [omni.sensors.nv.radar.wpm_dmatapprox.plugin] MotionBVH activation state 0 doesn\'t match requested state 1
[Fatal] [omni.sensors.nv.radar.wpm_dmatapprox.plugin] Running radar without MotionBVH is disallowed, to force it use --/app/sensors/nv/radar/runWithoutMBVH=true
If you are running on Windows and have motionBVH enabled, be sure to enable Vulkan as well by passing --vulkan

  11. Error regarding failure to process writer attach request when playing scene containing an OmniGraph, after changing timeCodesPerSecond setting. To resolve, save the scene, reopen it, then play it again.

    [Error] [omni.graph] Invalid Node object passed to Graph.get_graph_from_node
[Error] [isaacsim.core.nodes.impl.base_writer_node] Could not process writer attach request (<omni.replicator.core.scripts.writers.NodeWriter object at 0x7355b3d175b0>, None), Invalid NodeObj object in Py_Node in getAttributes

  12. In omni.replicator.object, the description file demo_shader_attributes_diffuse.yaml can have a corrupted JPEG error that the picture is not written before it’s used. We are looking into fixing it.

    [Error] [gpu.foundation.plugin] Couldn\'t process /tmp/carb.F3srY8/randomized_output.jpg, it might not have written completely. Reason: Failed to load image: Corrupt JPEG

  13. omni.usd LoadModule errors can be ignored.

    [Error] [omni.usd] USD_MDL: in LoadModule ...

  14. Following errors can be ignored when running ROS 2 Navigation and ROS_Nav2_Waypoint_Follower ActionGraph. These errors occur when starting and stopping simulation without ticking the impulse node inside the ROS_Nav2_Waypoint_Follower ActionGraph. The prim attribute for OrientReadPrimAttribute and TranslateReadPrimAttribute nodes will be set after ticking this graph.

    [Error] [omni.graph.core.plugin] /Graph/ROS_Nav2_Waypoint_Follower/OrientReadPrimAttribute: [/Graph/ROS_Nav2_Waypoint_Follower] Prim has not been set
[Error] [omni.graph.core.plugin] /Graph/ROS_Nav2_Waypoint_Follower/TranslateReadPrimAttribute: [/Graph/ROS_Nav2_Waypoint_Follower] Prim has not been set

  15. When aux_output_level is set on an RTX sensor authoring class (isaacsim.sensors.experimental.rtx.Lidar, isaacsim.sensors.experimental.rtx.Radar, or isaacsim.sensors.experimental.rtx.Acoustic), the following warning may appear in the log. This is harmless — the usdrt Fabric cache does not mirror VtArray<std::string> attributes, but the attribute is correctly set on the USD prim and read by the Replicator pipeline.

    [Warning] [usdrt.population.plugin] [UsdNoticeHandler] Unhandled attribute type VtArray<std::string> (prim attribute: _replicator:rendervar:GenericModelOutput:channels)

  16. CUDA driver failures from the omni.sensors.nv.lidar.lidar_core.plugin (example below) on Ubuntu may be due to a system-level CUDA installation mismatch with the omni.sensors runtime-compiled libraries.

    [Error] [omni.sensors.nv.lidar.lidar_core.plugin] CUDA Driver CALL FAILED at line 522: the provided PTX was compiled with an unsupported toolchain.
[Error] [omni.sensors.nv.lidar.lidar_core.plugin] CUDA Driver CALL FAILED at line 548: named symbol not found

    One workaround may be to set the LD_LIBRARY_PATH enviroment variable as follows, where </path/to/isaac_sim_installation> should be replaced with the path to your local Isaac Sim installation.

    export LD_LIBRARY_PATH=</path/to/isaac_sim_installation>/extscache/omni.sensors.nv.common-2.5.0-coreapi+lx64.r.cp310/bin:$LD_LIBRARY_PATH

  17. Physics Inspector “failed to find internal joint” errors for robots with mimic joints does not affect the functionality of the mimic joints and can be ignored.

    [Error] [omni.physx.plugin] Usd Physics: failed to find internal joint object for PhysxMimicJointAPI at /Franka/panda_hand/panda_finger_joint2. Please ensure that the prim is a supported joint type and is part of an articulation.

  18. The omni.kit.telemetry extension startup error with code (error = 206) on Windows is caused by a file path exceeding the length limit. Verify that the file path of omni.telemetry.transmitter.exe does not exceed 260 characters.

  19. Python errors related to tkinter like the following indicate the user is attempting to use tkinter with the Python distribution shipped with Isaac Sim. This is not supported.

    File "/path/to/isaac_sim/installation/kit/python/lib/python3.11/tkinter/__init__.py", line 38, in <module>
    import _tkinter # If this fails your Python may not be configured for Tk
    ^^^^^^^^^^^^^^^
ModuleNotFoundError: No module named '_tkinter'

  20. Error when using depth sensor AOVs. The AOV number (eg. 38 below) may change, depending on the selected AOV.

    [Error] [rtx.postprocessing.plugin] DepthSensor: Texture sizes do not match: inColorTexDesc 1920x1080x1:11@0 inDepthTexDesc 1500x843x1:33@0
[Error] [rtx.postprocessing.plugin] DepthSensor: Failed to allocate view resources for view 1 device 0
[Error] [carb.scenerenderer-rtx.plugin] Failed to export AOV 38 to render product. The renderer did not generate the AOV texture

Hang#

  1. On windows, when using the extension manager, clicking on the dependencies window will cause the viewport to go black and hang.

  2. The WebRTC Client on Firefox may appear to hang after a few seconds when clicking the Play button. Using the Google Chrome or Chromium browser is recommended.

Crash#

  1. Using compound nodes in OmniGraph may lead to a crash, we do not recommend using compound nodes in OmniGraph.

  2. On multi-GPU systems, building a MobilityGen scenario and then closing the stage (File > New / Open / Exit) can cause a fatal crash during teardown. As a workaround, pin CUDA work and rendering to the GPU that drives your display. Set CUDA_VISIBLE_DEVICES to its physical GPU index, then select CUDA device index 0 for rendering:

    CUDA_VISIBLE_DEVICES=<physical_render_gpu_index> ./isaac-sim.sh --/renderer/multiGpu/activeCudaGpus=0,

    For standalone SimulationApp workflows, pass "active_cuda_gpus": [0] in the launch configuration. See Multi-GPU for details about physical and CUDA GPU indexing.