Hi @maikelborys
I was reading your solution, but how did you fix it?
Best,
Raffaello
Hi @Raffaello
Let me break down my solution and setup in case it helps you or others.
π» My System:
- Laptop: ASUS TUF15 RTX 4070
- OS: Ubuntu 22.04
- ROS 2: Humble
- Camera: Realsense D455
π οΈ Root Issues Identified:
- β Wrong Realsense firmware/driver
The camera was not working correctly with default librealsense installed via apt.
β Fix: Downgraded firmware to 5.13.0.50 as recommended by NVIDIA.
Steps:
- Download firmware:
https://www.intelrealsense.com/download/19295/ - Update camera firmware (with only 1 camera connected):
rs-fw-update -f Signed_Image_UVC_5_13_0_50.bin
Reference:
- β Mismatch between CUDA version and Docker image
I had CUDA 12.2 installed, while Isaac ROS containers were using CUDA 12.6.
β Fix: Installed CUDA 12.6.2 to match the container:
- β Old NVIDIA drivers
I had to uninstall all NVIDIA components and install driver version 575 to match my RTX 4070 properly.
RESUMEN
I completely removed Isaac ROS, all NVIDIA-related components, and Docker images. Then reinstalled everything from scratch
π§Ή Full Cleanup Performed:
- Removed all:
- NVIDIA drivers
- CUDA versions
- Docker images (Isaac ROS containers)
- isaac_ros-dev workspace
- Rebooted the system
β Fresh Installation Steps:
- Installed NVIDIA Driver 575
- Installed CUDA 12.6.2
- Followed official environment setup:
Developer Environment Setup β isaac_ros_docs documentation - Verified Realsense firmware = 5.13.0.50
- Followed Realsense ROS setup:
Isaac ROS RealSense Setup β isaac_ros_docs documentation - Continued with cuVSLAM/Nvblox tutorial:
Tutorial for Visual SLAM Using a RealSense Camera with Integrated IMU β isaac_ros_docs documentation
Since then, everything works perfectly β camera feeds, VSLAM, and Nvblox integration inside Docker containers.
Happy to share more details if needed!
Gracias,
Maikel
2 Likes
Thank you very much for sharing!