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

Fwd: Re: [SeattleRobotics] Re: Scientific Methods in Robotics

Expand Messages
  • rfscheer@speakeasy.net
    Jeff has hit the bullseye. It would be great to see an amateur study of the balancing bot in which aspects of stability were characterized that could be
    Message 1 of 2 , Nov 2, 2007
    • 0 Attachment
      Jeff has hit the bullseye. It would be great to see an amateur study of the balancing bot in which aspects of stability were characterized that could be directly compared with other bots. Can't talk now. Am travelling. Yay Jeff! - Robert

      On Fri Nov 2 6:14 , 'Jeffrey Birt' sent:

      The basic premise is that one cannot prove things (theories) to be true, they can only be proven false. A corollary is: Theories are neither true nor false, they are simply useful or not (IIsac Asimov.) The point of course, is not disproving theories but developing better ones. That is, developing a better understanding of the subject at hand. While we can debate as to what 'science' is, the methods employed in science are directly applicable to our robotics hobby.

      Let's say you develop a new control algorithm for a balancing robot. You would not be satisfied that it could balance if unperturbed, rather you would design a set of experiments designed to make it fall over. So, you cannot prove that it will never fall over but you can determine the bounds of its stability. Jeff Birt
    • John Palmisano
      I figure I should solve this dispute ;) So many of you are arguing about how to calculate how one robot design is optimal over another, how to determine energy
      Message 2 of 2 , Nov 3, 2007
      • 0 Attachment
        I figure I should solve this dispute ;)

        So many of you are arguing about how to calculate how one robot design is
        optimal over another, how to determine energy and strength and how high it
        can climb, etc.

        This is what a mechanical engineer does! Its all basic math, really. We
        don't guess robot designs, we optimize by calculating forces and stresses
        and speeds all quantitatively.

        I took the liberty to put these calculations here:
        http://www.societyofrobots.com/calculator.shtml

        You enter your robot specs and it will tell you quantitatively how well your
        robot will work. Of course it is slightly generalized because it cannot
        account for every possible robot design, and friction is almost impossible
        to calculate (no, normal force times friction coefficient is highly
        inaccurate and requires experiments).

        There are many other techniques I use when optimizing mechanically,
        including computationally intensive CFD, FEM, numerical methods, and genetic
        algorithms. This isn't really science, more just engineering. Its already
        solved scientific problems years to decades ago, Im just applying
        the science for new applications.

        The next calculator I am working on will help you optimize robot arm design
        (how much will it sag? what is the maximum weight it can lift at Y
        extension? etc.). If anyone has more ideas on what they would like as an
        automatic calculator, let me know.

        John
        societyofrobots.com


        2007/11/2, rfscheer@... <rfscheer@...>:
        >
        > Jeff has hit the bullseye. It would be great to see an amateur study of
        > the balancing bot in which aspects of stability were characterized that
        > could be directly compared with other bots. Can't talk now. Am travelling.
        > Yay Jeff! - Robert
        >
        > On Fri Nov 2 6:14 , 'Jeffrey Birt' sent:
        >
        > The basic premise is that one cannot prove things (theories) to be true,
        > they can only be proven false. A corollary is: Theories are neither true nor
        > false, they are simply useful or not (IIsac Asimov.) The point of course, is
        > not disproving theories but developing better ones. That is, developing a
        > better understanding of the subject at hand. While we can debate as to what
        > 'science' is, the methods employed in science are directly applicable to our
        > robotics hobby.
        >
        > Let's say you develop a new control algorithm for a balancing robot. You
        > would not be satisfied that it could balance if unperturbed, rather you
        > would design a set of experiments designed to make it fall over. So, you
        > cannot prove that it will never fall over but you can determine the bounds
        > of its stability. Jeff Birt
        >
        >
        >



        --
        John Palmisano

        Robotics Specialist
        Naval Research Laboratory, Washington, DC
        B.S. Mech. Eng., Robotics, Carnegie Mellon University
        www.societyofrobots.com


        [Non-text portions of this message have been removed]
      Your message has been successfully submitted and would be delivered to recipients shortly.