Robot Operating System 2.0: Ardent Apalone Release | Hacker News Reader

Robot Operating System 2.0: Ardent Apalone Release

github.com

72 pointsbmc750511 comments

Good to see that coming out the door. It's still in beta, though.

They need to get their act together and get this into the standard Ubuntu repositories for installation using the standard GUI tools. Or install into non-system folders. ROS is notorious for breaking Ubuntu auto-updating because of bad packaging.

I tried following their instructions, and the install didn't happen because, for some reason, their repository wasn't recognized. Don't have time to fix this; will wait until they're out of beta.

ROS has historically been such an install pain that some people install it in a Docker container. That has its own headaches.

I haven't used ROS in two years but it was such a painful experience at the time. We were stuck on an old, old build because we found that some of the drivers we used didn't work on anything newer. I spent so much time trying different Ubuntu versions paired with different ROS versions.

And this was with pretty popular consumer hardware. I can't imagine the experience with anything exotic.

Did they fix the build system or are they still going to use the weird catkin build system? I read the article, but it didn't mention it. I ask, because it has been consistently the most difficult part of using ROS for myself and everyone I know. Alternatively, is the difficulty I've experienced more of an indication of build systems being complicated things?

They are moving to ament. http://design.ros2.org/articles/ament.html

I think a lot of the difficulty is in the build system being federated. This makes it complex as it supports multiple build other systems for other ROS packages. I haven't looked into it much yet, but I'm curious about ROS 2 also supporting ROS on smaller micros now. I think this would make the build system more complex.

Modifying the how the devel space is handled seems like an improvement and making it more modular seems like it should help though.

I have only used ROS for a few simple projects, but I found catkin (which is actually a thin wrapper around cmake) relatively easy to grasp. As far as I know, catkin is actually an improvement over previous build system, rosbuild.

You can look up more informed and lengthy threads/holywars about build systems, but in my experience, they always end up being complex for large projects. Some IDEs try to hide the complexity behind dozens of GUI tabs (build settings), but that usually makes headless/CI builds complicated, and sometimes even impossible.

Also, what I have seen a lot is: when people are struggling with build systems, they usually don't have a clear understanding of major underlying steps that need to be performed to build the executable: invoke compiler to compile individual source files to corresponding object files (with correct include paths and compiler flags for architecture) and then invoke linker to link them and any required libraries together, producing an output executable.

Once this process is clear, reading the documentation of particular build system to achieve what you need gets a lot easier.

Changelog between ros 1 and ros 2 here http://design.ros2.org/articles/changes.html, and you can find the general explanation for ros 2 design here http://design.ros2.org/

As a lead developer in a ROS-based project, I'm not sure whether we'll adopt ROS2 in a couple of years or abandon it altogether. It seems like with ROS2 supporting integrated CMake builds will be a lot more difficult, and in our company we strongly prefer monorepos and integrated builds due to ease of debugging and optimization.

On the other hand, ROS2 aims to provide RTOS support and pluggable transport layers, which are rather serious benefits.

Also, their decision to abandon XML-based roslaunch is the right one: we had to roll out our own startup infrastructure as well, as it's much more convenient to have a definite init order with sequential checks and waits e.g. for hardware availability.

I think a lot of work has been undertaken for a ROS1->2 bridge. It should be feasible to use this to have ROS on a micro and bridge to an existing ROS 1 code base.

Has anyone played with the RTOS implementation? I used ROS a ton for my thesis but was always left longing for true real time actions. I'd love to hear the story of someone who has tried to do a real project with it.

> Why ROS 2.0? http://design.ros2.org/articles/why_ros2.html

ROS = Robot Operating System

Reply on news.ycombinator.com