Newsgroups: comp.robotics
Path: brunix!sgiblab!spool.mu.edu!howland.reston.ans.net!usc!elroy.jpl.nasa.gov!decwrl!netcomsv!netcom.com!nagle
From: nagle@netcom.com (John Nagle)
Subject: Re: Drive/steering wheel configurations?
Message-ID: <nagleCH2G08.KLJ@netcom.com>
Keywords: wheels drive steering
Organization: NETCOM On-line Communication Services (408 241-9760 guest)
References: <1993Nov25.143300.10664@julian.uwo.ca>
Date: Thu, 25 Nov 1993 21:07:19 GMT
Lines: 71

mheim@mind.cogsci.uwo.ca (Michael Hiemstra) writes:
>I attended the last BEAM robotic olympics with a college class
>entry.  This year I'm in the working world and would love to
>have another entry.

>My main hurdle at the moment is deciding on a wheel configuration.
>This time around, I want to develope something small, quick,
>and agile.
>My questions are:
>1) Has anyone found one wheel configuration to be better?
>2) Is there any research papers on the topic (which can be
>   ftp'd)?
>3) What's your personal experience with different applications
>   and different configurations.

      Basically, there are two approaches: some kind of 
omnidirectional drive, and building something that works like a
real vehicle.  

      There are a number of omnidirectional drive mechanisms.  The
fanciest ones can simultaneously rotate and translate in any direction.
This involves big wheels with rollers around their perimeter.  
Vehicles with 3 such wheels were tried at Stanford, and Moravec built
a 4-wheeler that way at CMU.  In general, that approach only works on
nice flat floors.  Outside travel is generally out.  Even carpet is a
problem.  The big advantage, though, is that you don't have to plan 
movement around the directional limitations of the drive system.

      I wouldn't recommend that approach.  The machines which use it
are clunky, and the complicated roller-studded wheels tend to have
problems with anything but the most benign floors.

      More common is omnidirectional drive which can move in any direction,
but can't change position and orientation independently.  This is usually
implemented as three wheels, all of which steer and drive together.
Cybermation builds a big base that works this way, and Real World 
Interface builds a little one.  In both cases, the upper body of the robot
rotates as the wheels turn, so it is always facing "forward".  Only two
motors are required, one for drive and one for steering.

      Generally a good compromise.  Cheap, simple, and reliable.  
Not well-suited for outdoor use, though, or for high-speed travel.
Control is straightforward, although the resulting motion is inelegant.

      The next step is tracked vehicles.  Great for rough terrain,
but hard on floors, especially carpet.  Many teleoperators are tracked.
Rare for research robots.  

      A variant on a tracked vehicle is paired multiple wheels.
Six-wheeled all-terrain vehicles are good off-road machines, and
toy versions are worth considering as robot bases.  Used occasionally
for research robots.  Has most of the advantages of tracks with
less floor damage.

      Two powered wheels and an unpowered trailer (a "caster") is a common
compromise.  It's the cheapest approach.  Androbot and Heath both used
this geometry for their early "home robots".  Control is straightforward.
The caster tends to get stuck in things, since it is usually a small wheel.

      A variant on this is two powered wheels and one steered wheel.
This is the forklift configuration.  Caterpillar makes an autonomous
robot vehicle in a forklift configuration.  This arrangement gives
good positional control, and can be built very ruggedly.  (Caterpillar's 
demo video shows these machines working in a plant yard during a snowstorm,
carrying a load of about a ton.)  The control geometry is non-trivial,
but if you need to do heavy work, it's a good configuration.

      Beyond that, you're into standard vehicle geometries, like
cars.

					John Nagle
