Re: Updated document on kin&dyn API...

Herman Bruyninckx writes: 

> On Sun, 29 Jun 2003, Pierangelo Masarati wrote: 
> 
> [...]
>> I have a question right from the beginning: why do you focus
>> on lumped parameter systems?  
> For several reasons:
> - most applications in robotics (and control) don;t need more.
> - infinite dimensional systems are quite a bit more complex, and
>   tackling everything at the same time is too much for an open source
>   project.
> - there are other projects out there (such as yours :-) 
> 
>> do you think deformable systems are out of scope, or do they
>> represent too hard a challenge, or what?
> Too hard. Not in principle, but in the robotics practice. And because
> I don't know enough of them :-) 
> 
>> Maybe if you don't consider them from the beginning you won't
>> be able to cast them into your scheme.  So, what's your concern
>> about deformability? 
> Of course I have been thinking about deformability and about infinite
> dimensional systems. But as far as I understand the problem, there
> isn't a common theory and code base anyway, so infinite dimensional
> systems will always be an 'add on', or a stand-alone thing. 
> 
> In our own lab, we have a big noise and vibration research group,
> which are also working on infinite dimensional problems; they work
> closely with a big CAE company (LMS, a spin-off of our department) and
> also they don't have a nice intregration between both domains (finite
> and infinite dimensional systems). 
> 
> In addition, many deformabilities in robotics can be quite well
> described as "lumped". For example, a flexible robot joint. These
> systems will be covered by the simple DAE approach, just be extending
> the state space of the model.

We are starting a research on a space robot with very soft and long
arms, with frequencies in the range of hertz; although all you said
above is correct about mixing finite and infinite dimensional systems,
I think we cannot simply use lumped systems to model this.  Of course
we will have to go to a lumped representation, but we need a physical
model.  Another problem: think of the lightweight robots that are
being developed at DLR; they operate at very high speeds and their
design is very lightweight.  I guess they may suffer from positioning
errors caused by deformability.  I recall the Modelica Multibody library
has been developed with these applications in mind (although, as far as
I remember, they still suffer from limitations in deformable parts
modeling).  I understand your limitations, but these aspects will be
more and more important also in robotics :) 

p. 

Dr. Pierangelo Masarati               | voice: +39 02 2399 8309
Dip. Ing. Aerospaziale                | fax:   +39 02 2399 8334
Politecnico di Milano                 | mailto:[email protected]
via La Masa 34, 20156 Milano, Italy   | http://www.aero.polimi.it/~masarati