Open Source Humanoid Robot Is Awesom-o

awesome-o

Coming from a lab in France is The Poppy Project, an open source humanoid robot that’s at least as cool as ASIMO.

Poppy was designed as an affordable bipedal robot for use in education and art. It’s a small robot at just over 80 cm in height, but it can walk, move its arms, rotate its torso, and interact with bags-of-meat humans with two cameras and an LCD face.

Although Poppy is open source, that doesn’t mean it’s exactly cheap; the current design includes twenty-one Robotis Dynamixels MX-28 robotic actuators, actually servos with magnetic encoders, temperature sensor, and an ARM microcontroller. These actuators sell for about $200, meaning Poppy contains $4000 in motors alone. The estimated cost of the entire robot is €7500-8000, or about $10,000 to $11,000 USD.

Still, there’s an incredible software platform that comes along with Poppy, and being open source any enterprising engineer can take up the project and attempt to bring the costs down. We’d love to take one out for a walk. Just get rid of the hands. That’s too far down the uncanny valley for us. Video below.

 

34 thoughts on “Open Source Humanoid Robot Is Awesom-o

      1. Robotis servomotors include a torque measurement, configurable PID, full digital control through a bus (depending on the model) and mainly, the “28” in the name refers to the available torque…. Your micro-servo is far far away from such power.

        Please don’t compare diamond with coal

        1. If you measure current drawn by the servo, you can calculate the torque. With some calibration, Torque can be found. Measuring current does not required you to modify the servo.

          The pot’s signal can also be tapped inside the servo to figure out the position.

          T = kT * Ia * Ø where Ia = armature current, kT = torque equation constant, Ø = machine’s total flux

    1. The MX24s have magnetic encoders with 4096 positions over a 358 deg range. The magnetic encoder has the advantage of being contactless, which means they won’t wear out like cheaper pot-based servos will.

      Although maybe I miss the point of your post :)

      1. Still, must be possible to get the same functionality, or hack something up that works, for a lot less money. Maybe use 3 Hall sensors and a magnet for triangulation, if it has to be contactless, and if reflecting light off something isn’t possible. Do the maths on whichever processor’s replacing their onboard ARMs.

        Seriously though, a servo motor with a 32-bit processor… WHY!?!?

    1. You can get some ideas what the RC servos are doing by monitoring their currents individually. If someone really want to get full control, He/she can gut a RC servo and control the motor/gear/feedback with a microcontroller.

  1. No mention of the South Park reference? Anyway, I see people mention basic servos but I think they would be too bulky to match the look they wanted. Aesthetics seem to be a factor in this project.

  2. Ways to start the video, in descending order of coolness:

    1) Poppy walks into the camera frame itself.
    2) The camera fades in on Poppy.
    3) Poppy gets dragged in and set down unceremoniously.

    Going with #3 wasn’t exactly a good first impression.

  3. At least as cool as Asimo? You kidding, right?
    This one doesn’t have one hundredth of Asimo’s capabilities nor its motion abilities (walking, climbing stairs, _running_, recognizing objects and people etc, all of this 100% untethered).
    Yes, it is truly interesting and kudos to the creators for releasing it as open source, but that description is misleading and would not help the Open Source philosophy by creating false expectations.

    1. Still, wheel-reinventing seems to be a big problem with robots. Having a few standards, so people can just concentrate on the parts they’re good at, will really help. I think it would need quite a bit of software development to work, not much use just defining motors.

    2. Asimo does nearly anything of what you mention. This is only an automated puppet with pre-programmed scripts. Don’t believe the marketing presentations of Honda and watch the video where it falls from it’s calibrated stairs…

      Aldebaran Robotics, Cybedroid, Boston Dynamics have much more evolved robots than Honda

  4. We are on the alpha testers list for this at the moment. It’s an exciting robot due to the cost for something so tall. We’ll be testing it in August after we go to robocup with the current Darwin-OP’s. We did note that it can’t walk yet, but that is a matter of creating a walk engine with good enough balance – it looks like it has plenty of torque to walk under its own power. Our build won’t be tethered either.

    As for the dynamixel/servo tradeoff: you can almost get there with a high torque (28kg.cm) servo and a feedback hack. Dynamixel has many more options though. You’d need to use openservo boards to get closer, and I still think the quality of Dynamixel would be better. If I was building at home, I’d go with the 21 century robot posted here previously as it’s a more stable design, and use feedback modified servos.

  5. Sparkfun had a few openservo boards left – you need to fix the fuse settings as they made them all with the wrong ones set. They work great. Alas RobotFuzz has been out of them for some time. http://www.robotfuzz.com/index.php?main_page=product_info&cPath=65&products_id=180

    Then use Servo Cities enhancements to increase torque and strength. and you get to use cheaper servos that have been turned digital. Or just use any servo in these products.
    https://www.servocity.com/html/actoboticstm.html

    I hope Sparkfun make some more openservo boards with proper fuses.
    Or you can follow along to a new generation V4 here (one day…):
    http://www.openservo.com/Forums/viewtopic.php?t=925&postdays=0&postorder=asc&start=180

    Don’t forget Adafruit has a servo mod to get that feedback pot back out to your own micro.
    http://learn.adafruit.com/analog-feedback-servos/about-servos-and-feedback

    Please note. I am not dissing the Robotis servomotors, they look great, but they are too expensive for me. I offer info for an alternaitive route.

  6. Looks great but I don’t agree with this:

    Affordable
    The overall materials needed to build your own Poppy robot costs around 7500€ (including motors, electronics and 3D printed parts).

    I’d go with relatively less expensive than existing models, but 7500€ is not affordable.

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.