This release includes packages for AI perception, image processing, navigation and adds:
cuVSLAM v11 package with performance improvements and HAWK support
Optimization in NITROS improving performance and reducing CPU load (+ corresponding fix in Humble)
Updated ESS DNN model for AI stereo disparity trained dataset of 100K real images
new IMU package (Bosch BMI088)
GPU accelerated stereo rectification
bug fixes
Isaac ROS update is available now at github.com/NVIDIA-ISAAC-ROS and is part of our commitment to provide features and hardware acceleration for autonomous robots.
Clone the repositories you need into your ROS workspace to build from source with colcon alongside your other ROS2 packages and leverage the pre-built ROS2 Humble + Nav2 container. Please note that this release has been tested on the NVIDIA Jetson AGX Orin & Xavier with JetPack 5.1.1 (Ubuntu 20.04).
Our next major update will be late summer for 1.0 GA release on Jetpack 5.1.2, and a update in Oct for Ubuntu 22.04 with Jetpack 6.0.
This is exciting news! But I have some questions. Generally, Jetson software is fairly far behind the latest. The current JetPack installs 20.04, even though 22.04 was released 18 months ago. And the ROS documentation strongly discourages installing a particular ROS version on untested OS releases; Humble is not officially supported on 20.04, as far as I can tell.
This post implies that Humble works just fine on 20.04, but has been tested on AGX Orin and Xavier. It also mentions building from source and leveraging the docker container. Finally, it mentions a major update for “late summer” and October.
So, my questions:
I have ros Galactic installed and seems to be working, although I have yet to do anything more than “hello world.” Can I safely install Humble?
I haven’t tried using ROS with Docker. The post above implies that’s a common way to use ROS, but that seems like a really tedious way to go about it, having to surface all sorts of interfaces and such. Can I use Isaac (and VSLAM, et al.) directly, without Docker?
Alternatively, is there a good guide for using Docker in the context of experimenting with an autonomous driving robot? I’m hoping to get stereo cameras, 2D lidar (3D if I can find one for under $2k), motor control, steering, cliff sensors, and perhaps a few other sensors working together on this little platform I built.
Off-topic, but can I expect JetPack 6.0 any time soon?
Just for info: I have successfully built Humble from source, using ROS2 docs as a guide, for Xavier NX & Orin NX. Jetpack 5.1 & 5.1.1 tried (20.04 based).
Yes. Humble is a great upgrade to Galactic, with significant improvements including enabling accelerated computing.
Yes, you can use Isaac ROS directly without docker.
Many ROS developers compile from source, or install pre-build packages, however there is a trend to docker, which has benefits for development and deployment covered in multiple discussions in discourse; perhaps less relevant to an individual developer whom knows and controls their setup, but of great value for consistency across systems with multiple developers, including cloud-native development at scale, which is why we use it.
If this is a personal project, you could do without docker, however if this is a exercise in learning for career benefit, skill with docker is an asset.
There are multiple resources online for starting with docker. Choose the one that’s best suits your needs.
Seems an interesting AMR project.
To keep costs down, sensors including 3D LIDAR’s are free in simulation.
Not off topic.
Jetpack 6 is in development, and you can track the Jetpack roadmap. We are targeting an Isaac ROS update for GTC in March 2024 with Jetpack 6.