Potential impulse-purchase alert!

Failure is not an option...it's bundled with the software.

Comments

  • That looks pretty cool. Too bad the video of it shows the backlash to be 400 times what they claim the resolution to be.

    The guy measures the play (not really backlash) at about 2mm, when it should be ~ 0.005mm.

    Also, in other videos of it in motion you can actually see the play. Not sure why, the actuators appear to be better than that...
  • Ah, yeah but I believe that those are the low-end units. I think the robot is only intended for training.

    480 x 270 - 43K
    Failure is not an option...it's bundled with the software.
  • W9GFOW9GFO Posts: 3,915
    edited 2019-11-30 - 01:31:09
    They claim that the actuators have 19 bits of resolution per rotation. That is 524,288 values; though they specify 589,824 steps for some reason.

    Half a million discreet positions spread over a circle with a 400mm radius is about 0.005mm per step. That's a couple orders of magnitude more precise than demonstrated.

    They show it threading a needle, but it is not very convincing since it approaches only from one direction. I'd bet that @erco could hot glue a couple cheap servos to some popsicle sticks and achieve the same.

    Anyway, there is a rather large discrepancy between the claimed resolution and the demonstrated resolution. If that discrepancy were substantially less, I would be making an argument to my accountant in favor of backing the project.

    Another thing, they have an artificially low goal of only $10,000, which they are well past. They don't need my money to move the project forward. They say they already have a factory nearly completed. I can wait until some unbiased reviews are available. I do hope that they deliver what they claim, if they do I will certainly be a customer.
  • W9GFO wrote: »
    I'd bet that erco could hot glue a couple cheap servos to some popsicle sticks and achieve the same...

    Well... @W9GFO that comment sure produced a rather epic “diet-coke-thru-the-nose” moment here. :)
  • JRoark wrote: »
    Well... @W9GFO that comment sure produced a rather epic “diet-coke-thru-the-nose” moment here. :)
    Good to see that the humor came through, although I do seriously believe he could do it. Actually I am sure that many people could, he is just one that is more likely to pull it off using said materials.
  • W9GFO wrote: »
    I'd bet that @erco could hot glue a couple cheap servos to some popsicle sticks and achieve the same.

    @W9GFO - I nominate this for quote of the decade!

    Whit+

    "We keep moving forward, opening new doors, and doing new things, because we're curious and curiosity keeps leading us down new paths." - Walt Disney
  • I agree with Rich, the claimed specs really don't match what they showed.
    It does look neat, but I am in a wait and see mode on it too.

    Also, if erco doesn't post a video of said servos & popsicle sticks threading a needle in the next week, I will be sad. :)
  • Ah, my pet peeve....

    [Client]: "I don't get it, I'm churning out nothing but scrap but here, look at the computer screen....it shows that the axes are dead nuts accurate"

    The feedback device (encoder, resolver) is mounted, rigidly, on the back end of the motor shaft. There is a fault somewhere in the transmission such as a busted coupling or a pinion with a broken tooth, key-way, etc. As far as the control is concerned, all is good.

    Any machine of my design that requires a gear reducer, gets "dual-loop" feedback; motor feedback for loop stability and load feedback for position accuracy & repeatability. Even with high-end "zero-backlash" reducers.

    The robot in the vid has their low-end gear reducer...made from composite materials, I think they said.

    Failure is not an option...it's bundled with the software.
  • Mickster wrote: »
    The robot in the vid has their low-end gear reducer...made from composite materials, I think they said.

    They have some weird language in the promo video. Something like "self developed composite..." which I am sure refers to the carbon fiber connecting structures and maybe the actuator casing.

    Nothing I have seen leads me to believe that there are any different versions of the actuators that they will offer. The whole product revolves around the low cost actuator.

    I looked again and did not see any mention of any other actuators that they are using or intend to use. If there are different versions of the actuator, can you point them out to me?
  • INNFOS SCAs include QDD Pro series (with harmonic reducer), QDD series (with planetary reducer), QDD Lite series (Planetary Reducer based on composite materials), and DD series (no reducer)

    Source: Robotics Tomorrow


    And then there is their website
    3728 x 1707 - 160K
    Failure is not an option...it's bundled with the software.
  • MicksterMickster Posts: 1,525
    edited 2019-11-30 - 11:25:27
    Regular CNC application


    Failure is not an option...it's bundled with the software.
  • My point is that ALL the information regarding the kickstarter product is based upon the QDD Lite actuators, and it is the specs for those "Lite" actuators that claim 19 bit resolution. The fact that the same company offers other actuators is irrelevant because none of those other actuators are mentioned in the kickstarter.
  • The point is that; this desktop robot is the closest thing that I have ever seen to a real industrial robot.

    Sure there are many articulated arms out there but using stepper motors.

    If one intends to teach/learn robotics.....why not learn about what one would find in the real world. True closed-loop servo control with torque limiting and compliance.

    IIRC, you can get the entire thing for the cost of a single harmonic/integrated motor.
    Failure is not an option...it's bundled with the software.
  • Mickster wrote: »
    The point is that; this desktop robot is the closest thing that I have ever seen to a real industrial robot...

    I thought you were saying that the reason for the slop was that they were using lower end actuators. My point was that the ones used were the same ones they claim have 19 bit resolution (0.005mm when the arm is extended). There seems to be a huge discrepancy between what they are claiming and what has been demonstrated.
    Mickster wrote: »
    If one intends to teach/learn robotics.....why not learn about what one would find in the real world. True closed-loop servo control with torque limiting and compliance.

    I think you can get those same things using the Dynamixel servos. These actuators seem to be a lot better and are much more similar to industrial robots, but a resolution of +/- 1mm with the arm extended is pretty disappointing. That is no better than a servo based robot arm.

    Why do you think they are claiming that 19 bit resolution has 589 thousand steps when 19 bits is actually only 524 thousand? That's kinda weird to me.



  • All I get from the video is "resolution of 0.00086"

    The encoder is 14bit = 16383

    360/16383/0.00086 suggests that the reducer ratio is 25.55:1

    Looking at the part numbers, the nearest that I see to 25 is "30" (don't know if the part number is supposed to indicate the ratio)

    Resolution is not the same as accuracy/repeatability. In this case it is the smallest possible increment of the axis, based on the above.

    A 6 DOF articulated arm, remember. That's a lot of cumulative error to consider. Were the arm segments aluminum, I'd wager that one would see way more variation with temperature fluctuations (in a previous life, I manufactured 5-axis articulated-arm CMMs).

    Dynamixel cannot be compared. Conventional motor with conventional reducer and a rudimentary integrated drive. It could never do compliance without host intervention. All one can do is limit the torque but the axis will still want to return to the last-commanded position. The SCAs, in compliant mode....one can push them and they will assume that position as the new command position.

    I am convinced that there is some adaptive torque limiting going on there which just inspired me :smiley:

    Also, a 6-pack of the relatively clunky Dynamixel drives costs more than two of the Gruon robot arms, currently offered on KS.

    As stated, many times; the Gruon is intended for education and research.
    Getting hung up with end-effector repeatability is totally missing the point. Getting away from rickety, dumb, open loop steppers and getting students involved in proper axis control is what is on offer here.
    Failure is not an option...it's bundled with the software.
  • About the the seeming discrepency between play and resolution:

    Could it be that the high resolution is still somewhat useful in a high-play gear train because, for a particular static situation, all play/backlash will be physically taken up in a predictable manner, and therefore all that resolution is still potentially meaningful?



    I am the Master, and technology my slave.
  • MicksterMickster Posts: 1,525
    edited 2019-12-01 - 13:29:09
    The_Master wrote: »
    About the the seeming discrepency between play and resolution:

    Could it be that the high resolution is still somewhat useful in a high-play gear train because, for a particular static situation, all play/backlash will be physically taken up in a predictable manner, and therefore all that resolution is still potentially meaningful?

    I was going to state this but you worded it better.

    I am currently on-site, retrofitting a piece of metal-forming equipment. There is a linear (horizontal) axis; servo-motor with integrated encoder, reducer and rack & pinion final drive.
    With power off, I rock the carriage back and forth by hand and can feel/hear/measure ~3mm lash. My resolution is 1mm/300 (300 encoder counts/mm).

    The nominal tolerance of this axis is +/- 0.1mm.

    I am easily achieving half of this tolerance because any external influence on the work-piece can only come from one direction and my pinion tooth is already pushing against the rack tooth...the lash is already behind.

    The maintenance guy was horrified at the thought that he would have to replace the entire transmission...nope, it still looks great and the lash is compensated for.

    Same thing applies to robot welding; the arm follows a consistent path. If any lash means that the tip coordinate is not absolutely true, who cares as long as the offset remains consistent. Tweak the program.

    Failure is not an option...it's bundled with the software.
  • ercoerco Posts: 19,560
    Roy Eltham wrote: »
    Also, if erco doesn't post a video of said servos & popsicle sticks threading a needle in the next week, I will be sad. :)

    Aw thanks Rich, Roy, Whit et al! I'll add it to my "to do" list. Kinda buried right now, but it does sound like a great challenge!

    Perhaps we should start a whole new "thread" akin to the long-lived Figure 8 challenge. :)

    MIAs Martin_H & Tommy Tailspin can jump in anytime. Feel free to do a figure 8 then thread the needle afterward.
    "When you make a thing, a thing that is new, it is so complicated making it that it is bound to be ugly. But those that make it after you, they don’t have to worry about making it. And they can make it pretty, and so everybody can like it when others make it after you."

    - Pablo Picasso
  • W9GFOW9GFO Posts: 3,915
    edited 2019-12-01 - 18:13:32
    Mickster wrote: »
    A 6 DOF articulated arm, remember. That's a lot of cumulative error to consider...

    In the case of the 2mm of play measured at the end effector, all of that play is in the base actuator. Have you seen this video?

    https://youtu.be/ZlJENPxR7yM?t=613
  • [Groan] I imagine this clown went straight back to his Xbox.

    How many servo parameters did we get to see? None.

    Anyone concerned about accuracy would have a look to see if any integral (Ki) was applied. Personally, I only apply it when I really need to hit a target and set it to zero as soon as it isn't required.

    When sitting idle, I also sometimes drop the position loop gain (Kp).

    Deadband is another possibility....Why have the drive PWM-ing to stand still.

    If the axis is within +/- deadband_counts then kill the PID loop, else enable PID loop. This will feel exactly like backlash.

    They need to have a servo/robotics expert review that product.
    Failure is not an option...it's bundled with the software.
  • Mickster wrote: »
    [Groan] I imagine this clown went straight back to his Xbox.

    I would like to think that it is a case of user error and not the limits of the actuator. Unfortunately, none of the other videos I have seen have convinced me otherwise.

    If it were me, and I was trying to sell this robot arm - I would absolutely be showing off it's accuracy and repeatability. What concerns me is the lack of such demonstrations. Is it simply an oversight, or is it because it is not up to the task? On the other hand, such demonstrations are clearly not required since a whole lot of people have committed to buy them already.
    Mickster wrote: »
    They need to have a servo/robotics expert review that product.
    Absolutely, as I said in my second post "I can wait until some unbiased reviews are available. I do hope that they deliver what they claim, if they do I will certainly be a customer."
  • Duane DegnDuane Degn Posts: 10,007
    edited 2019-12-04 - 22:00:45
    FWIW, Skyentific posted a video about the actuators.

    https://youtube.com/watch?v=0sgR_RaxYu4

    While I agree the "reviews" from this channel are lacking in detail, I still like Skyentific. I get the impression he's not pretending to more than he is.

    In order to hold an "early bird" price, I submitted a pledge for their 6-axis arm. I have until Christmas to cancel it (which is my current plan).

    I keep imagining a hexapod with 18 of those actuators.
    Hobby servos work well enough to make a hexapod walk but hobby servo pots wear out with extended use. A hexapod based on 18 of those actuators should be a lot of fun. Though I doubt they'd be as much fun as other ways of spending that kind of money.

    Now I'm torn between cancelling the pledge for the 6-axis arm and adding 12 actuators to my pledge.

    As I write this, there are still 47 of the 6-axis early bird options available. I'd be more likely to follow through with purchase if other forumistas were joining in the fun.

    Edit: Trying to get YouTube video to link.
  • Yeah, he still has that backlash stated from when he performed that measurement but I am not convinced that he was seeing backlash. I imagine the arm being shipped with soft servo gains because having them too high would result in the thing vibrating all over the place.

    He also states that the actual reducer ratio is 26:1 which is close enough to the 25.5:1 that I came up with.

    I would only be interested in one of these for demonstrating concepts (machine loading/unloading). I have no intention of putting it to work. It wouldn't take much to scale-up and use their "pro" range of reducers.

    I was tempted to go for the 2-pack offer but I know that I would probably not even get around to unboxing....fully booked well into 2020 and 7 days/week.
    Failure is not an option...it's bundled with the software.
  • Duane Degn wrote: »
    FWIW, Skyentific posted a video about the actuators.

    https://youtube.com/watch?v=0sgR_RaxYu4

    He says the backlash is 18 arc minutes (nearly a third of one degree), which corresponds nicely to the +/- 1mm movement that he showed in the other video.

    The encoder must be reading the motor instead of the output shaft, unlike servos such as the Dynamixels which have the encoder reading the output shaft. That would explain the backlash.
  • Failure is not an option...it's bundled with the software.
Sign In or Register to comment.