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

Herman Bruyninckx wrote:
> On Sun, 29 Jun 2003, Pierangelo Masarati wrote:
> 
> [...]
> 
>>As I've pointed out in a paper we're presenting at the
>>Multibody Dynamics 2003 conference in Lisboa, coming next week,
>>http://mbdyn.aero.polimi.it/~masarati/Publications/os2003.pdf
> 
> 
> Great! I have been looking for such a paper about mbdyn :-)

Thanks.  Of course, it's not a technical paper in the classical
sense; I think it's my first "political" paper...

> 
> [...]
> 
>>One critical point is that niche projects
>>do not attract users.  This is particularly true for multibody
>>software, shich is very specialistic: 
> 
> 
> I am quite optimistic that this will change in the near future. The
> only problem at this moment is that the users of these kind of sofware
> lag behind a year or two-three with respect to the users that use and
> hack the Linux operating system. Anyway, I am very interested in
> contributing to mbdyn.

You're welcome.  I think the best way for academics
to contribute is to have a student (an undergraduate?)
try to make it work and model some problem of interest.

Everybody can contact me, or the project, for any
build/install problem or for modeling suggestions.

> I also do have some quite worked out software
> engineering suggestions, such as the Object-Port-Connector pattern
> that was worked out in a master thesis this year
>  <http://people.mech.kuleuven.ac.be/~bruyninc/thesis-cosaert.pdf>

I've looked at it with interest.  I guess the basic idea
is somewhere hidden in the way MBDyn treats nodes for
generic elements: most of the entities that generate
outputs, including nodes of any sort, measures on element
properties and more, can be used as inputs to generic
elements by masquerading them under the scalar degree of
freedom interface.  If the element is used in equations,
it must be related to a dof, otherwise it can be anything.
Unfortunately, this has pot been implemented as generally
as it would have been required, but it gives all the
required flexibility when connecting entities not explicitly
designed for connection (e.g. a force to some fancy measure,
or the elongation of an actuator to the desired rotational
speed of a helicopter rotor).

> As soon as mbdyn is decoupled  from its initial target application
> (aerodynamic simulations) and features a common core on top of which
> any sort of dynamics application can be implemented,

I think it is decoupled (at least from a user's perspective).
You start froma  blank page and need to add nodes (public dofs)
and elements (write equations).  You build your system by adding
elementary blocks, aerodynamics comes on top if you need it.
You can model a hydraulic circuit (there's a tutorial about a beam
actuated by a piston, taken from the literature), a control
circuit or whatever (mechanical systems are preferred, anyway).

> the user response
> will be much larger. Now you only advertise mbdyn mainly as an
> aerodynamics package, 

actually, I don't see so much aerodynamics in the web page;
it is only mentioned as a "bias" because it is something we
use a lot (and we are developing a lot, but in a general
sense).

> which ``scares away'' visitors of the website
> that are interested in other dynamics domains.

...

> 
> [...]
> 
>>I expect robotics software to have a much wider basin of potential
>>users; 
> 
> Not really that much wider than general dynamics simulation!
> 
> 
>>However, both projects de facto address
>>niche problems, so ddo not expect much feedback unless you provide
>>from the beginning something that works well enough to attract users. 
> 
> 
> I know, and that's why I added this ironic remark about "the
> Cathedral" :-)
> 
> Thanks for your feedback anyway! (Even if it was "off topic" in the
> context of my post :-)
> But it is not really off topic in the Orocos context: you might have
> seen the stream of posts during the last week, discussin the "Orocos
> II" proposal; in that proposal, simulation of dynamical systems is one
> of the work packages, for robotic systems but also other "bond graph
> like" dynamical systems.

We are considering Orocos II, maybe we don't have resources
for this project that is a little off-topic for us.  But we
look at it with interest, because of potential contaminations,
and as a source of ideas.

And thanks for your feedback as well...

p.

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