Loading ...
Sorry, an error occurred while loading the content.

42173Re: [SeattleRobotics] Looking to share an order of Polymorph (Polycaprolactone/ShapeLock/themo set plastic/low melting-temperature polymer)

Expand Messages
  • David Buckley
    Mar 1, 2010
    • 0 Attachment
      Hi Peter
      I agree with everything you said although the Amblers and TecFoots are all under development - still. Just bought some mid range Sharp IR rangers to put on TF-Condor, one to add to the object detection and more importanly so it doesn't walk off a table when I forget and turn round. Fortunately because of the 4mm Birch Ply repair was easy. Although the 'Ironware' is more or less fixed the software and hardware are still changing.
       
      For the record Simplhex's legs were aluminium (hand filed) with a joggle in them to reduce the axial torque loads on the servos, would I do it that way again - no, plywood is so much easier even though for me they took very little time..
       
      The standard Hexapod design and in fact the standard robot design of attaching the limb to the servo is reaching a dead end even though that is the way Industry does it. Passive compliance will make everything work better providing standard IK techniques are abandoned.
       
      I do have a tub of Polymorph and some sheets of similar material used in ct-scan support courtesy of Roger and have been thinking it would make good dragon(dinosaur)  feet which are passively compliant. Could probably make pockets for QTpills!
       
      David
       
       
       
       
      ----- Original Message -----
      Sent: Monday, March 01, 2010 1:15 PM
      Subject: Re: [SeattleRobotics] Looking to share an order of Polymorph (Polycaprolactone/ShapeLock/themo set plastic/low melting-temperature polymer)

       

      David

      Thank you for all those figures.

      10-15km is 2-3 million steps. That requires resonably good engineering. 10
      hrs for Wild-Eagle is maybe 20,000 steps. I'm sure polymorph would have no
      problems with that.

      To me, I think it depends on why you're building the robot. I suppose most
      people's anwer would be to learn - for their own interest or to advance
      robotics research. So the robots that you've had on display are retired.
      They may be working harder than they did when they had a job but they are no
      longer active for your learning. (Of course, if you build a robot for, say,
      a museum display then its job is to move all day, every day and it must be
      built to do that. But only one in a thousand home-brew or university robots
      has that sort of life.)

      So my guess is still that polymorph hinges are perfectly good for 99.9% of
      (non-industrial) robots.

      > A pertinant question is how much energy does it take to flex a polymorph
      > hinge?

      Energy or torque? The hinge is springy so any energy you put into it to flex
      it is recovered when it straightens. I kept feeling the hinge while it was
      being flexed for its 10,000 trials and it didn't get warm so I guess that it
      wastes little energy.

      A friend suggested that if it begins to fail, it will be self-healing.
      Cracks can't propagate because the extra energy loss at the crack tip will
      raise the temperature and soften the polymorph, hence healing it. It's a
      nice idea if it's true.

      Of course, the next question is, how do you recover the stored energy? How
      do you make use of the springyness? If the hinge is connected to a servo
      (via a typical model-aircraft style wire con-rod) then the servo will
      probably be unable to make use of that stored energy. I guess it will depend
      on the design.

      The torque needed to bend the hinge is very roughly 0.15kg.cm/radian. The
      hinge region is 0.65mm thick less than 1mm long and 10mm wide. I suppose it
      all scales roughly linearly. (Or will it go up as the square of the
      thickness? - it's a long time since I learned about bending beams.)

      If a hinge typically bends 1rad during a stride then 0.15kg.cm is about a
      twentieth of the torque of a standard servo. If you've designed the leg
      properly then maybe that 0.15kg.cm can be used to partially support the
      weight of the robot.

      > It may be desirable in say a foot which needs to deform to absorb footdown
      > loads, but otherwise?

      Otherwise? You mean why bother with a polymorph design rather than the
      standard hexapod legs where the servos act as the joints? Standard hexapod
      legs may be the best design but I feel too few robots explore other
      possibilities. Standard hexapods are a cliche. Cliches aren't neccessarily
      wrong but it's always good to question them. Homebrew and university robots
      are for learning and you don't learn much about the physical design of a
      robot from Yet Another Standard Hexapod.

      If you're using polymorph then (I think) you have to use model-aircraft
      style con-rods. They don't have to be wire con-rods; they could be part of
      the single piece polymorph leg. So the new design has two independent
      features:

      (1) it uses con-rods
      (2) it's made of polymorph

      Con-rods have certain advantages. You can put the servos in the body - not
      on the joints. You can use the full rotation of the servo: if a typical
      servo can manage almost 180deg but the leg only swings 60deg then the
      con-rod can give you almost a 3:1 mechanical advantage. You can arrange the
      servos to give "unusual" motions to the leg. (I find it hard to design
      linkages like that from scratch so I've written a two-servo linkage
      editor/simulator. I'll upload it if people are interested.)

      Is polymorph worth using? I don't have any experience. As I've often said
      here before, robots are for learning. If you're learning then you don't know
      what you're doing; possibly no-one knows. Engineering is for when you do
      know what you're doing. Therefore 99.9% of robots should *not* be
      engineered. They should have a design life of hours, not years. Robots
      should be made of popsicle sticks and hot-glue. Polymorph *may* be a good
      material for lash-ups. If you don't like the design, it's cheap and easy to
      soften the polymorph and change it. How often have you thought "well, the
      design looked good on paper but I wish I'd made that part 1cm longer and now
      it's too much trouble to change it"?

      (I know that a lot of people like standing at their milling machine making
      beautifully crafted/engineered designs. I enjoy it too - it's a form of
      zen-like meditative relaxation. But that's a separate issue from making
      robots.)

      Personally, I spent many years working with crabs and thinking what
      fantastic robots they are. The few polymorph robots on the web have a
      wonderful organic look. It's probably a great material for making robot
      crabs.

      As I asked before:
      > If you were making a
      > 3 DOF one-piece polymorph leg, what sort of design would you choose?

      The answer is not obvious to me.

      Peter

    • Show all 24 messages in this topic