In search of questions to ask

  • Hi everyone,

    I have just joined your forum in the hopes of better understanding the complex world of industrial robots.

    In short, I'm an analyst looking at this field for the time and I wondered if people could help steer me in the right direction. I am doing an analysis piece on wider automation and ABB v Fanuc v Kuka v Yaskawa.

    I know that 'which is the best robot' will be a highly subjective topic, so I am rather asking for questions you believe I should be asking myself at this early stage. What do you believe are things I should be looking at to understand how the industry works and what different players bring to the table? What drives one's decision to automate? How does one choose which robot to buy? These sorts of things.

    I realize that this is a very broad and open-ended topic, so feel free to channel your answer in specific areas you are most interested in.

    Thanks a lot for your help!


    Edited once, last by Aza ().

  • AD
  • TCO is probably the biggest single item, the "long pole in the tent." It gets pretty complex to calculate in advance, though. Many customers that have sunk costs in a particular robot brand tend heavily to use that robot for tasks that a different (usually more expensive) robot would be more ideal for, because of the associated additional costs of a separate, parallel training/spare parts/maintenance track.

    Likewise, a task that would be very easy to automate may end up too expensive for a customer who has no established talent/manpower pool that could handle the upkeep and maintenance. For example, imagine a small machine shop that could improve its throughput simply by adding a machine-tending robot to put raw stock into a CNC machine, and remove finished parts. The robot might be affordable, but once you add in the costs of safety, guarding, additional manpower and/or training specific to the programming and operating of the robot...

    The latest trend in "light" industrial robotics has been collaborative robots, which attempt to resolve the previous case by making robots that don't need guarding, and are much easier to set up and operate for basic tasks. But so far, these robots are limited to low-payload, low-speed, low-accuracy applications.

    One trap many new users fall into is buying an older used robot, then discovering the hard way that spare parts and skilled maintenance are hard and/or expensive to come by. Another is buying a cheap robot that ends up being inadequate to the task, for whatever reason, or naively assuming that a robot can be programmed to perform any task a human does, equally well. It's NOT that easy -- one reason robots are used to much more in automotive production than in other industries is that the robots, and the production side of the automotive industry, have co-evolved over the last few decades to converge on parts and processes that are optimized to the strengths of industrial robots (repeatability, consistency) and avoid their weaknesses (random variations in parts, complex "judgement calls").

    Integrating industrial robots is still as much an art as a science in many ways, which is why professional integrators are so often found as middlemen between robot manufacturers and end users. They can save an inexperienced end user from an expensive dead end, but are also an additional cost.

    The thing to keep in mind is that the robot is often the lesser cost driver, as long as you avoid a serious mis-match between the task requirements and the robot's capabilities (and always take the robot spec sheets with a hefty grain of salt). A robotic system designed for use by professionals will be less expensive than one that has to reach iPad levels of user-friendliness. One rule of thumb is that getting the task to work takes perhaps 30% (or less) of the programming man-hours, but getting it to work well, reliably, long-term, and capable of handling (or failing gracefully) for all the myriad foreseen (and UNforeseen) failure modes, takes 70% or more.

  • Thanks SkyeFire. This is all very interesting.

    Here are some questions that came to mind as I was reading your post. Feel free to answer as many as you want. This is quite a long list and some are probably not very intelligent questions.

    Anyone else reading this is, of course, more than welcome to answer my questions or better perhaps, steer me towards more interesting ones. I'm really just here to learn.

    1) You mention sunk-cost (infrastructure, training etc) as a reason for sub-optimal purchases. I already have a Kuka robot doing x. I'll get another one for y even though ABB is slightly better there. So many question is: is technology differentiation a worthwhile factor to focus on or do other dynamics such as the availability of trained operators or a reliable local after-sales service, really what customers focus on when buying a robot?

    2) Any ball park estimate of how much the TCO compares to the initial robot price? I have seen somewhere that installation is basically as expensive as the robot itself. And what about maintenance? Would that, over the robot's lifetime, be around as much as the initial purchase price?

    3) What would be consumer considerations when buying that first robot? Is there any robot company out there that could be seen as having the upper-hand if you exclude all legacy costs from training etc?

    4) What about once you have bought your first robot and you expand your installation base? Are there brands out there that stand out as having better usability, better after-sales service?

    5) You mentioned customer stickiness being caused by training. Are there certain brands that have visibly more people familiar with their systems? Will this greatly change by geography? For instance, if factory X uses ABB, will company Y, 5 miles down the road, be more inclined to use ABB too?

    6) When reading through hyped-up reports on automation, collaborative robots are always shown front and centre. "here is the future" sort of thing. What do you think needs to happen for these slow, inaccurate robots to eventually gain steam? Or perhaps do you not believe they will?

    7) It's interesting to observe that collaborative robots are mostly coming from smaller companies like Rethink Robotics or Universal Robots. Any thoughts as to why that is? Do you believe the Big Four could also enter that space, perhaps with a more robust, reliable, simple product?

    8) What I get from my research (and a lot of reading on your forum) is that introducing robots is often a high-friction event. To me, that means the best robot companies are those that make this switch and subsequent expansion as smooth as possible. Do you think customers would be more willing to buy collaborative Fanuc robots than more high-tech ones from Rethink Robotics?

    9) I'm not quite sure I understood your point about iPad levels of user-friendliness. Are you saying that it is better to have a system designed for use by professionals not because it is slightly cheaper but because it will save money down the line by reducing the number of hours needed for programming?

    Thanks a lot for your help!

    Edited once, last by Aza ().

  • 1. Since for most common tasks, one brand is about as good as another, often the extra logistical burden of training, spares, etc for two different robot brands just isn't worth it. Most major buyers (the automotive industry) have company-wide standardized packages -- a particular brand robot, bundled with a specific set of option packages optimized for that customer. So that one can (in theory) walk into any of that company's plants anywhere worldwide and find the very same infrastructure. In reality, it's never quite that good, but that's the ideal.
    OTOH, sometimes you need to bite that bullet. 20 years ago, Chrysler was using almost nothing but Nachi, but for their glass-decking robots, they bought KUKAs, because applying the sealer required fine motion control that the Nachis of that era couldn't provide. Still, tasks that require a particular brand tend to be at the outer edges of the bell curve.

    2. TCO depends on so many external factors, it's impossible to guesstimate. The support structure for a single robot could easily costs as much as the robot, or 2-5x that much, depending on many factors. With more robots working together, economies of scale can kick in. Maintenance... well, it really depends. Even a cheap robot can last a good long time if it's taken care of well, and a top-line robot can be worked to death in 5 years or less if it's abused and neglected. It's much like an automobile in that respect. There are "lemons," so avoid things like the first year or two of a brand-new model, and get some independent user testimonials.

    3. For a first-time robot buyer, I would think that finding a good integration partner is a higher priority than selecting a particular robot brand. For one thing, a general integrator that has experience with multiple brands can probably help you find the optimum price/capability choice. For some very complex realtime-control jobs, you might need a KUKA with special option packages, but for simple pick&place or spot-welding, a Nachi or Kawasaki will probably do well enough.

    4. I'm generally biased towards KUKAs, but this will vary with geography. The overall system, and the integration between then different sub-sections, is more likely to drive cost that the individual sub-units. So unless you get a lemon robot, or really abuse your robot, your maintenance costs from your integrator will likely be a bigger factor than that of the robot(s). A good integrator can help you with that robot-level maintenance as well, with their inside contacts.

    5. Varies wildly with geography, and with industry. Comau robots, as far as I know, are only used in-house (including in Chrysler, now that Fiat (who owns Comau) owns them), but Chrylser went through an all-ABB stage briefly, and before that was Nachi-only for a long time. GM went from Fanuc-only to KUKA-only for a while, and back to Fanuc (IIRC). Ford was all-Kawasaki for along time, in North America, while VW and BMW were almost exclusively KUKA for many years. Also depends on whether you're in Europe, NA, or Asia. KUKAs appear to dominate in Aerospace so far, but that's a very young market.

    6. Cobots are the perfect entry-level robot for small, low-payload jobs that need to avoid the costly guarding and training associated with their big cousins. But while they're cheaper, easier, and safer, they're not going to replace the big dogs anytime soon. More likely we'll see a gradual upwards creep of Cobot-type capability into larger robots over time. So if you need a robot to operate without a safety fence, and be taught by people with utterly minimal training, and don't need high speed/payload/accuracy, and need something that will run off wall-socket power (as opposed to 480VAC 3-phase), Cobots may be what you're looking for. Like, a small mom&pop bakery that starts getting more orders than they can handle -- you could, maybe, just plop a Cobot in front of the oven and let it handle pulling finished trays out and putting fresh trays in, while the human workers concentrate on keeping the dough going. Nothing like industrial production rates, but nothing like industrial cost, either, and no one's going to get killed if they get between the robot and the oven. And you don't have to fence off half the room, or rebuild the floor.

    7. Well, the Big 4 (and their competitors) have a lot of sunk cost, and a lot of ingrained habits and groupthink. So for something really off-the-wall innovative, it's always easier for a newcomer looking for a new, empty niche to get into the game. Cobots fly directly in the face of a lot of established wisdom about robots, but that's because they're attacking a very different market segment, with different priorities and specialities. The big 4 have been chasing bigger payload, higher speed, and tighter accuracy for so long (because that's what the big customers wanted) that it became almost self-fulfilling. This left a market niche for smaller players, once the technology existed to make small Cobots easy, safe, and cheap. Whether the Big 4 will chase that market will depend on how much profit potential they see there -- right now, it looks like they're leaving it to UR and Rethink.

    8. That's going to depend more on what the customer has used previously. If they use Fanucs, they'll be much more likely to try a Cobot by Fanuc than one of the new companies. A customer who is a "robot virgin" is probably going to be neutral on that, beyond any potential brand name recognition. This is part of why UR, and especially Rethink, have put effort into making their robots not just user-friendly, but "friendly" and non-threatening -- the Baster Cobot's 'face' display being a major example. The main target for UR and Rethink (so far) is people who've never used a robot before and have no idea what they're getting into, although there is also some penetration into new parts of companies that have used robots in the past -- for example, most robot use in Automotive has been in the welding shop -- trim&final has been primarily human work. But a few Cobots appear to be making inroads there, though it's early days yet.

    9. More like getting over the training hump. Getting to minimal competency on a Nachi, KUKA, Kawasaki, Fanuc, ABB, takes at least a week of (expensive) training, and for anything advanced probably needs 2-3. Which usually requires flying out to the vendor's training facility, staying in a hotel, etc etc. A Baxter, OTOH, can be plopped down on a normal floor, plugged into a wall socket, and probably be taught some basic tasks with an hour's worth of RTFM. The barrier to entry is just much lower. It's a bit like the difference between learning to use a consumer-grade 3D printer, and a low-end CNC mahine. It's much easier (and safer, and hence less intimidating) to learn the former, even though many of the underlying skills and technologies are shared between both. A 3DP can probably serve as a "gateway drug" for CNC machines, once the 3DP user starts bumping into the limits of the technology. Similarly, Cobots might well be an easier "on ramp" to automation for customers who just couldn't make the startup investment (in time, money, risk, training, etc) directly to big robots.

  • Thanks for the detailed answer SkyeFire. I was not expecting to get so much out of this forum! This is great. :bravo:

    I’ll try my luck and throw another salvo of questions your way. Again, everyone’s welcome and don’t hesitate to tell me if I’m asking the wrong questions and should rather be thinking about X Y Z rather than A B C.

    I have two broad fields of enquiry at this point: 1) dynamics that prompt automation with traditional ‘dogs’ as you call them, and 2) rapid expansion of the industry because of cobots.

    Traditional dogs

    1) How do you control a robot arm exactly? I know Fanuc makes CNCs but these are for mills lathes etc. Does each robot arm manufacturer make its own operating system, or could I control an ABB arm with Fanuc controllers?

    2) Is there a brand that, regardless of geography, would be considered simpler to install, use and repair?

    3) Is there a Big Four that is generally cheaper than the other 3?

    4) Surveys I read indicate the main obstacles to installing robots are cost and ease of use (training, installation etc), but that doesn’t explain why some countries are lot more automated than others. Why is it that Germany has a higher ‘robot density’ in general industry (excluding automotive) than the US? Is it just because it makes more stuff? (btw, is robot density at all a good way to think about this?)

    5) Same question for different industries. Obviously automotive is wayyy higher than any other industry I can find. Is there something that makes this industry more automatable than others? What are characteristics I should be looking out for to understand why industries have automated and why new industries might automate in the future?

    6) Short follow-up from 5. Does new automation usually come because new robots can do more things or because people gradually become too expensive (in this latter case, the tech of the robots would already be somewhat mature)?

    Cobot theme

    Here are some things I take away from your previous post
    - Cobots are much cheaper overall because of usability (little training needed) and they don’t need a pen
    - Cobots are probably not going to replace big dogs, but smart tech will rather creep upwards
    - New entrants like UR and Rethink are targeting ‘robot virgins’, who are indifferent to brands
    - Big 4 have been pursuing different goals

    Here are my questions

    1) Do you see cobots becoming a large market, possibly comparable to bigger dogs? If so, would most applications come from new customers or industries already using robots? In other words, which do you perceive as most automatable and also most willing to automate?

    2) As you hinted earlier, no company can really claim a technology supremacy. Buying a Nachi or ABB usually comes down to path dependency and support networks. Should we expect this to change with cobots? In other words, is the tech here a lot more important? I guess this goes back to asking "what do customers want": new exciting robots or simple, reliable.

    3) If cobot tech starts creeping upwards, could we imagine a scenario where UR and Rethink eventually steal Big 4 sales because their technology is better? I’m thinking of your Chrysler ‘biting the dust’ example here. Or conversely, is there a case to be made for Big 4 players waking up to the need for technology that will eventually be a part of the entire product line? I am thinking Fanuc Green robots or KUKA LBR iiwa. I know any answer can only be speculative but I am interesting in hearing how you think, in light of the growth opportunities from question 1, the industry will evolve. Will the Big Four have an IBM moment where their focus on mainframes makes them incapable of competing against PC makers, or is the progress more incremental and thus puts a Big 4 at an advantage (brand, repair network, trained users etc)? Or are both coexisting peacefully? Fanuc’s CR-35iA has a promo video showing it helping an auto worker, not a restaurant chef.

    Edited once, last by Aza ().

  • Hm... how did I miss this for two weeks? Oh, well, better late than never....

    1. This gets into forward/inverse kinametics, Denavit-Hartenberg parameters, and lots of other things. In general, industrial robots never got the degree of cross-brand standardization that, say, CNC machines got with G-Code. There is essentially zero cross-compatibility between brands in either hardware or software -- the only place the brands can work together is at the peripheral-communications level, and that's because every brand has no choice but to support the major fieldbus standards used across all their customers. Every brand has their own OS and programming language, and guard their secret sauce jealously.
    Now, in theory, one could control any robot arm from any computer with enough mathematical horsepower, but in reality, there's so much low-level dynamic control and parameterization, not to mention hardware differences, that it's usually much easier to just buy the arm and controller as a package. ROS/Orocos is about the only system I've ever seen that takes an "open source" approach to robot control, but it just doesn't have any traction in industrial use. "Generic" multi-axis controllers like the Siemens 840D can, and have, been used to replace brand-specific robot controllers, but there's generally not much value added there -- a more common approach has been to take a full robot&controller package, and add an option to it that allows you to "slave" it to the control system of your choice (KUKA robots with RSI or FRI, the LWR series, and so on, have been done this way).

    2. Not really, no.

    3. Nachi is, as far as I know, the cheapest of the Big 4, although I think Fanuc actually sells cheaper per-unit for large-scale purchases (but they make it back, with interest, on parts&service). There's an entire (small) banking sector dedicated to financing this stuff, on the large scale.

    4. Again, it's mostly TCO. There's a lot of connected costs to automation (not "hidden", as such, but not obvious to the newcomer), above and beyond the simple purchase cost of the hardware. Robots aren't good at manufacturing processes that take a lot of hands-on "look and feel" judgement-call work -- that's why you find the body-welding assembly lines in automotive almost 100% robotic, but the trim/chassis/final (where all the "pretty stuff" is added, like dashboards, seats, trim pieces, rubber moldings, etc) are mostly human-run still. Robots are good at processes that can be done by simple rote, repeating the same motions every time, every 30sec, for years, or jobs where "chasing the moving target" can be reduced to a simple mathematical algorithm. Things that involved lots of hand-eye coordination, and wiggling/fitting finicky things into place, are still more reliably done by humans. One reason Automotive is so robot-dense is that the entire design of the parts and assembly has been driven towards optimizing for robotic performance over decades. Aerospace, on the other hand, has issues automating because all of their production processes and parts are still purely designed for being done by hand, with every part being more "crafted" than "assembled".
    So... for any industry to automate, it needs a few prerequisites: a "foot in the door" process, one that can be automated relatively easily, and at low risk (you don't want to have to completely rebuild your production facility if the robot in the corner doesn't work out). Processes that are more expensive to do with humans than robots -- for a long time, automotive paint was done by robots even though humans were actually somewhat better at it, because of the health&safety costs of human-operated paint booths. Reliable infrastructure -- you can't automate in a location with a bad power grid, or semi-regular floods. A labor pool that has the education to keep the automation going in between costly field visits by the manufacturer -- fewer bodies, but more expensive per-body.
    Political issues can also have an effect -- some countries have labor laws, or militant unions, that can make the "friction" of trying to automate so high that it gets given up on.
    Japan has an increasing problem with lack of labor, combined with (for whatever reason) a near-fetishistic love of robots (I blame Astro Boy, personally). Germany tends to be pursue quality and productivity, and goes into automation more as a production leverage, but uses more expensive automation than most other "automation-dense" countries. China is starting to turn towards automation because they're starting to see some manufacturing work slip away to other countries who haven't (yet) experienced China's wage inflation. And in the USA, re-shoring is becoming a thing in part because automation can offset the apparent (but arguably inflated) cost savings of sending work to cheap-labor nations.

    5. Phew. I'm not an historian or economist, but my worm's-eye view: Automotive benefitted from a confluence of:
    A. major demand and growth, and competition (lots of three-shift operations for popular models)
    B. Parts and processes already oriented towards assembly-line operations (lots of small repetitive operations)
    C. Very high production rates (robots do small repetitive tasks faster and longer than humans)
    D. lots of jobs that were Dirty, Dull, and/or Dangerous (an entire generation of auto workers tended to have missing fingers or other scars)
    E. labor union issues, rising labor costs, combined with issues finding enough labor (see A)
    F. Changes every model year. This may seem counterintuitive, but while robots can't handle unit-to-unit variance very well, they're better at being reprogrammed for a new model car every year or three, than "hard" automation. So the automotive industry needed the flexibility of robots to be "re-tooled" for new tasks periodically, while also having long-duration part runs where nothing changed much that were friendly to robotic repeatability and lack of fatigue.
    G. All this happening as the technology for the first really practical robot arms hit the scene.

    Aerospace, by comparison, has never had the production rates, and is only now really finding a need for greater economies of scale. Agriculture had too many variables and moving parts, and is only recently starting to see serious inroads of robotics automation (as opposed to "hard" automation like mass milking systems, which still required human hands to attach/detach the fiddly bits). And there's lots of industries that are heavily roboticized, but less obviously so -- chip fabs, for example, where clean-room requirements meant few humans ever had those jobs, but the small/fast payloads make for robots that don't really impinge on public consciousness as much.
    Then there's "midway" machines -- a CNC machine, or 3D printer, is a "robot" in most ways that count, but a very specialized robot -- while what we usually "think" of as "robots" (outside of SF like R2D2 and C3P0) are "arms", which can be switched relatively easily from doing welding, to paint, to palletizing... there's a lot of "robots" out there that we just don't think of as "robots." And lots of automation, like smart HVAC systems for buildings, the ABS or engine controller on most cars, smart cruise missiles... the list goes on.

    6. Both. There are things I can do with robots now that I wouldn't have dared even try when I started out 20 years ago. And the "cost" side can be driven by either labor cost going up, or automation cost going down. For a parallel, look at Gaming PCs -- the cost of a good gaming rig has stayed about the same for many years now, but the power at that price point keeps going up. Similarily, robots haven't gotten a lot cheaper, but the bang/buck keeps increasing, to the point that Cobots have now started to enter the market at much lower TCO levels.

    1. Cobots are drilling up into the market's "soft underbelly," where robotics simply haven't been available or affordable. I expect the convergence to come from both ends, as the Big Dogs see a new market segment and try to get into it, while the NewBotics companies keep making their machines smarter, faster, and stronger to chip away at the low end of the Big Dogs' market share.

    2. It's going to be a PC/Mac type of question. Most buyers will want bots that "just work," no muss, no fuss. But there will always be the customers on the edge of the bell curve that want something just a little bit beyond the machine's comfort zone, and are willing to pay (to a degree) to get it. These latter customers will be part of what drives the R&D sector of the Cobot market.

    3. The market seems ripe for disruption. New players (like FoxCon) are coming into the "big dog" market, and UR/Rethink and their ilk are slipping in the back door. For the immediate future, I would expect Cobot growth to be more "horizontal" than "vertical" -- there are hard limits on Cobot safety relating to speed/strength/inertia, which will be hard to get over, but the current models' inherent safety and user-friendliness will probably keep opening doors among new customers that are "under" the low end of the Big Dog's target market. So I'd expect to see Cobots expanding laterally from light manufacturing to greenhouses, bakeries, fast food, etc, before they start colliding with the Big Dogs. Then again, when ARPANET got started, no one predicted the World Wide Web, so....

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account
Sign up for a new account in our community. It's easy!
Register a new account
Sign in
Already have an account? Sign in here.
Sign in Now