ROS Resources: Documentation | Support | Discussion Forum | Service Status | Q&A

Isaac ROS, hardware acceleration for autonomous robots

Open Robotics and NVIDIA are collaborating on hardware acceleration.

As part of this effort, we announced Isaac ROS at ROS World 2021 to deliver performant hardware acceleration packages to ROS developers for industrial grade autonomous robotics applications.

Isaac ROS provides native ROS2 packages of modular nodes which makes it easy to integrate high-performance computing into existing ROS applications. These nodes are flexible to incorporate in a pipeline or as a single node all while supporting the debugging and visualization tools available in ROS2.

Our latest release includes ROS2 Foxy packages for AI perception with image processing and deep learning, tested on Jetson AGX Xavier with JetPack 4.6.

  • Stereo visual inertial odometry (60 fps at 720p)
  • DNN model inference for custom and pre-trained DNNs with included examples for DOPE 3D pose estimation & U-NET semantic image segmentation (pre-trained PeopleSemSegNet 25fps at 544p)
  • AprilTag detection (52fps at 1080p)
  • Image pre-processing (lens distortion correction, color space conversion, scaling)
  • Stereo depth estimation (disparity and point cloud)
  • Camera support (CSI & GMSL interface imagers)

Isaac ROS is available now at Clone the repositories you need into your ROS workspace to build from source with colcon alongside your other ROS2 packages.


The packages are great.
Can they be used in a commercial system? The license is not clear like BSD or Apache is.

1 Like

Thank you. The packages bring great value for robotics.

Yes, it can be used for commercial applications.

Great, can you show me where it says that in the license? Commercial companies need legal clarity before they can integrate a package.

1 Like

This is an early access release for developers to use, and developers/companies can use to develop a commercial product. We will update the license in a release planned for later this year to enable distribution too as part of a commercial product.

Thanks for the explanation.

Thank You for great work, but the package is not working for CUDA backend for maximum disparity greater than 64.
How we can solve this problem?

Thank you for making use of it! We can dig into the problem better in a GitHub Issue.

Thanks for answer. We have already raised an issue here github issue

1 Like

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.