Meeting of "GUI workshop", Leuven June 20th 2003

Here is the draft of the meeting minutes. Please, people that
attended, feel free to mail corrections or additions; I will collect
them on the webpage <http://www.orocos.org/meeting-jun2003.html>.

Herman Bruyninckx
===========================================
Present: Herman Bruyninckx (KULeuven), Sara Fleury (LAAS), Christian
Schlegel (FAW), Boris Kluge (FAW), Anthony Mallet (LAAS), Marco Gritti
(Politecnico di Milano), Monica Reggiani (University of Parma),
Francesco Monica (University of Parma), Peter Soetens (KULeuven),
Panagiotis Issaris (KULeuven).
                                                                                
There were presentations by Herman Bruyninckx, Marco Gritti, Boris
Kluge and Monica Reggiani. They discussed requirements, current
systems, design issues, etc. Here is a list of some of the more
general discussion points:
                                                                                
- there is a rather clear understanding of "clients" and "servers" for
  visualisation, and that the activities of both are not to difficult
  to decouple.
- there are two big multi-platform 3D APIs: Java3D and OpenGL (+ the
  formats and tools built on top of that, such as OpenInventor and
  VRML). Many of these GUI clients are available in the community.
- the main reason why there is so few real integration is the lack of
  common "Models" of the robotics objects that everybody uses: rigid
  bodies, kinematic chains, laser scans, velocities, etc.
- XML Schemas provide an appropriate way to encode all robotics
  models. The same information can be re-used for several purposes
  (possibly with automatic generation from the XML files): object
  communication through the network ("marshalling"), class hierarchies
  in C++ or Java, educational documentation, etc.
- the "View" and "Control" parts of the Model-View-Control pattern are
  not much of a practical problem; the Models are.
- the SmartSoft-based implementations cover a very large set of
  robotics needs. Some of the problem areas: the system is fully
  event-based, which implies some overhead in areas where a "polling
  based" client-server interaction would be more appropriate; the Qt
  runtime is not fully thread-safe; drawings are fully repainted
  everytime.
- the SmartSoft design decouples the "drawing widget" from the
  "visualisation painter"; the latter one is fully widget independent.
- "streaming" data for visualisation could be implemented with the
  CORBA or TAO "pluggable protocols".
                                                                                
The meeting resulted in several suggestions for concrete actions. The
one with the largest chance for success is withheld for posting a
"call for participation" on the mailinglist: the XML encoding of
robotics objects.

Herman Bruyninckx