Newsgroups: comp.robotics
Path: brunix!uunet!newsgate.watson.ibm.com!news.ans.net!malgudi.oar.net!zaphod.mps.ohio-state.edu!sdd.hp.com!hp-cv!hp-pcd!hpcvca!hpcvcec.cv.hp.com!abrown
From: abrown@hpcvcec.cv.hp.com (Allen Brown)
Subject: robot parts: info and question
Message-ID: <1993Jan31.054717.2297@hpcvca.cv.hp.com>
Sender: nobody@hpcvca.cv.hp.com
Nntp-Posting-Host: hpcvcec.cv.hp.com
Organization: Hewlett-Packard Company, Corvallis, Oregon USA
Date: Sun, 31 Jan 1993 05:47:17 GMT
Lines: 120

I am assembling the people, expertise, and parts to make a robot.  It
will be a cat sized six legged creature modeled after a beetle.

References so far:
1. Minimalist Mobile Robitics by Jonathan H Connel
   ISBN 0-12-185230-X
   This talks about a Brooks inspired wheeled robot designed to
   collect soda cans.  What is striking about it is that the control
   system includes no internal communication between modules.  All
   interaction is communicated thru effects on the real world.  Also
   modules have essentially no memory.  I wouldn't generally recommend
   this approach, at least not to this extreme.  But I recommend the
   book.  It gives a valuable perspective on what can be accomplished
   when the robot has a very limited intelligence.
2. Machines That Walk by Shin-Min Song and Kenneth J Waldron
   ISBN 0-262-19274-8
   This book is strong on gaits.  I have only read part of it.  But it
   gives valuable information on maximizing the static stability of
   the gait.

Power will come from NiCads.  I have a place I can get sufficient
quantity of these batteries super cheap.

We have selected the main motors.  We will be using the TRW 405A223.
At more than $20, these motors would normally be too expensive, but I
found a friend with just enough of these that he was willing to sell
cheaply.

We also bought for evaluation about ten motors from a couple of
surplus houses.  Several of those motors were approximately what we
were looking for.  I recommend you get catalogs from:

1. American Science and Surplus
   (708) 475-8440
2. Mendelson Electronics Co., Inc
   (800) 422-3525

We considered stepper motors but settled on DC motors.  Steppers tend
to take too much power, even when doing nothing.  They tend to be more
expensive than DC motors.  Steppers have the advantage of one pulse
translating to a fixed amount of motion.  But you have to have some
additional sensor to find out where you are when you start.  (This is
generally a limit switch.)  But if you are going to install an
additional sensor, why not use a pot and measure position directly?

We also considered hydraulics and pneumatics.  We ruled out hydraulics
because of cost and because we couldn't guarantee it wouldn't spring a
leak on the front room carpet.  And I didn't like pneumatics because
of the lack of control.  Basically all you get is bang-bang.

Nitinol (memory wire) is out because of energy inefficiency and lack
of fine motor control.  Too bad.  It would have been nice to have
direct linear.

We have discussed various ways to translate the energy from the motor
to the leg.  We haven't found a source for gears at anywhere near a
reasonable price.  Partly for that reason we will probably go for a
cable driven by a screw.  I worked on that today.  It still needs
work.  I plan to drive 6-32 threaded rod directly from the motor.
This will push and pull a steel block by way of a threaded hole thru
it.  The cable will be attached to the steel block.  I would like to
use acrylic plastic as the housing that the steel block slides within.
Oh, I know that is not ideal as a bearing surface, but I want to be
able to see that the cable is not buckling.  I figure this will make
debugging easier.

The main structural components such as body and legs will probably be
made of model aircraft plywood and fiberglass.  High strength and low
weight are what makes me lean toward them.

The uC will be the Pic.  (I don't know if that is how it is spelled.
Others in my group are familiar with this computer.)  The Pic is in a
quartz lid 18 pin DIP.  It has 8 digital I/O lines and 4 analog input
lines.  There is a 8 bit ADC multiplexed onto the analog inputs.  As I
recall the Pic has 512 bytes EPROM and 32 bytes RAM.  I expect we will
use one Pic per leg and one to supervise.
  uC communication will be partitioned as:
    digital
      7 - Commands Serial In
      6 - Commands Serial Out
      5 - Motor3 CW
      4 - Motor3 CCW
      3 - Motor2 CW
      2 - Motor2 CCW
      1 - Motor1 CW
      0 - Motor1 CCW
    analog
      3 - Foot Force
      2 - Motor3 position pot
      1 - Motor3 position pot
      0 - Motor3 position pot

We do not yet have the motor drive electronics.  I am hoping we find a
suitable IC.

We have also not worked out how to measure foot force.  But we know
two of the constraints: its gotta be small and cheap.

And we have not yet planned any sensors.  My guess is that we will
include light sensors and possibly a sound sensor.  We will also
probably include either a radio link or IR.

How much will all of this cost?  Hell if I know.  My original
guesstimate was $2000.00.  But we have been coming in substantially
under target.  That is a trend I want to maintain!


Now the question:
Any advice?  Do you know of a good cheap source for gears or pullys?
(Like about $0.50 per gear.)  Do you know of a better way to deliver
force to the legs?  I don't really want to design and build my own
motor to cable interface.  Do you know where I can get this for less
than $2.00?  How about a small cheap strain guage?

Are there others in the Willamette Vally working on small legged
robots who want to compare notes?
--
  Allen C. Brown  abrown@cv.hp.com or hplabs!hpcvca!abrown or "Hey you!"
      Not representing my employer.
    "Don't dream it... Be it." Rocky Horror Picture Show
