There is some really good discussion on this in the 2018 thread, but I'd really like to see a ROS 3 that conforms a little more to the more widely-used alternatives to the stuff they had to roll for themselves back in the day.
Agreed. There are countless things in ROS1 which feel like a less polished version of X. I believe ROS2 is trying to use a more commodity stack for IPC and serde.
Kind of a weird choice for the name, the title of the piece is "The hottest thing in robotics is an open source project you've never heard of" and "The Robot Operating System (ROS) has moved out of academics" is a sub-section (and makes more sense as such, since it's in contrast to ROS's origins, not any recent event). Aside from letting you know ROS exists this piece is fairly free of content, but I guess it's a neat thing to be aware of.
I didn't know it was considered as more academic, seeing as I saw it in production on some next level stuff. At least it was dockerized and I didn't have to wrangle it.
Is ROS any easier and more reliable than it used to be? I've tried a few times, out of curiosity, to tinker with it, but have never been able to get the thing to work well enough to go through tutorials.
I always found Oxford's MOOS to be easier to work with. That plus something like MIT's IvP Helm objective controller will get you an autonomous mobile platform very quickly.
Cool thing about ROS is that is works for all sorts of robots, while MOOS [0] is focussed on autonomous marine vehicles.
With ROS, I can have motion planning for a manipulator in hours, then put in on a mobile robot with whatever sort of locomotion etc all in the same system.
> MOOS [0] is focussed on autonomous marine vehicles
Core MOOS is just an efficient low-overhead IPC middleware, exactly like core ROS but, in my experience, smaller and faster and easier to use. You can use it for literally any communication between multiple distributed processes, robotic or otherwise. I think ROS became more successful at first because of better branding and promotion by Willow Garage. Now it's all network effects.
MIT's Ocean AI lab, by virtue of being an Ocean AI lab, uses MOOS for marine vehicles, but other groups have used it for other things as well.
The answer from the old-fashioned people who didn't believe in open-source would be Microsoft Robotics Studio. MS killed that years ago.
Now Microsoft is contributing to ROS2 IIRC with ROS2 running on windows too.
Three smaller threads from 2017: https://news.ycombinator.com/item?id=15890213
https://news.ycombinator.com/item?id=15749546
https://news.ycombinator.com/item?id=15530813
Quite small from 2016, but nicely counters the OP's baity title: https://news.ycombinator.com/item?id=12080752
2013: https://news.ycombinator.com/item?id=6867480
A similar article from last year: https://svrobo.org/a-history-of-ros-robot-operating-system/
(via https://news.ycombinator.com/item?id=20490740, but no comments there)