Newsgroups: comp.robotics
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!news.sprintlink.net!pipex!uknet!gdt!uwe-bristol!wj-wray
From: wj-wray@uwe.ac.uk (WJ Wray)
Subject: Re: [RESEARCH Q] How do *you* model manipulator joint friction ?
Message-ID: <1995Mar6.224121.12155@pat.uwe.ac.uk>
Sender: usenet@pat.uwe.ac.uk (uwe nntp usenet poster)
Nntp-Posting-Host: usenet@pat.uwe.ac.uk (uwe nntp usenet poster)
Organization: University of the West of England, Bristol
X-Newsreader: Tin 1.1 PL5
References: <3jel5u$mf5@ztivax.zfe.siemens.de>
Date: Mon, 6 Mar 1995 22:41:21 GMT
Lines: 46

 
More generally this question relates to modelling of hybrid systems
in which continuous dynamics and discrete events are mixed.
How these models are simulated is another question, though this
debate shows some confusion between `modelling' and `simulation'.

The issue is not one of low level modelling; the theoretical models
for friction are simple equations easily visualized. Take your pick.
There are plenty of commercial simulators which can handle events,
such as a transition from static to dynamic friction, so there's
no need to write your own simulation code, except for lack of $s :(
The problem is that todays simulators require their equations to be
in particular forms and it's the job of the modeller to reduce the
set of model equations to a form suitable for simulation. This job
is not `modelling' or `simulation' but `munging' or `manipulation'.
It is tedious, error-prone, hard and needs to be automated.

Physical systems like manipulators are ideally suited to an object-
oriented modelling methodology in which joints, links, couplings,
drives and so on are specified as separate `classes' which can be
`instantiated' and then connected to form the complete model.
The object-oriented modelling approach has been very successful for
pure discrete event systems because the high level model can easily
be directly simulated, after all the simulator is just a scheduler.
For continuous systems the object-oriented approach leads to a huge
set of equations, resulting from the connections between objects,
which can't be directly simulated using current integration methods
but which much first be reduced to ODE or low-order DAE form.

Recently a couple of tools have appeared which address this issue.
The approach is to use object oriented languages to specify models
in an equation-based manner independent of simulation techniques.

I recommend you read `Continuous System Modeling' by F.E.Cellier
(springer verlag 91, 0-387-97502-0) for more on modelling.

The paper `Object-oriented Modeling of Hybrid Systems' ESS'93,
European Simulation Symposium, Delft, The Netherlands, Oct 93, by
H.Elmqvist, F.E.Cellier and M.Otter includes dry friction models.

BTW *I* model manipulator joint friction using a multi-body
library developed for the modelling language Dymola.

  Will Wray
   IAS Lab
     UWE
