Newsgroups: comp.robotics
Path: brunix!uunet!pmafire!mica.inel.gov!ux1!news.byu.edu!hamblin.math.byu.edu!sol.ctr.columbia.edu!usc!cs.utexas.edu!bcm!aio!drseus.jsc.nasa.gov!graves
From: graves@drseus.jsc.nasa.gov (Philip Graves)
Subject: Re: Modeling Slip, Mobile Robotics, Kinematics, Simulation
Message-ID: <1992Nov17.144824.26839@aio.jsc.nasa.gov>
Sender: news@aio.jsc.nasa.gov (USENET News System)
Organization: Lockheed Engineering & Sciences Company
References: <1e8eknINNdn6@clover.csv.warwick.ac.uk>
Date: Tue, 17 Nov 1992 14:48:24 GMT
Lines: 43

In article <1e8eknINNdn6@clover.csv.warwick.ac.uk> esrbu@csv.warwick.ac.uk (Mr I M Hunter) writes:
>
>        Has anyone attempted to model the effects of slip: What happens to
>an omnidirectional robot when wheels are misaligned or differ in size? 

We have two Cybermotion K2A mobile platforms which have a special drive
train mechanism called "Synchro-Drive". The three wheels of this
system are steered together so that they are always parallel (if aligned).
Our experience is that when the wheels are not aligned, the system
will always drive in an arc, instead of a staight line. One of the systems
was modified to make it narrow enough to fit through a door (by moving
the wheels towards the center of the platform). This caused the arcing
problem to be worse, and the wheels have to be aligned more accurately
to get it to drive a straight line.

>do you model which wheel (or wheels) slip, by what degree and what is
>the resultant force on the body of the robot? How would YOU model static
>and dynamic friction? And this is after any errors within the robot itself

I have used classical models of static and dynamic frictions, (requiring 
iterative solutions which are a major pain because they are 
not well suited for simulation). There are several sophisticated 
tire models used by automotive companies to model vehicles, and are 
available as part of commercial dynamics modeling and simulation 
packages such as ADAMS. 

>Am I simply a misguided man attempting
>to model the (up to now) unmodelable? All I actually want to do is model a
>couple of effects without trying to create a 'perfect' model (cos that is a
>looser: what about differences in floor surfaces or angle of floor etc..)

In general, you can only get a qualitative idea of what might happen
from any model of this type. It is impossible to find the perfect model
because the friction depends on a plethora of factors, including 
everything from temperature to how long it has been since the floor 
was mopped and waxed.

Good Luck.

-- 
--------------------------------------------------------------
Philip Lee Graves, graves@drseus.jsc.nasa.gov
--------------------------------------------------------------
