When I run cuvSLAM v3.2.0 version, there will be an error message "cudaError Unknown"

Robotics & Edge Computing Robotics - Isaac Isaac ROS
,
1

After updating isaac ros visual slam to release 3.2.0, when I compile the source code, the following error message will appear.
Moreover, CUDA has already been installed in the environment, and I am running it using Rosbag data.
What are the reasons for this message and how is it possible to avoid it?

ros2 launch isaac_ros_examples isaac_ros_examples.launch.py launch_fragments:=visual_slam interface_specs_file:=${ISAAC_ROS_WS}/isaac_ros_assets/isaac_ros_visual_slam/quickstart_interface_specs.json rectified_images:=false
[INFO] [launch]: All log files can be found below /root/.ros/log/2024-12-30-17-21-02-693214-fusion-N16PR1-34278
[INFO] [launch]: Default logging verbosity is set to INFO
[INFO] [component_container_mt-1]: process started with pid [34290]
[component_container_mt-1] [INFO] [1735550463.020889808] [isaac_ros_examples.container]: Load Library: /opt/ros/humble/lib/libimage_format_converter_node.so
[component_container_mt-1] [INFO] [1735550463.067458938] [isaac_ros_examples.container]: Found class: rclcpp_components::NodeFactoryTemplate<nvidia::isaac_ros::image_proc::ImageFormatConverterNode>
[component_container_mt-1] [INFO] [1735550463.067523511] [isaac_ros_examples.container]: Instantiate class: rclcpp_components::NodeFactoryTemplate<nvidia::isaac_ros::image_proc::ImageFormatConverterNode>
[component_container_mt-1] [ERROR] [1735550463.086753065] [NitrosContext]: cudaErrorUnknown (unknown error)
[component_container_mt-1] [ERROR] [1735550463.086787336] [NitrosContext]: [NitrosContext] setCUDAMemoryPoolSize Error: GXF_FAILURE
[component_container_mt-1] [INFO] [1735550463.086797179] [image_format_node_left]: [NitrosNode] Initializing NitrosNode
[component_container_mt-1] [INFO] [1735550463.087091565] [NitrosContext]: [NitrosContext] Loading extension: gxf/lib/std/libgxf_std.so
[component_container_mt-1] [INFO] [1735550463.090015952] [NitrosContext]: [NitrosContext] Loading extension: gxf/lib/libgxf_isaac_gxf_helpers.so
[component_container_mt-1] [INFO] [1735550463.096294162] [NitrosContext]: [NitrosContext] Loading extension: gxf/lib/libgxf_isaac_sight.so
[component_container_mt-1] [INFO] [1735550463.103046970] [NitrosContext]: [NitrosContext] Loading extension: gxf/lib/libgxf_isaac_atlas.so
[component_container_mt-1] 2024-12-30 17:21:03.116 WARN  gxf/std/program.cpp@538: No GXF scheduler specified.
[component_container_mt-1] [INFO] [1735550463.116266448] [image_format_node_left]: [ImageFormatConverterNode] Set output data format to: "nitros_image_mono8"
[component_container_mt-1] [INFO] [1735550463.116357155] [image_format_node_left]: [NitrosNode] Starting NitrosNode
[component_container_mt-1] [INFO] [1735550463.121553565] [image_format_node_left]: [NitrosNode] Loading extensions
[component_container_mt-1] [INFO] [1735550463.121602871] [image_format_node_left]: [NitrosContext] Loading extension: gxf/lib/multimedia/libgxf_multimedia.so
[ERROR] [component_container_mt-1]: process has died [pid 34290, exit code -11, cmd '/opt/ros/humble/lib/rclcpp_components/component_container_mt --ros-args -r __node:=container -r __ns:=/isaac_ros_examples'].
2

Hi @corneliayu1028

Did you install everything? It’s may missing the VPI
https://nvidia-isaac-ros.github.io/getting_started/hardware_setup/compute/jetson_vpi.html

Best,
Raffaello

3

Thank you for replying!
It run on x86 platform, which don’t need the VPI.

4

Hi @corneliayu1028

Could you run “nvidia-smi” to check version of CUDA and NVIDIA GPU Driver if they meet the requirements below? If the problem persists, would you mind re-installing the GPU Driver then reboot to try isaac ros visual slam again?

  1. Check discrete NVIDIA GPU with the following specs:
  • supports CUDA 12.6+ with Ampere NVIDIA GPU Architecture or newer
  • minimum 8GB of VRAM and recommended 12GB+
  1. Install the latest NVIDIA GPU Driver from official instructions (here). (Minimum 560+ is required, see here).

Let me know.

Best,
Ahung