How to handle ROS skeptics

Hey folks!

I have a simple question: How do you handle co-workers that are skeptical of ROS?

Background: I’m trying to make the case for ROS2 at work but one of the key stakeholders is skeptical of ROS. He comes from a long career in aerospace where everything is thought through and there’s a huge focus on determinism, availability and dependability.

He seems to think that ROS2 isn’t a “big boy” framework but he’s also open to learning more.

In his words: “I need to be convinced on ROS, but that’s more of a lack of detailed knowledge vs. religion. It’s probably worth a virtual sit-down and have you give me a guided tour. The ROS stuff that you find online is a little too rah-rah for my taste”

9 Likes

This is a good question, although personally I don’t like to use a word like convince, as it always sounds like there is an involuntary aspect to it.

What I’ve noticed is that people tend to have a few specific things they’re unsure about. If you can find out what those are, that would help, as it narrows down the problem significantly.

This might be something to ask: what exactly qualifies a framework as a big boy framework? Is it the level of use? Scale of deployment? Number of products sold using it? Companies using it? Specific development processes or standards used? Having gone through specific types of audits or certification? Etc.

It’s easier to answer concrete questions and dealing with facts than sentiments, as the latter are really difficult to change.

Additionally, I’ve also found that – but this may be due to the audience I typically talk to – talking about actual concrete results and showing videos about them helps.

So instead of talking about how things could be with ROS (1 or 2, doesn’t matter), show what they actually already are. This changes the perspective from some potentially idealised picture to concrete, verifiable facts.

Bonus points if you can discuss use- and edge-cases coming from the domain your audience has experience with/in. They should be able to relate to those, and the solutions chosen, much more than say a TB2 driving around the bedroom of a PhD student.

Finally: be honest. As was discussed in ROS2 Default Behavior (Wifi), things are not perfect, fault free or TRL9 everywhere but that’s ok. Any engineer should be able to understand that technology is never completely as it seems on the outside, or made out to be in videos (n==1 of course). The important thing is it’s being recognised and worked on, and in the meantime not marketed as a perfect solution.

However, even with things being as they are, there may still be sufficient ROI or advantage for someone (or multiple someones) to start using it, simply because it would seem not using it would be more costly.

9 Likes

It really depends on his motivation.

  • One over-arching issue is the permissive license where ROS users aren’t really compelled to talk about their wins and successes (even though I hear about them all the time off the record).
  • If the root of the issue is the classic open source debate that has been settled long ago for most applications except a few verticals (graphics / games, and CAD).
  • One logical failure I hear all the time is the belief that ROS is a monolith; where it is an all or nothing solution, which is patently false. Generally ROS can be easily integrated with whatever proprietary system floats your boat.
  • I’ve seen a number of orgs attempt to replicate open source solutions because of quality concerns, only to get mired in years of development. Writing anything from scratch is only a worthwhile endeavor if you value your time at zero and aren’t concerned with time to market.
  • Along those lines, the things we do as robotocists are hard. They require research, hypothesis testing, and iteration. That’s what ROS allows to happen quickly. You can’t iterate if you are stuck building infrastructure.
  • There is still a mindset, based in ROS’s history, where people believe it is an academic enterprise. I usually point people to the TSC as proof that this isn’t the case. If companies like Amazon, Toyota, Ford, Fraunhoffer, the US Government, and Bosch, being active ROS contributors doesn’t convince you I don’t know what will.
  • In Aerospace, Spirit Aerospace gave a great talk at ROS-I early in the year. They are big converts. I would suggest you reach out to ROS Industrial for an intro.

I compiled a list of a few metrics to help out with social proof. I would suggest that most of them are big under counts. We’re trying to work on a few more things to tamp down on the FUD.

image

13 Likes

Here’s another data point, it is a bit on the marketing side, but I just went through this list and about half the organizations in this list either use ROS, support a ROS interface, or work with Open Robotics directly.

The other thing I wanted to add is that like Python there are a lot of ROS packages out there, some of them written by the smartest people I know, and some of them written by people just starting their software journey. If you judge a federated project by its least skilled contributors you’re doomed to throw the baby out with the bath water. We’re aware of this issue and most of the community is working to address it and make those distinctions clearer.

As an aside, your boss dismissing a whole community as “not a big boy framework” could be interpreted as both ageist and sexist, and might be indicative of a sub-optimal engineering culture. :wink:

9 Likes

This information that you show in the image is available?

1 Like

It should come out in a media document soon. Is there something in particular you are looking for?

2 Likes

In the coming document a few more of those stats are added but most of those numbers are available in the annual metrics reports or on https://metrics.ros.org with more details related stats and more context.

4 Likes

This is a bit an argument by authority. If we allow those, perhaps the list of companies on the ROS-Industrial membership could also help. Many of the names @Katherine_Scott mentions have their logos in the montage on that page.

4 Likes

I think the broader narrative here is when you go through the lists of businesses using ROS there are larger orgs, that are capable of acquisitions, who have shown a vested interest in ROS. There is also an ecosystem of smaller companies that can be acquired to bring innovation to market. Moreover, the community size means that there is a labor pool that are capable of working with ROS based systems. Hiring and training are major concerns when you are trying to innovate and beat competitors to market. The support of these large orgs signals a willingness to use ROS as a defacto standard for innovation. What some might call an appeal to authority can also be considered a viable market opportunity.

1 Like

Not now, but the times that I did some presentation about ROS I didn’t find, but good to know

Humm, really nice Tully, I think that, when I made the presentation I couldn’t remember of this repository, but I’ll keep it in my radar. Thank you very much!

Forgive the cynicism, but it feels like, for a lot of more older… huh I mean more experienced people (:wink:), a “big boy framework” is one that is old (to the point of being antiquated), hard to use, and expensive to use/maintain, but it’s “just how things are.” Which basically means it’s because they don’t really know any better!

It’s very similar for people that keep saying “ROS 2 isn’t ready” (vs. ROS 1) but the last time they actually looked at it was 2+ years ago, so perhaps we need to have a post like this but specific to ROS 2 vs. ROS 1?

@You_Know Your question is related to this thread probably: Consolidated ROS2 feature status In particular: “And how the feature status of ROS2 is w.r.t. functionality required in domains in which ROS1 features have been non-ideal/missing.”

1 Like

Wow what a great discussion @Katherine_Scott @gavanderhoorn @tfoote @fkromer @felix.widmaier @pxalcantara!

I’m blown away by the activity and immediate support in this discourse. Why didn’t I join sooner?

@Katherine_Scott You made a lot of great points! I’ve taken your points and put it into a cheat sheet for when I have the conversation later today. PDF of my cheat sheet attached.

While putting this document together…I realized it would be nice to have a ROS “developer evangelist guide” that helps folks like me who are in companies “sell” ROS into their organizations. Most companies that have been doing robotics before ROS have a hard time making the leap from handcoded TCP middleware and apps written in basic C++ to ROS. Perhaps the guide could have your points and a list of questions that @gavanderhoorn brought up? Developers aren’t sales people by nature and I think the approach that @gavanderhoorn brought is a good way to explore people’s objections and resolve them. If I wanted to start something like this with the goal of putting it up on ROS.org…how would I go about this?

Case for ROS.pdf (488.2 KB)

2 Likes

I totally agree especially on these two points.

Whether they actually think that it’s a monolith or communicate it as such, this is a major problem I’ve seen. It’s open source. They should use the parts they want and leave the rest. There are clean APIs and other interfaces to use what they want.

I’ve also seen way too many custom frameworks. It begins innocently enough to “maintain control of the source” but again there are ways to maintain control by using ROS as well. I also think ROS has the added benefit of being successful in a wide variety of use cases. These successes mean that it will be more flexible to future changes in your project. It has the breadth and depth to allow a large amount of flexibility which is key especially in changing business climates.

I’ve just never understood these two arguments. I think those making these assumptions should take a step back and really spend some time evaluating the situation further.

1 Like

:+1:

This is great. I think we can cook up an evangelism guide down the line. If you do create some sort of glossy sell sheet feel free to send it to me (this is a good draft but it needs a few revisions, probably with a designer). I have some plans for ROS.org that are brewing but I can’t disclose them just this second. Suffice it to say I have a few things in flight that might get us the resources to update ROS.org with the specific intent of improving professionalism and visibility while addressing business and community concerns.

This is a great question. I have been trying to get scientists (biologists, neuroscientists, and chemists) to use ROS for their experimental rigs since the boxturtle days 10 years ago. It has been a very difficult sell. There have been several well-known biology experiments and papers using ROS, but it has not become as popular as I would have hoped. I feel like there could be a tipping-point soon, though, and it could take over the science world if enough scientists see how much it could improve their work and their data. Most scientists are not trained in computer science and fear Linux, C++, and command lines. Proprietary software has a strangle hold on many academic labs unfortunately. ROS 2 could connect all of this software to the open source robotics world, though, and join together these disparate communities. Scientists would both benefit from the fantastic work all of you have been doing for all these years and be able to contribute back their unique expertise to make the ROS community even bigger and better.

1 Like

Anecdotally, in my work with the Open Source Hardware Association, we’re seeing a slow but steady up-tick int he number of open hardware projects used for research. This, along with the push for open access to journals seems to be reaching a tipping point. I think a good way to push for this is to demonstrate ROS as a suitable replacement for LabView. Unfortunately, I don’t think we, as a community, do enough work in this direction. If someone in the academic sciences wanted to actually do a push in this direction, showing 1:1 correspondence to LabView features, I would love to support it.

1 Like

But there is no such correspondence. At least not on the UX side of things, which is what “fear of Linux, C++ and command lines” basically boils down to for lots of potential users.

With Labview (and other NI software), as much as I don’t like it, it is possible to hook up a sensor board to a Windows laptop over USB, click an interface together, connect some wires to the AD converters and you’ll be recording data and plotting figures in 5 minutes, to give one example.

Controlling motors? As long as your motor board is supported, here is a USB cable and a Rio board, off you go. And it’s all hard real-time as well.

Whether something is possible – with additional development – is not important.

I fear a comparison between software such as Labview and ROS 1 or 2 would not be as positive as we may expect/hope.


Edit: I realise this sounds rather negative, especially coming from someone who has been saying similar things for so many years (ie: everyone should use ROS if it makes sense for them), but I also feel it’s good to stay realistic.

2 Likes

Yeah agree the nascent capability is there but the UX isn’t. When I say 1:1 correspondence I mean examples of how to replicate common features with an open stack of tools (that are presently imperfect). If another organization wanted to find NSF/EU funding to extend ROS in that direction I would whole heatedly support it.