Re: Orocos meeting: design presentation?

The following comes from a private information exchange between
GianAntonio Magnani and myself. I bring it to the mailinglist because
it contains relevant remarks for discussion.

Herman

PS I got GianAntonio's consent to violate the privacy of his email :-)

---------- Forwarded message ----------
At 13.39 01/08/01 +0200, you wrote:
>On Wed, 1 Aug 2001, GianAntonio Magnani wrote:
>
> > At 15.18 31/07/01 +0200, you wrote:
> >
> > >Would you be interested/willing to give a short overview of your
> > >design ideas and experiences during the Saturday afternoon session of
> > >the Orocos meeting? (I hope you do!)
> >
> > I am happy to share my experience related to industry world. I could give
> > first a short overview of the Comau C3G controller (the standard and,
> > especially, the open version) then give my point of view on design
> > requirements and approach. I think 15 minutes are enough for my
> > presentation, since the C3G controller is for sure not the latest in
> > technology, and it may not be so interesting for attendees.
>
>Fine, thanks!
>
> > An issue could worth some attention, that I didn't find in my quick reading
> > of your design
> > `document' on-line on the Orocos website: That of the "robot controller
> > reference model". I mean something similar to the ISO/OSI model of the
> > communication systems or, in other words, a sort of canonical form of the
> > controller, as it is called in RCS: a reference model architecture for
> > intelligent control, Albus, J.S., Computer , Volume: 25 Issue: 5 , May
> > 1992, Page(s): 56 -59.
>
>Well, this is an interesting topic :-) We (and many others) have
>become very critical about this reference model, because of its
>inflexibility. We are now more thinking along the way of ``software
>agents'' (holons, components, whatever you want to call them).
>
>But my hope and expectation is that (part of) the Orocos code will be
>useful for different architectures; I mean, basic libraries such as
>kinematics and planning etc. Anyway, it is also not my intention to
>come up with one single architecture, or `the' best design: I hope we
>can build pieces that form the basic infrastructure of any robot
>system architecture you will want to make.
>
>But this point is certainly one of the most important ones for this
>meeting.
>
> > A more detailed description of the NIST (National Institute of Standards -
> > USA) reference model is given in Hierarchical control of intelligent
> > machines applied to space station telerobots, Albus, J.S.; Lumia, R.;
> > McCain, H., Aerospace and Electronic Systems, IEEE Transactions on ,
> > Volume: 24 Issue: 5 , Sept. 1988, Page(s): 535 -541.
>
>NIST is using this architecture for another open source control
>project: EMC <http://www.isd.cme.nist.gov/projects/emc/>.
>I have been in (email) contact with the leader of this project, but
>our views didn't match. Of course, I have no objection against sharing
>code whenever that would be possible.
>
> > A model (the Functional Reference Model) was also defined by ESA/ESTEC as
> > an essential part of their (automation and robotics systems) Control
> > Development Methodology. I will ask Mr. Peter Putz (ESTEC) to forward me
> > the relevant electronic documentation.
>
>We (not me personally, but a number of colleagues) have been working
>with ESA in this framework... and that has been one of our most
>disappointing experiences :-)
>
> > I am not particularly an enthusiast of reference models, but I think
> > adopting one could help in a wide scope project like Orocos. If you wish, I
> > could briefly introduce this topic in my presentation.
>
>I would welcome any of your ideas about the topic: we should start the
>project without prejudices :-)
>