Is there an official guide to install ros2 (foxy) on jetson nano AND ubuntu 18.04?
Thank
1 Like
Hi @leogarberoglio, there are these ROS2 Foxy containers for 18.04:
- https://github.com/dusty-nv/jetson-containers#ros-containers
- https://nvidia-ai-iot.github.io/ros2_jetson/ros2-jetson-dockers/
It’s recommended to use the containers, but if you prefer not to, you can follow how ROS2 Foxy was installed in the Dockerfile of the containers.
I didn’t know about dokerfiles. It would be the same as a typical installation from the performance point of view? I want to use it to test ROS2 apriltag.
Another question, is jetson nano compatible with ubuntu 20.04? I mean, can I install a regular ubuntu 20 and a regular ros2 foxy from binaries?
In the same way is apriltag_ros2 compatible with ubunu18.04 + ros2 eloquent?
What could be the best combination in order to get best apriltag_ros2 performance?
Thank for your help!!!
Docker containers have the same performance as if it were installed outside of containers. Containers run natively on the device (i.e. not in a virtual machine or some other kind of emulation). Ubuntu 20.04 is not yet officially supported in JetPack, so it’s recommended to stick with Ubuntu 18.04 if you plan on using CUDA and the GPU.
Hello @dusty_nv, I am experimenting with ros2-foxy in the following repo
which supposed to be very similar to the repo you provided above.
I found out that there is not many pre-built packages, hence I would have to build them from source in the Nano which is very slow. NVIDIA may want to considering building various images of ros2 (base, core, perception, desktop) as a support, or simply creating a apt-repo for pre-built packages.
P.S. Appreciate your effort for helping supporting the community with the Ubuntu 18.04. I knew there is not much you can do when the road-map of the NVIDIA for supporting Ubuntu 20.04 is still a year away.
Hi @pvmilk, you are correct in that the Foxy packages need built from source. I’ll take a look at building container images for ros-desktop in addition to ros-base. Also, can you point to the perception package you are referring to?
@dusty_nv Thanks for a quick response.
My bad on the perception packages; apparently it is only for ros1.
You might want to refer to these tags for your reference.
https://hub.docker.com/r/arm64v8/ros