ROS Resources: Documentation | Support | Discussion Forum | Service Status | Q&A answers.ros.org

GSoC and the fate of OpenEmbedded

gsoc
ros2

#1

Hello! I am currently writing an application for the upcoming GSoC which involves ROS2 and OpenEmbedded. Is it fine if I go into detail about the experiences and my vision for ROS and OpenEmbedded? I would be interested in discussing my ideas and how they line up with OSRF, I just dont know with whom and where exactly. Also what part of that vision could be implemented during the GSoC. I would stay on after that and maintain/update regularly to achieve them.


#2

Exciting to see students interested in this field @braineniac! I’d suggest to review what’s been done already. You’ll definitely find pending things to do but ensure that it contributes. To list a few:

I also think that LG engineers are commiting resources. Have a look at https://github.com/ros2/ros2/issues/607#issuecomment-469502151. @abilbaotm and @LanderU, feel free to jump in and advice @braineniac in case there’s something you consider would be beneficial and isn’t “ready” just yet.


#3

Thanks @vmayoral . I have already experimented with meta layers from jumpnowtek and meta-ros(to generate a beaglebone and raspberry image with ROS). I find it kind of weird that its still on indigo.
Also was there any work done on integrating rmw-microxrcedds from @BorjaOuterelo? Putting ROS on microcontrollers and sbcs should kinda be the point of OpenEmbedded.
Also why is the meta-ros2 GPL3 licensed? Isn’t the point of ROS2 that companies can use it, modify the layer if they want to and ship their product close sourced?


#4

Maintainers simply may not have enough time to contribute. Most of them work full-time for projects. Updating this is probably a fantastic way to demonstrate your experience with OpenEmbedded :wink: (it always helps for the application, it did in my case!)

That’s wrong. The point of OpenEmbedded isn’t that one. Moreover this was attempted in the past and you’d find that getting a fully fledged DDS implementation running in a microcontroller is somewhat complicated and not really useful. That’s what the micro-ROS project is for which builds on top of previous experiences.


#5

Understandable, I will try to set up a nice workflow and try to fix the issues on meta-ros first.

Can you elaborate why it wouldn’t be useful? Offloading intensive, high frequency calculations to a microcontroller(f.e. sensor data processing) then sending back the results sounds pretty useful. Or did I misunderstand something?


#6

Ah, I found out about design.ros.org and found the relevant discussion. I will post it here in case someone else is also interested.


#7

@braineniac I think it would be cool to get superflore to generate open embedded builds properly – I did quite a bit of work with this while I was at Open Robotics (since it is similar to Gentoo in recipe format), but lacked the proper open embedded experience to get it to function fully.

I got a little bit further than getting roscore to run, but could not quite get things to be stable/repeatable. If you have more experience here, you might talk to @tfoote about this.


#8

I am just a student and currently don’t have extensive OpenEmbedded experience, but I would like to get involved and learn as I go. First get the the superflore and bloom integration during the GSoC, then work on generating proper OpenEmbedded builds with ROS using already available layers(like the ones from jumpnowtek or similar).
I want to focus on embedded systems in my career, so I could work on this in my free time for quite a while and I should get better at it over time :stuck_out_tongue:
I will just submit my GSoC application at the end of march and if I get accepted, it will make more sense to talk about future plans knowing my full background.


#9

This post was flagged by the community and is temporarily hidden.


#10

Hello @braineniac @allenh1 (couldn’t tag more than 2 users :slight_smile: , vmayoral)! We at LG are working to get superflore to automatically generate buildable Bitbake recipes of ROS2 index packages by extending the work initiated by @allenh1 . As an intermediate target, we want to build BeanBirdBot set of ROS2 bouncy packages with the automatically generated recipes as opposed to the ones we have in order to verify the conversion is good on an actual project using them and make sure it still works. Then next step will be generating and building all ROS2 crystal packages. Here are some links to the work in progress:





It’s great to have a place to discuss ideas, issues and collaborate for this work here and you’re welcome to join!


#11

@andre.goddard wow! That’s great work! Looking forward to reviewing that upstream!

I also appreciate that you didn’t just ignore the Gentoo side of things in your efforts.

Cheers!


#12

Thanks for the update @andre.goddard! @LanderU and @abilbaotm may want to sync with you on how to align our two groups to avoid re-doing work.

Edit: actually, @LanderU, does it make sense for you to reach @andre.goddard and maybe set up a call?


#13

Sure @vmayoral, I’m more than happy to discuss/test/help wherever is necessary.
I’ll have a look into the above links to have a better understanding what you’re doing @andre.goddard. I’m open to have a call and discuss everything.

Regards,

@LanderU


#14

Don’t mean to cross post, but we’ve just started a thread with a proposal for extending OpenEmbedded support in superflore and will really appreciate if you could please take a look and join us. It’ll be our pleasure to have a call to chat further.