My company wants to move over to robotics and they want me to make it happen. Where do I start??

  • Hello everyone,

    It was unclear to me whether or not this was the right forum for this thread. I do apologize if its not.

    So, the company I work for is in the sound and light business. Buying, selling, installing and operating systems in this vain. My employer has had enough and now wants to move over to robotics. He will support and help me if I ask him to, but in general this thing has been made my responsibility.

    Of course it would be better to hire a professional "robot guy" to help us, but we haven't found one yet, the country I live in doesn't have many robots in industry. What kind of qualifications am I looking for in such a "robot guy" anyway?

    I have spent a few hours online doing research and feel completely overwhelmed, not least because most my google searches have yielded nothing useful - I suspect I lack some basic terminology...

    Finding vendors who sell used and refurbished robots was easy enough (we are looking mostly for KUKA or ABB), but I cant seem to find anywhere that sells E.O.A.T or "turntables".

    Obviously, purchasing a robot is the least of our worries, we also need to learn how to operate it. We have some basic CAD skills and I have been given the time to learn how to use fusion 360 (I can do that, I'm good with learning how to use programs), but I don't even know what I don't know... any help or pointers will be greatly appreciated.

    So here's the basic gist:

    At the beginning we are looking to use the robots mostly for cutting sheet metal and plastics. To that end, we want to purchase a fiber laser cutter able to cut through metals of ~6mm thickness. We need an arm with a reach of ~2m and the precision of ~0.06mm. We are also looking for a spindle tool that will be used to cut plastics of similar dimensions. We are also exploring the idea of robot CNC and milling of foam sculptures.

    To make my question a little more specific, the first task given to the robot will probably be cutting sheet metal in 2D. For example, cutting a hoop of a certain dimension, then a long rectangle (which will later be bent around the hoop and welded, not necessarily by a robot), then a similar hoop out of plastic. Effectively making the parts for a lamp, to be assembled later by a technician. As our understanding of the robot grows, so the tasks given to the robot will become more complex. I understand the basic requirements I need to look for in the robot: load, reach, repeatability. What other requirements should I be looking for? What kind of skills do we need to acquire (by our selves, or by hiring people who have already acquired them) for a process such as mentioned above? What programs do we (or hired help) need to be familiar with? What do I not know that I do not know?

    I realize these are some very vague and general questions, any help will be greatly appreciated. I feel like I'm groping in the dark, I just need to be pointed in the right direction.

    Thank you,


  • Welcome to the forum!

    If you are looking for a Robot guy then it would be preferable to find someone with experience with your application. If you want to do machining, routing, or laser cutting then ask for someone who has done that before.

    Here are a few questions to ask yourself.

    How important is accuracy? You mentioned holding a precision of .06mm. A good industrial robot can be repeatable to .06 mm but will never hold accuracy to .06mm. For cutting plastic and sheet metal to a high accuracy then you would be much better off purchasing a CNC. There are also CNC laser machines you can purchase that would be much more accurate than a robot. Have you considered a CNC?

    What type of EOAT are you wanting, Drill, Spindle motor, gripper? There are components you can purchase but generally you have to design the mounting plates yourself that attach the motor or gripper to the end of the robot. Then you either need to machine the mounting plate yourself or hire a machine shop to do it for you.

    What about electrical controls? Do you have any experience with wiring, 3 phase power, VFD's, I/O? Someone will have to wire power to the robot, wire power to the spindle motor, setup the VFD and IO.

    What about Safety? Do you have any experience with industrial safety, guarding, E-stops?

  • Thank you for the response!

    The thing is, where I come from there aren't many robots. I'm assuming there are some in the heavy industries, but its not a common thing. This is why finding a guy is proving to be so difficult, the people who installed the existing bots were probably from abroad.. My employers long term end-game is to one day become the authority on industrial robots ourselves.

    I'm going to be honest, I kinda thought repeatability and accuracy were the same thing =/. Accuracy IS important to us, but I couldn't find an "accuracy spec" on any of the robots spec sheets, or am I missing something? I was looking for some sort of spec about the "steps" in the bots movements, like, whats the smallest movement the bot can make. Either way, a robot is more important to us than the highest accuracy. We rather get a relatively accurate robot than an extremely accurate CNC. I know the example of a task I gave is better suited for CNC, but thats just for the beginning while we learn. Eventually the tasks will be more complex and 3D.

    The E.O.A.T we are looking for are spindle, and fiber laser cutter. Hopefully getting an automatic tool changer later on. Fabrication is not a problem for us. We're pretty good at it ourselves, and if we feel we're not up for the task we have some good contacts for things like that. We have a lot of experience with wiring, 3 phase power isn't an issue. We don't have any experience with VFD's and if I'm completely honest, I don't even know what you mean by I/O? In/out? Input/output? How is this related?

    As for safety, theres no experience. Is there a standard I can look up and educate myself? I'm assuming E-stops are programmed stops in the bots movement that limit it from moving into a wall and things like that? That type of thing (and anything related to programming) is something I will teach myself.

    Can you recommend any resources for further information? By no means do I intend on being lazy and asking silly questions on forums to get other people to answer me. I Want to research things myself, its just that I can't seem to find alot of the information I am looking for and I don't completely know what it is that I need to research. I also understand this endeavor is too much for complete amateurs and that we will need some help from professionals, but I do want to be educated on the issue as much as possible first. I don't want to depend on others completely.

    Thank you so much for your help

  • Where is it that you are from? The term to search for is Robot Integrator, Systems Integrator, or Robot Programmer.

    If you do a search for precision vs accuracy and look at the bullseye pictures that will give you a good mental image of repeatability vs accuracy. Robot manufactures don't publish accuracy because robots aren't very accurate and it depends on many factors out of their control (such as payload). Generally you can expect an industrial robot to have accuracy around 1 mm. Some are better, some are worse.

    For EOAT you just need to select a reputable company that makes spindles and lasers, then fabricate a mounting bracket to attach them. Make sure that the mounting is accurately made and use dowel pins. VFD = Variable frequency drive. That will be needed to control RPM on a 3 phase spindle motor. I/O = inputs and outputs, generally 24 VDC signals to control other devices or receive input from other devices. For example 24V output could control a solenoid valve for a gripper to open or close. Or a 24V output could go to a VFD or motor starter to turn a motor on. 24V input could come back from the VFD to say that the motor is actually running.

    E-stop is an electrical safety device, red mushroom head button that stops the robot. If you have other devices like spindle motor or laser you would want to stop them also. Robot not running into wall could be done by safety rated software that limits position. You can review RIA for robotic safety standards.

    Honestly, there are not a lot of resources out there for industrial robotics. This forum is the best place to get information. The more you get into it the more specific you can make your questions and get good quality answers back on this forum. Each industrial robot manufacturer has a proprietary system and programming language so it will be important to have the manual when you purchase a robot.

    Good luck!

  • Repeatability vs Accuracy. The "target shot" graphic is particularly useful. Robots are good at repeating motions exactly the same way every time -- even if the motion is wrong, it will be wrong in exactly the wrong way time after time.

    Basically, if you hand-program the robot, and keep tweaking the programmed positions until the path is "perfect," that "perfection" will be repeated to within the repeatability limits of the robot forever (or until the robot begins physically wearing out). If you try to go CAD-to-Path, you're immediately in the "accuracy" enviromnent.

    Also, compared to CNC machines, robots have a lot more backlash. If you programmed a mathematically perfect circular cut, it'll turn out to be slightly oval. This can be corrected, but requires trial-and-error.

    Robots also have much less rigidity than CNC machines. A robot with a cutting laser will be much more accurate and repeatable than the same robot using a spindle, b/c the spindle will induce "recoil" forces that are not very consistent.

    I have one of these self-correcting router tools, and one of my ambitions is to find a way to strap it to a robot to do large-scale CNC-ish operations without struggling with the robot accuracy issues. But it'll probably remain a dream forever...

  • HawkME and SkyeFire , I would like to bring a bit of nuance on the Repeatability vs Accuracy thing. And I invite you to correct me if you think I'm wrong.

    Everything that was said is right, but in the case a 2D cutting on a sheet of material, the precision of the part create should be quite good if it's not too big of a part. The absolute precision with respect to the base of the robot won't be good, but the distance between the points should be quite good. In other words, as long as the "XY" plane is define properly and the part is not huge, the cutting path could be shifted 1 mm on the sheet itself, but the part should be quite precise. In a case like that, we are kind of in the middle of the realm of repeatability and precision. Especially in the case of a 2D model where you don't switch for one configuration to the other on the same part.

    This is just my personal experience working with offline programming software as we constantly struggle with robot accuracy, we even developed a calibration method using a laser tracker to improve robot accuracy. But if you are working in a quite small work area of an existing piece or if you are cutting on a plain sheet, a good frame definition should do the trick. Especially if the goal is to cut parts to build a lamp.

    dodkipod First of all, welcome to the world of robotics. It's such an amazing field where you will learn new things every single day. This forum is, in my opinion, the best resource on the web in the field of robotics. There's a lot of very knowledgeable and active members.

    I don't want to look like the guy that push his product (I'm not a sales man, I don't get any commission or anything) but a good way of getting your feet wet without breaking the bank and/of breaking anything physically speaking would be to use a simulation software to get started. There a few software that gives you free trial. You can create your station, create the program, run it in the simulation environment and generate the real robot program at the end.
    Some robot manufacturer have their own specific simulator:
    ABB -> RobotStudio
    Fanuc -> Roboguide
    But some are also robot agnostic:
    RoboDK, RobotMaster, Octopus, etc.
    This is exactly how I started robotics about 10 years ago. I downloaded a trial version and learned a lot (I also did a whole lot of virtual robot crashes! Good thing I was not playing with the real robot!)

    RoboDK even have a direct link with Fusion 360, you can draw you laser cutting path in Fusion and automatically transfer it to RoboDK, it will then convert it to a usable robot program in the language required by your robot controller.
    Most of the robot simulators will have some kind of Wizard to help you create complex program without requiring the complete robot programming knowledge of an expert.

    Have a great day and good luck.

    RoboDK - Simulation and Offline programming software for industrial and collaborative robots.

    Visit us at
    Take a look at our tutorial videos on our YouTube channel.

  • Pay very close attention to safety, especially if you are putting a laser into the equation. There are ISO standards for robot safety, and RIA (Robotic Industries Association) standards as well. Of course, to obtain copies, they will want you to pay for them. Welcome to the wonderful world of robot safety. If you want to buy a new robot, many if not most of those manufacturers are also integrators. You might expect to pay a premium for their services, though. But they should know their own robots and do a good job. Spend the time to clearly define what you will need very specifically from your proposed system. It is very difficult to give a customer what they want when they really do not know what they want. Or if those wants are often changing. Oh, and did i mention SAFETY?

  • Jeremy you are correct on the nuance of accuracy. It really depends what they are trying to do, small area vs large area, and if dealing with cutting forces from a bit vs laser. The laser tracker calibration makes a huge improvement in accuracy but definitely adds cost.

    It will still not be as rigid and have more backlash than a CNC. All depends on how good is good enough.

  • I forgot to mention that some robot brands have enhanced or "absolute accuracy" options. ABB calls it absolute accuracy. It is a special calibration method for the robot which takes into account things like deflection of the arm and such. ABB also has "advanced shape tuning" for things like cutting holes and what not. I am sure that some of the others have their equivalent.

  • HawkME Thanks for the confirmation.
    And just to clarify, I was not suggesting that a laser tracker system was the solution for this process, just that precision is such an issue that, for us, it was worth developing a solution for it.
    But so that you know, the laser tracker is only needed for the calibration process, therefore we have partners that can come on site with their laser tracker for a few thousands dollars. It can generally be done in a day or less (smaller the robot, faster the process). You end up with a calibrated model of the robot (kinematic including flex under payload. * Not under applied force). You can use it to created filtered programs for the robot using our software.
    This way you don't need to buy the expensive toy that is a laser tracker.

    RoboDK - Simulation and Offline programming software for industrial and collaborative robots.

    Visit us at
    Take a look at our tutorial videos on our YouTube channel.

  • Thank you all for the great responses.

    One question about accuracy though, granted, we don't need anything super accurate and I'm sure we will be fine. But out of curiosity, if bots accuracy is such a thing, how is it that they can create such impressive sculptures such as these (image from a google search): digital-stone-carving-robot.jpg

    Ive seen many videos on instagram of robots milling/sculpting in foam with a large spindle on a large robot. From reading the responses in this thread I would assume these should be very inaccurate, yet the end result looks great.

  • And/or additional 3rd-party calibrations, like SARCA , or a lot of trial-end error fine-tuning on their CAM programs. Or just raw experience on the part of the CAM programmers -- there's various tricks you can use to minimize the effects of robot inaccuracy over small areas, and with experience you can figure out how to "hide" the worst effects in areas that aren't very obvious, or are easy to post-process "blend in" with some sandpaper.

  • Ok so I think I understand. If my robots repeatability is +/- 0.06mm then it can perform the same task, each iteration being within 0.06mm to the last one. If the robot is inaccurate, the path it makes may differ from the path I asked it to make, but in every iteration, it will make the "same" mistake right? So basically, with low accuracy, mistakes will be made. However, the "same" mistake will be made every time, within 0.06mm.

    In order to improve accuracy, I can run some tests, sending the robot tooltip to a certain point and seeing how close it gets, thereby learning the range of error and then incorporate the error into the path I program. So say I set it to move to x/y/z and I see that it ends up in x-1/y+2/z-2, I can take these things into account when programming, making the end result more accurate (within 0.06mm every iteration). Did I get that right?

    Can you guys give me a gross estimate of the range of inaccuracy? We're not going to do any precision milling, its (at least at this moment) looking like we will be using the robot for foam/wood milling, making sculptures and designs for show-business, hotels and the likes. Were not looking for pin point precision.

    Edit: bad manners asking questions before researching, I know, I do apologize. From what I can tell the accuracy is something like +/- 1mm, does that sound about right? How much worse does this get when milling harder materials such as wood?

    they use positionally accurate robots (HA or ABS variants)

    Can you elaborate on this a tiny bit please? All the milling videos I have seen use "regular" robots, for example KUKA kr180, but when I search for positionally accurate robots I do find specific models.

    Edited once, last by dodkipod ().

  • Robots are made from parts. and all those parts have tolerances (including length of joints!). As long as tolerances are within some limits, those are all good parts and can still be used to make robots. But this still means that every robot made will be slightly different from each other and therefore not a perfect match for model used in software (all robot of one type use same software model, there is no individual tweaking).

    So what was just described are so called standard robots (STD). Making tighter tolerances would work better for accuracy but it would mean way more rejected robot parts.

    In case of KUKA, robots, two "fancy grade" options are available: HA and ABS.

    HA robots are like STD but made from carefully handpicked parts (every part is tested and every now and then you get one that matches design perfectly or at least a lot closer than average part).

    So right from the start HA robots (high accuracy) are more precise since they match model much better than average robot units (STD).

    This is why if you want HA version, you need to order it as such. To be an HA version, arm must be MADE from those perfect parts which can only be done in the factory. Plus it is not like those fancy grade parts are just sitting around, it takes some time to gather all perfect parts (must run some volume production). in other words if you want HA version of some rare model, it may tage some extra time....

    But that was only the first step. The next step is to better measurement of the arm under load to make unique "fingerprint" compensation.

    All this means HA robots are like STD but more accurate and a bit more expensive and a bit longer lead time. HA is the top notch, there is nothing better in commercial, volume robots.

    Then there are ABS robots. They are basically STD robots in terms of parts tolerances. And while parts are same as standard (and therefore have same tolerances), these robots do get measured to get that compensation "fingerprint" (known as PID file), just like HA models. They get pretty close to HA but without special/rare parts. And unlike HA units, ABS upgrade can be also done on site - just need to measure robot, not rebuild it with rare parts.

    1) read pinned topic: READ FIRST...

    2) if you have an issue with robot, post question in the correct forum section... do NOT contact me directly

    3) read 1 and 2

  • Fanuc also offers a high accuracy "Signature" calibration.

    When routing wood the robots accuracy doesn't necessarily change but you get effects from vibration, chatter, backlash,cutting forces that can affect the cut quality.

    If you can afford higher accuracy then I strongly recommend it. Also buy good quality motors and cutting tools that are specifically made for the material you are cutting.

    Wood can be done with good results but requires effort and some trial and error.

  • Any machining with a robot on any material harder than styrofoam should probably be assumed to need some "spring passes".

    There's no hard number for accuracy, although you can generally get the data for a particular robot model from the manufacturer. But that will generally have caveats for payload, inertia, and reach distance, among other things.

    In my experience, a 10:1 ratio between accuracy and repeatbility has been a decent general rule of thumb. This assumes no steps have been taken to improve accuracy.

  • Robots are made from parts..thise parts have tolerances. As long as tolerances are within some limits, those are good parts and can be used to make robots. But this means that every robot made will be slightly different and not perfect match for model used in software...

    This was extremely helpful, thank you so much

    I'm afraid we can't afford it, but it seems to me like its not completely necessary.

    In my experience, a 10:1 ratio between accuracy and repeatbility has been a decent general rule of thumb. This assumes no steps have been taken to improve accuracy.

    So as a general rule of thumb, if I get a robot with +/- 0.06mm repeatability, its accuracy would be roughly +/- 0.6mm? This still seems pretty good to me

  • Hi,

    that is not a problem you don't have robotic knowledge . There is a big evolution in robotic world to permit anyone to use a robot.

    This may imply avoiding complex robots and have a look at robots solution like universal robots, where programming is very easy (you can teach using it on their website) . But that really depends on your needs.

    From my opinion, RoboDk is the tool you need for your application, you can first evaluate your project with it and then buy required hardware.


Advertising from our partners