Willow Garage and the Making of the Robot Operating System | Hacker News Reader

Willow Garage and the Making of the Robot Operating System

spectrum.ieee.org

75 pointsilamont19 comments

It's not exactly encouraging to see, on the Willow Garage ROS website, that the page seems to have been last updated in 2013 (http://www.willowgarage.com/pages/software/ros-platform). You'll instead want to try http://www.ros.org/ or go directly to the installers, located at: http://wiki.ros.org/lunar/Installation

However, the latest version (Lunar) on OSX is described "This is a work in progress! It really won't work right now... " and I've verified their instructions don't work. Instead, you'll want to use this shim: https://github.com/mikepurvis/ros-install-osx

Hope this saves some folks time.

Willow Garage is defunct. They closed doors in early 2014. This point is not really made on their website.

What's more troubling to me is I've been hearing about ROS 2.0 since 2014, and no progress (at least outwardly) has been made.

ROS 2.0 is under active development, with many talks on it last ROSCon.

You can follow it at https://github.com/ros2 and https://discourse.ros.org/c/ng-ros

It's going for a Beta release this December 8 [0]

Some of the main goals of ROS 2.0 are support for QoS and (small) embedded devices, Windows support

[0] https://github.com/ros2/ros2/wiki/Roadmap

ROS probably should go under Linux Foundation as well.

ROS 2.0 is rebuilding a lot of infrastructure. Plenty of development is happening; the github repos as I type this show a lot of repo updates that are minutes to days old. That said, the current state of ROS 2.0 is probably well described as "developer only". Current ROS-based commercial robots are still on ROS 1, Kinetic or even Indigo.

That's good to know. I guess if you're in the ROS community this stuff is apparent, but on the periphery it's not clear that ROS 2 is even a thing judging by the ROS website. I just thought in 3 years, I would have seen it mentioned more prominently on the ros.org domain, so I concluded it was stalled.

Luckily, lot's of commercial parties seem very interested in ROS 2.0 which gives me some good hopes for a less painful transition from ROS 1 to 2 than from Python 2 to 3...

But given the federated development of many non-OSRF maintained packages, the bridge between ROS 1 and 2 is going to be used for a long time, I'm afraid. But it's possible at least to mix the versions.

The Willow Garage page is dead because the company went out of business.

Trying to run ROS on anything but Linux is a waste of time.

It works well under Windows via the Linux Subsystem, at least the transport layer (which is what most people I've talked to use it for anyway.)

Ah, OK. I checked, and while transport is fine, it didn't seem to support cameras or other hardware.

I'd prefer it to be Windows-native, rather than WSL.

I've had weird bugs with ROS on WSL. RosPy works, ROS CPP however seems to fail to subscribe.

I run it on my Mac at work all day. It's not a waste of time but you really do need to commit to the maintenance. Mike Purvis' shim listed in a comment above helps a lot.

Yes, well, a version of ROS has always been tightly paired with single version of Ubuntu, for better or worse. There is only so much bandwidth to support so much. These days, Kinetic (the current LTS) and Lunar are pretty portable to many Linux distros, but anything else is probably an exercise for the intrepid, although ros.org will happily take patches. But if you are just starting out learning ROS, then my advice is to hew as closely as possible to the documented/supported configurations. There is enough other stuff to learn to keep you busy without fighting other subtleties.

Even on the correct version of Ubuntu, there always seems to be some version clash / version pinning problem which breaks the Ubuntu updater. ROS itself is three things - some message-passing middleware, the Gazebo world simulator, and a distro of various robotics-related packages from others. The distro part consists of software from maybe a hundred different sources, more or less duct-taped together. Some of the packages in that distro work better than others.

A good way around this is to run ROS exclusively from Docker containers. With the right permissions set up, it works fine with USB devices too.

https://hub.docker.com/_/ros/

I used the ROS a few years ago with Professor Jenkins at Brown University. It was great, I sat in my wheelchair in the UK and flew a drone remotely in the Lab. in Rhode Island using just my head and right index finger. It was great :-)

It was easy to set up and use and pretty reliable at the time. In a similar vein I used NodeCopter to do something similar between Yorkshire and London at a Wired conference as well. That used a web interface and lots of JavaScript goodness to achieve roughly the same affect. It was loads of fun, but a very high pressure thing to do to try not to hit journalists in the head with a drone!

My next big step is to go for the whole FPV experience with my set of limitations, I started trying to do just that last year but with one thing and another (read: hospitals and stupid quadriplegic stuff) it didn't come to pass. It definitely looks like fun though as I had some really cool advice from Kevin and a few other people. I'm really hoping to get back off the ground again at some point soon. (See what I did there!)

Here's the video of the nearly hitting the editor of Wired magazine in the head with a drone in case you're interested:

https://www.youtube.com/watch?v=_XM97VvRSRM

Edited to add: if anybody from ROS needs a quadriplegic geek to help test anything, get in touch. :-)

Most of my life is figuring out how to get people who've made a whole bunch of stuff on ROS a way to migrate that stuff, and a platform to migrate it to, when they get to a point where all that cobbled together R&D work needs to turn into something safety-critical.

I've always thought ROS was largely meant for prototyping. An http transport layer for all nodes (even on the same machine) doesn't seem like the right choice for a production system.

Kind of. I've seen the approach of building on top of ROS and incrementally replacing the weakest link work well. You end up with something that feels academically wrong to be in production spaces, but you also end up with something that actually sold and shipped. Establishing a product feedback loop sooner is more valuable than selling a technically "correct" product.

Also it's TCP (or UDP if you want) not HTTP. ROS2 does one better with DDS.

Reply on news.ycombinator.com