ROS 2 Raspberry Pi 4 image with the real-time kernel as presented in ROScon 2022 and future work (NVIDIA?)

[shuhao]

As some of you have heard me present at ROScon 2022, the ROS real-time working group has worked on a Raspberry Pi 4 image with both ROS 2 and the real-time kernel built-in. We are hoping this image will help make it easier for beginner and expert users to get started with ROS 2, real-time, or both.

I was encouraged to post here to get more visibility on the project so more people can benefit from the work we have done. The image we created can be downloaded from Releases · ros-realtime/ros-realtime-rpi4-image · GitHub.

Additionally, we also released the image builder that builds the image above. This image builder is designed in a generic manner and should be modifiable such that it supports other single board computers.

One thing that I kept getting asked about at ROScon is the possibility of supporting NVIDIA jetson boards. While this is something we are interested in, the real-time working group is mostly a volunteer-driven organization and none of us have an NVIDIA single board computer to play with. Further, it is my understanding that the NVIDIA jetson software stack doesn’t play well with the PREEMPT_RT kernel (my sources are from casual reading, so I could be wrong). That said, if there are community interest and involvement, we are happy to help with the bring-up effort.

cc: @razr @carlossv @LanderU

[baqwas]

@shuhao!

Many, many thanks for your efforts! This is a great opportunity for a beginner like myself to finally work in an environment that is most comfortable.

Regards.

[dashpod]

That’s great. Will there be a substantial difference when I use RPi4 with real-time kernel when I integrate it with a real robot (turtlebot 3 for example)?

[shuhao]

I don’t think there’ll be any substantial difference in performance. My understanding is that most of the nodes launched by ROS are using the normal process scheduler (SCHED_OTHER). It should neither see improvements nor substantial performance degradation.

This is why we are OK with recommending this image as a general purpose image even though it has a special kernel installed.