Coding Robots of industrial production line using ROS Industrial
Hello, I work for an international company that sells automotive parts to the Automotive industry such as Renault, Peugeot.. They have over 70 production lines in their warehouse on the site where I am currently.
Their automotive parts are assembled using robots (up to 7 roobots on one Production line). The robots are coded using Staubli language VAL3 on software SRS, sometimes other robot brands are used. There are vision systems, PLCs and sensors communicating with other PLCs that send the states of the product/line to robots, so the robots check what to do next.
Those production lines are usually developped by integrators (other small companies that develop the lines and sell them to the international company I work for) this is done for garantee purposes and so on..
In the maintenance process time is always wasted because of robot points that should be retaken for different reasons. I find this super annoying because it's like regenerating the trajectory of the robot.
I would like to know if developing the robots and coding them with ROS would make the process better or worse, in terms of cycle time, safety , reliability... Would it add a layer in terms of hardware/software and make things "unsafe" or slower like some people tell me?
I love coding with ROS2 and I do not understand if its better to code the robots of an industrial production line with the robot language like the integrators do or if its more beneficial to code with ROS2, notably ROS Industrial.
I want reasons to try implementing a production line with ROS. Is it already implemented in some industries in PRODUCTION LINES (not stand alone applications where the robot does things alone and does not rely on other parameters of an entire line like for example medical robotics or mobile robotics). I have worked with ROS2 on mobile robots (LIDARS, Cameras, IMUs..) and I know that ROS is very robust.
I want to know if in this specific application it is usedful and beneficial.
Thanks!
I expect you'll get quite a few subjective replies and anecdotes in response to your post here.
To start with one (ie: a subjective answer): There isn't really any other answer than: yes, it is being used for production work. The scale however would probably be one of the main differences. The kind of traditional automation you describe can include tens to hundreds of robots on one line. That's not something I've seen in the context of "ROS used with industrial robots". I'm also not sure it would make sense for those kinds of applications though.
As to the rest of your questions: anything can be done, or made to work. It's all software, just as the proprietary OS of Staubli (and other OEMs) is. The bottleneck will be how much time and effort you are willing to spend to get something to work ...(more)
[..] But we should not misrepresent things: no, I don't believe anyone would claim ROS (1 or 2) is at the same level of certification, safety, robustness, etc as any of the proprietary solutions the traditional automation OEMs offer. How could it be? They've been developing their solutions for over 60 (or more) years now, with huge budgets and customer demand. That's not something an OSS community can match, or at least not in just 10 years.
On the other hand, we should also not pretend the proprietary solutions of OEMs are perfect. Far from it actually. I've personally encountered enough 'unexplained' behaviour and missing functionality in Staublis, KUKAs, FANUCs, Yaskawas and Siemens controllers (to name a few) that I believe it'd be more honest to say all software has problems, and that ROS is just one more tool in your toolbox, and it could be ...(more)
[..] Have a problem where traditional automation falls short? ROS could be worth a try.
Perfectly happy with what your preferred colour-of-robot OEM provides you? Please continue using it.
Of course it's slower. Anything which adds anything more than "simple" interpolation between joint/Cartesian poses and toggles a few IOs while moving an EEF around will be slower than what can be done with traditional automation.
(Exaggeration, but that's basically what industrial robot motion controllers do)
I believe ROS only starts to make sense when you have a (largely) unstructured environment, unstructured or unknown work objects, partly unknown processes, processes which require online control based on sensor data, motion which depends on sensor data (and I'm not referring to seam-tracking), production runs of a single object (which necessitate on-line program generation based on CAD/3D scan ...(more)
[..] As nice as VAL3 is (or KRL, INFORM, TP, G-Code or RAPID are), implementing any of the above in those environments is either impossible or very complex, or requires special support (ie: custom development) from the OEM in question.
Summarising: for the type of automation you describe, I don't believe the value proposition of ROS (1 nor 2) is currently interesting. For anything beyond what traditional automation can do, it does make sense to invest time in it.
PS: I've posted this as a series of comments as it's not an answer to your main question (ie: "could I use ROS for [this task]?").