I have an application with a simple warehouse room and an RTX lidar placed within it.
I am using a python script node that collects the latest Lidar result and parses the data to a ROS2 point cloud2 message node.
I specifically use the generic ROS2 publisher node because it allows additional message fields like “Intensity“.
I am also using the following carb settings to ensure that the RTX lidar intensity data responds to materials: --/app/sensors/nv/lidar/enableRtxReflectanceInformation=true --/app/sensors/nv/lidar/enableAdditionalRtxReturnInformation=true --/rtx/materialDb/nonVisualMaterialSemantics/prefix=omni:simready:nonvisual --/app/sensors/nv/materials/enableMaterialInterLayerContribution=true --/rtx/materialDb/nonVisualMaterialCSV/enabled=false --/app/sensors/nv/lidar/enablePolarization=false
In the below images you can see the applications starting memory usage while running the lidar at a very high resolution 10Hz (2048 * 128ch).
After 2 hours the application reaches 81GB which is rather odd.
This program does have a carb warning that will show each simulation tick: 2025-11-17T12:22:03Z [2,018,720ms] [Warning] [omni.graph.core._impl.utils] Could not sync USD on attribute inputs:simTimesToWrite - Error in 'pxrInternal_v0_24__pxrReserved__::UsdStage::_SetValueImpl' at line 6703 in file C:\g\201060795\USD\pxr\usd\usd\stage.cpp : 'Empty typeName for </Render/PostProcess/SDGPipeline/DispatchSync.inputs:simTimesToWrite>'
Could this be causing the memory issue?
Below is the point cloud data visualized with RVIS2 for those who are curious as to the result.
Thanks for reporting your experience with expanding memory usage when publishing RTX Lidar data to ROS2 in Isaac Sim. Your setup and workflow—using a Python script node to collect Lidar results, parse to ROS2 PointCloud2, and leveraging the generic ROS2 publisher for custom fields like “Intensity”—closely match the recommended integration patterns in Isaac Sim.
Are you using Isaac Sim 5.1.0? It’s possible that this issue can be reproduced using RTX Lidar Sensors — Isaac Sim Documentation as it follows a similar process: setting up RTX Lidar sensors, streaming data, and publishing over ROS2 using the bridge. If this memory usage behavior is present during standard tutorial steps—especially with high Lidar resolution and publishing frequencies—it may indicate a deeper bug or inefficiency in either sensor data management or the ROS2 publishing pipeline.
To help diagnose and escalate this issue, could you confirm if you observe similar memory expansion when following the tutorial workflow with default settings? If so, sharing your steps and monitoring details can be valuable for the team.
We noticed that this topic hasn’t received any recent responses, so we are closing it for now to help keep the forum organized.
If you’re still experiencing this issue or have additional questions, please feel free to create a new topic with updated details. When doing so, we recommend mentioning or linking to this original topic in your new post—this helps provide context and makes it easier for others to assist you.
Thank you for being part of the NVIDIA Isaac Sim community.
