Ask Your Question
0

ROS for industry is a good choice or not?

asked 2020-05-05 11:56:38 -0500

Hyperion gravatar image

I made exactly same question on ROS Course on Udemy with Edouard, but I wanted to hear from you about this...

This question is based more on the experience than the pure knowledge of ROS. I heard that ROS is not complete thread-safe when working with multiple subscribers/publishers and is not good to implement in machinery control for robust automation (like create an automation for a machine with ROS and put in main line production as part of automated system p.e.). What you think of that? Is ROS1 good for create robust automation? Is better to wait for ROS2 or ROS Industrial? Or you have another technology that would you choose instead of ROS (not talking about PLC or similar) in order to develop robots or robust machinery? A friend of mine in China tells me that ROS is used only for tests or in universities for robotics studies, but he never saw a robust machinery using ROS 24/7 on a shop floor (he works as supplier for Delta, Advantech and Staubli automation).

PS.: recently we saw the Acutronics closing the doors for not find investors for his H-ROS (https://www.therobotreport.com/acutro...). That's is not directly related to my question in some way?

Sorry for my bad english...

edit retag flag offensive close merge delete

Comments

Whether it is a good choice or not is going to be very heavily influenced by the exact requirements of the application, the context in which it is going to be used and the skills of the engineers and developers involved.

As your question does not really provide any of those details, I'll only link you to the ROS-Industrial website. I would recommend to take a look at the public videos and slides of the ROS-Industrial Consortium Americas Annual Meeting and the videos and slides of the ROS-Industrial Conference 2019. The best evidence of the applicability of ROS in industrial contexts -- except visiting the factories and talking to the developers -- is probably watching those.

gvdhoorn gravatar image gvdhoorn  ( 2020-05-05 14:04:19 -0500 )edit
1

A friend of mine [..] tells me [..] he never saw a robust machinery using ROS 24/7 on a shop floor (he works as supplier for Delta, Advantech and Staubli automation).

There isn't really any appropriate response to this sort of statement, as too many details are lacking, and it's impossible to make a statement like your friend did without qualifying it much more (in which contexts? Which requirements? What sort of applications? Etc, etc).

A friend of mine in China tells me that ROS is used only for tests or in universities for robotics studies

Interestingly, surveys show something else (from here):

Finally: ROS is certainly not perfect, neither ROS 1 nor ROS 2. But it doesn't have to be.

It has to be good enough.

gvdhoorn gravatar image gvdhoorn  ( 2020-05-05 14:09:02 -0500 )edit
1

I purposefully did not post my response as an answer, as I don't believe there is an answer.

If 10 years talking about ROS in industry have taught me anything, it's that the scope of potential application and of (non)functional requirements is just to broad, there is no single answer to the question whether ROS can (or should) be used.

I could have answered with that, but I doubt that is what @Hyperion came here to hear/read.

And of course: posting this on ROS Answers is most likely going to result in somewhat biased answers/comments. Not just because ppl posting here will have (great) experience with ROS, but at least the professional users have a tendency to be working for or with a company which uses ROS.

Dissenting opinions are rare on ROS Answers, but I can point to one: @Sebastian's answer to #q316916 ...(more)

gvdhoorn gravatar image gvdhoorn  ( 2020-05-06 03:19:17 -0500 )edit

Some related Q&As: #q237173 and #q52254 (old).

gvdhoorn gravatar image gvdhoorn  ( 2020-05-06 03:29:26 -0500 )edit

Ah that was a nice thread, I got another good chuckle reading that again.

stevemacenski gravatar image stevemacenski  ( 2020-05-06 13:24:36 -0500 )edit

1 Answer

Sort by » oldest newest most voted
5

answered 2020-05-05 20:37:16 -0500

updated 2020-05-05 20:37:50 -0500

I'm answering this so verbosely not only pointed at your question, but since I get this question alot, I want to write something out so I can forward this to future question askers. Its a friendly rant :-)

In addition to the good metrics from the report that @gvdhoorn posted, I can tell you that out in Silicon Valley I can hardly name a robotics company who's product _isn't_ running ROS. I think I was able to come up with 3 that I know for sure don't use it in any way. All are heavily research focused MIT or CMU spin outs, so that's not shocking to me.

  • Simbe
  • Fetch
  • Savioke
  • Vicarious
  • Built
  • Boeing
  • Cruise (autonomous cars)
  • Apex (autonomous cars)
  • Bossa Nova
  • Bosch
  • LG
  • Toyota Research Institute
  • Honda
  • BMW's industrial automation groups
  • Amazon Robotics and AWS
  • Certain groups in Google
  • iRobot recently
  • and I work for Samsung Research, which you can make whatever implications you like from.

The list goes on. There's even massive industrial giants forming the ROS TSC (https://index.ros.org/doc/ros2/Govern...) including LG, Samsung, Amazon, and Microsoft. From talking even to autonomous driving companies, many more of them use ROS(2) or ROS derived tools, more than you might think as a casual observer. Additionally with the entrance of folks like Apex, I think you'll see even that space converge back into ROS-land.

On your specific points:

I heard that ROS is not complete thread-safe when working with multiple subscribers/publishers

If you write bad code, you'll have bad ROS code. That's not a ROS problem, that's a bad use of ROS problem. I can say that I've had robots with over a hundred days of constant up time on a ROS designed system. Some are probably approaching a year.

is not good to implement in machinery control for robust automation

That's all based on your requirements. Though most reasons why I would say "well, maybe" aren't true anymore. ROS2 takes care of most of those. Its suitable for those. If you're trying to spin something at 100KHz, you may be right, its not the most optimal communication framework for that sort of low level control, but that makes up remarkably little of any autonomous robot system.

Is better to wait for ROS2 or ROS Industrial

Well to be clear, ROS2 is here, we're on the second LTS release and there are people building products with no ROS(1) in it at all anymore. ROS-I is a consortium of companies working on standards and projects, its not a separate thing from ROS itself.

edit flag offensive delete link more

Comments

Hi people.

Before anything, I loved the ROS! Mainly due to the possibility of being as a wrapper for my robots, enabling me many possibilities for communication and virtualization through Industry 4.0. I even started to read and use V-REP, RoboDK and MORSE to learn more about it.

I appreciate the answers and the links, thank you all .

I realized that in a way, my question made you very uncomfortable and defensive about the use of ROS. As if my question was to denigrate the ROS in a way, when it is actually the opposite. I genuinely would like to hear from you (experienced in the subject) about the statements I RECEIVED from my contacts and my experience on the ROS.

Before I explain the reasons for my previous question I would like to say a few things:

  • I'm automation developer in Brazil. Working in one of the ...
(more)
Hyperion gravatar image Hyperion  ( 2020-05-06 08:57:26 -0500 )edit
  • My supervisory systems usually I make using C ++ and Java;
  • From 2019 until a short time, experienced (at the request of one of my clients) develop without using PLC and an industrial PC with Python and ROS;
  • I had previously had contact with the ROS on my master's degree in computer science, creating an AGV to the Pioneer 2-DX. However, the machine that I built now is very different from an AGV.

I can not give details of the machine development as it is a confidential information and is owned by my client. However, assure that the machine was developed with the ROS, C ++, Python and EtherLAB and is working. It operates with its virtualization including Industrie 4.0, OPC-UA and publishing real-time informations like +120 I/O's, encoders, control / status words and many other variables.

Well, here's my report of what I saw on the ROS ...(more)

Hyperion gravatar image Hyperion  ( 2020-05-06 08:58:31 -0500 )edit

First, I believe the report that many companies like Samsung and LG for example using the ROS should be on R&D and not shop floor environments (correct me if wrong please). For what I see is quite the contrary. Since I started to develop this machine, not only my suppliers did not support the ROS to even have heard of it.

I asked if they knew at least one case study to base myself and nothing looks ... And I asked people working in Beckhoff, Keyence, Kollmorgen, Oriental Motors, Advantech and others and living in numerous companies shop floor . And that was one of my main barriers: particular problems of ROS, PyQt and EtherLAB with my project. Unfortunately I didn't receive any response except from this community (the EtherLAB I had no answer, solved more than 95% of the problems alone). Just sit in the chair and study (this is ...(more)

Hyperion gravatar image Hyperion  ( 2020-05-06 09:04:37 -0500 )edit

After that, I took up the idea of ​​building other robots with ROS, including for clients such as Samsung, LG, 3M and others. And they felt extremely insecure about using ROS, even though I have a functional and feasible case. They still prefer to pay +$100k in semi turnkey solutions than investing in a new product for themselves - they already work with such technologies and there is always that shock. And NONE of these companies heard of ROS also - at least not in local sectors of R&D and on the shop floor. I do not take their reason, after all the business solutions such as Rockwell, LS, Schneider and others are very good and comfortable to use. Which leads me to believe that my machine is the first, if not the only machine developed with ROS of all industrial center here in the area.

Hyperion gravatar image Hyperion  ( 2020-05-06 09:07:53 -0500 )edit

So I inquired myself for this and did such a question. They were able to understand the nature of my doubts? After all, I'm newbie in ROS and need to learn more about it. However, was put in my place and tell me what you think about this situation?

Hyperion gravatar image Hyperion  ( 2020-05-06 09:08:33 -0500 )edit

And again, i apologize if i passed a wrong impression of ROS being bad or something. That is not what I meant. What I want is simply to know the views, biased or not, about ROS. After all, I can read and come to a conclusion for myself. As in a debate. Thanks in advance.

Hyperion gravatar image Hyperion  ( 2020-05-06 09:14:58 -0500 )edit

I’m not going to be able to get to all of that detail, but I think it boils down to why isn’t ROS used in specifically your corner of the robotics world, industrial automation.

Industrial automation is used to buying a black box that does XYZ (controllers, PLCs, etc) and you buy N black boxes to plug them in to create your application. It typically doesn’t involve alot of development and software know-how to do the “systems integration” of that area.

Many times to those folks, software is the enemy and they want as much as hardware logic as humanly possible.

Its not surprising to me there’s push back or lack of interest because they’d rather just buy a black box which everyone else uses and works immediately vs developing a custom solution which can take longer and more prone to error. Those big automation ...(more)

stevemacenski gravatar image stevemacenski  ( 2020-05-06 13:36:03 -0500 )edit

I realized that in a way, my question made you very uncomfortable and defensive about the use of ROS. [..] And again, i apologize if i passed a wrong impression of ROS being bad or something. That is not what I meant.

I did not interpret your question like an attack or derogatoray at all.

If you understood my comment as being defensive then I believe that to be due to "the email effect" (ie: it's hard to express nuances with plain text alone, and they are easily lost).

It's perfectly OK to ask these questions. You should ask them.

gvdhoorn gravatar image gvdhoorn  ( 2020-05-06 13:37:54 -0500 )edit

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Question Tools

1 follower

Stats

Asked: 2020-05-05 11:56:38 -0500

Seen: 354 times

Last updated: May 05 '20