Home Built Delta Robot


Theres an interesting thread going on at RoboterNetz.de (translated) about building delta robots. For those who haven’t seen delta robots before, be sure to check the video after the break for some really impressive agility displays. A delta robot usually has 3 arms connected to a single point at the end. This configuration allows the end point to maintain its “level” while the whole unit is moved, usually very quickly. There are a few people building them in this forum. The one pictured above is [Asurofilmchen]’s version, but you should also check out [Radbruch]’s posts of a robot made by [Jamison].


[via Hacked Gadgets]

39 thoughts on “Home Built Delta Robot

  1. @therian

    It’s not that I doubt human sexual ingenuity, but how do you mean? the side to side motion of the movement would be pretty rough on the inside of the vagina or anus, a back and forth motion would be much more usable.

  2. I wonder if a delta robot couldn’t be useful as a milling machine? Mount a dremel or some such on the head, and you have all three axes. I don’t know how much harder these would be to drive accurately, though, as I have not attempted either.

    In addition, IIRC, you can use this system to ‘record’ a movement in three dimensions, and then play it back. You could draw a shape with the head and replay it with a router bit or some other tool.

    1. I am planning on building a cnc router and people seem to want more lighter passes, and I program delta robot at work. I don’t know how to compare without building.

  3. Indeed. They have a version of a Delta Robot that is on the shelf at Fry’s around here, which is being marketed as a three-dimensional force-feedback input device. I suppose each axis can both determine its position (input) and change it as well (force feedback)

    Ever since they started calling vibrator motors “force feedback” i’ve been put off of the whole idea. I want the steering wheel to jerk out of my hands, dammit, not make my lady-friend squeal.

  4. That’s great. CNC with no linear bearings. I’d like to have a go at making one of these and putting a router on the end of it. Now, are those rose joints all round, or just at the ‘head’ end? Is there only one degree of freedom at the elbows? I’m thinking three windshield wiper motors and three optical feedback disks could drive this.

    I wonder how hard it would be to set up EMC to control it.

  5. I can imagine that any play in the joints would futz things up. Even so, a well made version seems like it could handle it. Mind you, I have no practical experience with milling, though I’d love to get into it. Living check to check with a family kinda puts a damper on hobbies.

  6. defex got it right: these machines are REALLY good for pick and place, since the items are usually of very light weight. However, bad for CNC as they are not very ridged, with little holding strength. To make matters worse, unlike simple X/Y machines, its strength is variable throughout its operating envelope.
    Accuracy for light loads (low mass) can be excellent (with the right actuators, sensors, and software).
    Difficult for ‘profiling’ (as in CNC path following) as the math involved to transcribe a simple XY motion to the three motor actuator is pretty complex algebra (possibly some calc)

    And, in the post, you don’t really mean ‘level.’
    In the case of the video shown, level is the vertical displacement of the effector within the operational envelope.
    you mean that the ‘orientation’ of the end effector always stays consistent. In fact, without some really difficult math, the ‘level’ of the end effector is never consistent… more like sine waves (it bobs high near the middle, and low near the sides of its envelope. in order to make it move a straight line, you would need to be able to calculate and output 3 dimensional sine wave profiles to motors. Tracing complex curves becomes even more difficult.

    pick and place works great though, since you dont care about the path, only the pickup point and the drop off point. (so long as you specify a mid point that is shallow enough such that you don’t crash into the work surface as you pass through the middle of the work envelope.

  7. PNP(pick’n’place)+CNC+3dprinter= replicator >:) with that built in to a cabinet… now if some one could do a DIY carousel then the problem of raw material would be solved

  8. @dd, @ast stewart platforms are a little different, they use 6 actuators to move in 6 degrees of freedom (x,y,z,roll,pitch,yaw) You can use linear actuators or servos (mine has servos because it’s cheap, but it makes the math very difficult) The deltas have only 3 DoF (x,y,z)

    Both are very cool, keep up the good work builders!

  9. water jet for stone cutting comes to mind.
    dunno if the force of the jet would throw off accuracy too much. if the frame was manufactured rigid enough and you had a goo flexible high pressure line this could work out making marble or granite busts. possibly crystal skulls..

  10. Problem with these for any sort of decent machining is lack of operating envelope. Good for fast, lightweight operations, but not with any force behind them or moving over large distances.

  11. @nuke:

    The video is legit.

    I was at a CNC convention last year in Los Angeles and watched this robot do it’s thing. These are very fast and very accurate.

    I’m not sure how many of these there are, but it looked exactly like this and was in a booth for a company that makes PLC’s and servo controllers.

  12. My understanding is that the primary goal of the delta arm robot concept was to increase speed. Toward that goal, the end-effector’s weight was reduced as much as possible by moving all the actuators to the stationary base. The flat orientation of the end-effector is beneficial, but I don’t think it is the main benefit. The ultimate speed and translation agility were the big boons.

    So I think putting a milling head on it, even a Dremel, would not be in line with the why you would choose a delta arm.

  13. This could be good for sheet cutting, laser, gas plasma or water where there are no side forces from the cutting process. The only difficulty i could see is (as mentioned above) working out the maths of the trajectories of the head. (keeping a level path parallel to the work piece)

  14. In my opinion, although the general design serves it’s purpose It would like to see more effort given to the programing of robotics, as that is where the greatest amount of time is lost. They ride so much on being exact with their movements, but are absolutely horrible at doing things with less precision. That is why machines with high tolerances, and high price tags prevail, that IS their advantage. We on the other hand are really very good at being inaccurate, and can adjust the trajectory of our limbs mid path at very high speed in order to achieve our goal. that is why humans are so much faster when accuracy isn’t the goal.


  15. I’m glad you all like my high speed Delta robot.

    I was wondering why all the hits on youtube, now I know!

    Some very astute observations made in this thread.
    Andar_B- These are on the edge of material strength, best suited for very light pick and place… you very quickly get into the snafu of needing to make it stronger, thereby getting heavier, thereby needing to make it stronger.. and round and round into Newton’s Law hell.

    Nes- It is triangulated and strong, but any cantilevered load (router bit) extended beyond the very center of the end effector gets into some wild dynamics, including monster torsion on the upper arm, and depending on where it is in the workspace, monster axial loads in the parallelogram. Not optimal for cnc.

    MRE- Great observations, though kinematics for this were simple trig, trajectory generation (path-planning) can be standard trapezoidal method, or G-code, but JensOverby (check his youtube channel) has a dynamics based trajectory planner that is even over the head of the ABB flexpicker engineers!

    kgartner- uh-oh, a serious PKM guy. My friend Ilian Bonev has just finished the first textbook for these mechanisms “Fundamentals of PKM’s”. It is good brain food for us.

    Dan and Nuke- I’ve made @10 different versions of this mechanism for Delta Tau, they always show them at the conventions. Some slower, some faster, some bigger, but all with different controls/amps/motors.

    KermyNZ- Industry is researching this, but the only reason industry is looking at these for laser/waterjet/plasma is due to cost (no bearings). Honestly, anything away from fast/light/faster/lighter begins to move away from the true strengths of the mechanism.

    DY- exactly

    In a nutshell, 10-15g’s of acceleration is the key to these, with cost and workspace close behind. No where else will you get 160 pick and places per minute over a 1m workspace for @$15k cost. Period.

    Glad you all liked, this is a growing field, very cool to be involved.


    1. At work I use adept quattros with a terrifyingly long gripper post (600mm) and loads of 2.5kg , still making 60+ pick and places with 1m travel. The cnc wood router builds I have been looking at are slower and less force and have only slightly smaller motors. I have seen a bunch of pick and place ones with an extending shaft for rotation, then you are only moving your cutter and side load again. Also in my case mounting the unit to the ceiling and software levaling and a movable table, then the unit is completely out of the way when off.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.