Newsgroups: comp.robotics
Path: brunix!uunet!destroyer!caen!uwm.edu!wupost!csus.edu!netcom.com!park
From: park@netcom.com (Bill Park)
Subject: Source for OmniWheels(tm?) ?
Message-ID: <1993Mar16.225135.7388@netcom.com>
Followup-To: comp.robotics
Summary: Who sells OmniWheels(tm)?
Cc: Park@netcom.com
Organization: Netcom Online Communications Services (408-241-9760 login: guest)
Date: Tue, 16 Mar 1993 22:51:35 GMT
Lines: 140

A quadraplegic friend of mine is looking for 12-inch diameter
OmniWheels(tm) to use as front traction casters slaved to the main
wheels in his patented wheelchair design.  The only known source,
Intex Medical (San Antonio, Texas, USA), no longer has a telephone
listing.  Any suggestions for alternate sources would be greatly
appreciated.

I have already been in contact with Prof. Larry Leifer and Machiel Van
der Loos at Stanford, Butler Hine at the NASA-Ames mobile robot demo
facility, and Brian Carlisle at Adept, Inc.  Nobody knows where to get
OmniWheels if Intex is out of business.

Several years ago I saw a picture of a Japanese OmniWheel-like
wheel design for vehicles.  I can't remember the source or the
Japanese company, however.  Popular Science or Popular Mechanics or
Design.  Perhaps Mitsubishi.  Does anyone in Japan know of these?

What is an OmniWheel?
=========================

An OmniWheel -- and I don't even know if that is actually
someone's trademark -- is a wheel with from four to twelve rollers all
around the rim.  The axes of the rollers lie in the plane of the
wheel, at right angles to the shaft of the wheel, and tangent to the
perimeter of the wheel.  (Wish we all had graphic capability in email
and Usenet!)  That means the wheel will provide traction if you put it
on the floor and turn it around its axle.  It will also slide freely
in the direction of the axle.  The rollers are slightly barrel-shaped
to provide a constant radius and smooth ride.  

Another variation is two of these wheels side-by-side fixed to the
same axle, but with one OmniWheel rotated around the axle so that the
centers of its rollers are opposite to the ends of the other
Omniwheel's rollers.  That allowed longer rollers and maybe a smoother
ride.  Yet another variation is a dual-wheel design with the rollers
axes at about 10-15 degrees to the plane of the wheel.  Why, I can't
imagine, but that is the Japanese design I recall seeing.

Some History
============

I'm told that OmniWheels were first patented back in 1910 (a double
wheel design with rollers each) and that they are in use in various
ways today.  For example, to allow pallets of cargo to be slid and
positioned easily in any direction while being loaded into a cargo
aircraft.  Another use is in conveyor lines.  A table whose top is
covered with OmniWheels forms a shunting station at which goods can be
slid sideways off the first conveyor and onto a second.  Perhaps if
these kind of OmniWheels come in a 12-inch diameter, the problem may
be solved.  Does anyone know of a vendor?

While at Stanford, Dr. William La conceived of using three




OmniWheels to make a wheeled mobile robot base that had all three
degrees of freedom over a floor--that is, it could move in any
direction while turning in either direction.  I heard he originally
intended to make an improved bumper car for amusement parks!  From a
standing start, it could head off in any direction instantly, without
first having to turn itself to face the direction of travel (as do
fork lifts and vehicles with two differentially-driven drive wheels
for steering, such as wheelchairs), or to rotate its wheels to drive
in that direction (as must, for example, the Cybermotion and Denning
mobile robots).  To achieve this, La mounted the three OmniWheels with
their axes horizontal and 120 degrees apart.  Driving the three
OmniWheels at different positive or negative speeds could produce any
combination of translatory and rotatory motion over the floor.  Dr. La
is now off in Florida somewhere getting an M.D. degree, and I haven't
tried to reach him about this.

It was an elegant design for omnidirectional mobility because it was
completely symmetrical and required the minimum number of motors:
three, one for each degree of freedom, though no one motor produced
either pure translation or pure rotation about the center of the
robot.  Other omnidirectional mobile robots have as many as six motors
to drive three wheels, adding to their cost, weight, and mechanical
complexity.

The only disdvantage I can see to the design is that the rotation
rates of the three OmniWheels must be tightly regulated in order to
accurately produce the desired robot motion.  That required
variable-speed servo control of each OmniWheel.  It would be an
interesting control system design exercise to decide how to close the
loop with navigation sensors to correct for slippage of the wheels on
the floor and errors in rotation rate.  Simply position-servoing each
wheel wouldn't necessarily get you where you wanted to go, because
even if transient wheel angular position errors netted to zero, the
final robot position could still be wrong.  Some interesting
nonlinearities lurk, I think.  Maybe a good neural net project for
some undergraduate.

As part of a project to develop rehabilitative aids for the disabled,
the Palo Alto Veterans' Hospital and Stanford built a mobile robot
base according to La's design, and even mounted a Puma 250?










manipulator on it.  The robot arm's controller did double duty: To
move the robot, it switched three of its inboard arm axis drive
ciruits over to drive the three servomotors that turned the
OmniWheels.  Brian Carlisle's robot company, Adept Inc., had it for a
while -- maybe they even built it.  They made a video showing it
rolling around one of their workshops with Brian riding on it.  The
video also demonstrated how its superior agility made it very easy to
drive the the mobile robot accurately into a docking fixture at a
fixed workstation for manufacturing applications.  It has wound up
with several other mobile robots at NASA-Ames Research Center
(Mountain View, CA), where it is part of a project to develop a
control system for multiple mobile robots.

Stanford's Mechanical Engineering Department also built a wheelchair
with the same wheel arrangement.  I think it was operated entirely by
moving your head: Two Polaroid sensors in the seat back tracked your
head motions and translated them into the steering commands.  Some
people who tried it said it was like driving a highly-manuverable
sports car.  It was easier to get in and out of a toilet stall with it
than with a regular powered wheelchair, for some reason that I forget.

I've often though that this was a fine design for commercial mobile
robots, wheelchairs, and automatically guided vehicles in factories,
but it has been almost completely ignored.  Except for Intex, I guess.

Thanks for any help you can provide. Please post to comp.robotics so
we can all see.

Bill Park
=========


-- 
Grandpaw Bill's High Technology Consulting & Live Bait, Inc.
