I’ve been putting together a ROS+OpenCV robot kit for hobbyist ie “Makers”. What does that mean? Well the thought is - simple ROS software stack that can be extended but relatively easy to kick off and have the “hardware” to something hello-worldy quickly; tolerant to a level of “scrappiness” ie wires hanging out, not drop-proof tolerant; lastly economical and extensible using hardware that’s readily available via any of your favorite online sources.
I feel that is similar to the “duckiebot” developed by duckietown.
Also, I think that image processing with a raspberry is a little complex because you would lost many frames (works like 10 fps if you apply some filters with decent quality (640x480)).
It could be interesting use a odroid ux4 or some main board with more power and a lower cost.
I actually got the Duckietown ROS software stack running on our Raspbian ROS OpenCV image. Our ROSbots kit is functionally equivalent to the duckiebot.
I reached out to the duckietown team members to see if we can work together. If you know anyone on the Duckietown team, would appreciate an intro. I believe our two efforts complement each other very well (we are focused on HW, them on curriculum framework for ROS).
As for vision on a RPi, totally agree. I’m thinking at having a ROS container and/or VM that runs on a laptop/server (maybe even in the cloud) do the heavy lifting computations, while pointing at the ROS_MASTER_URI on the ROSbots Pi running ROS. Assuming you have the bandwidth to stream the images from the Pi Camera to the machine at 15-30 FPS (which shouldn’t be a problem compressed at 640x480), you can run vision, localization, whatevers with another ROSbots “brain” living off the actual RPi. Hope that makes sense!
Appreciate the odroid suggestion but I’m trying to avoid requiring anyone to buy a >$50 SBC and/or co-processor.
Let me know if you have further feedback or suggestions.
BTW - what are you working on using ROS? Love to learn more.
I’m participate of duckietown Chile, but i don’t know anybody in the MIT team or another team.
That make sense, is a very good idea process the data in a powerful computer to get a better result, could be in a docker it’s a easy and quickly form to test.
I am a undergraduate of Computer Science student and I am working with a Parrot Ardrone with [pharo] (https://pharo.org/) and [LRP] (https://pleiad.cl/research/software/lrp) , a live programming language designed for the creation of the behavioral layer of robots. I am also part of duckietown Chile where we use robots to teach ROS to second grade engineering students.