[Modeling] AUML diagrams

Bernhard Bauer bernhard.bauer@informatik.uni-augsburg.de
Mon, 14 Apr 2003 17:59:02 +0200

Dear Sehl Mellouli,

>Dear Mellouli,

>Could you provide more detail on you opinions?
>Perhaps, you could also contribute some of your own ideas and papers --
or write chapters 
>for AUML specifications even now. Also, you say wish (as we do)
"solutions that can be 
>mapped to code easily."  Can you give us more detail?  (Particularly,
when there are 
>members of the AUML team that are now transforming AUML diagrams to

>At this point, we need participants that will provide tangible ideas
and contributions 
>that can be turned into specifications.  We hope that you will help in
this manner.


>Jim Odell

On 4/11/03 6:14 PM, "sehl mellouli" indited:

> Dear All,
> I have some comments on AUML diagrams. AUML is based on UML with some 
> deviation/extension on notations. Nevertheless, IMO, the difference 
> between AUML and UML is not only notation but also the agent concept 
> itself. Agents are not objects; they have another structure (mental
> structure) that objects have not.
> An object is described by attributes and functions, but not an agent. 
> An agent has its beliefs, desires and intentions (this is an agent 
> structure), if we consider BDI architecture.

I tried to point out these agent specifica in my paper on class diagrams

allowing the specifier to describe BDI agents. Using wff for the
description of the 
state of an agent. 

> This structure can be 
> described by attributes and functions itself but the agent structure 
> is not attributes and functions.

Yes, but attributes describing the BDIs of an agent...

> Agents can share knowledge, can play many roles, can see their roles 
> changing during execution. Does a class diagram with some new 
> stereotypes consider these aspects? IMO, I don't tnink so.

Sharing knowledge is not addressed with the agent class diagrams. We had
discussed this issue elsewhere. We hope to find a solution for this

For sure roles are an important aspect and can change during execution.
However we 
think this role change can be described elsewhere. See e.g. the paper of
Jim et al. 
on this subject.

> IMO, object oriented technology is widely used because object-oriented

> methodlogies propose concrete solutions, that are solutions that can 
> be mapped to code easily.
> Looking at the AUML diagrams, are they easily mapped to code? can an 
> agent be developed as a class? I think no.

Well having a Java based agent platform in mind, like JADE-LEAP, we have
to map
an agent into a specific agent class. Thus, I strongly believe that a
code generation
is possible to some extend.

> Developing diagrams is vital to design MAS but also we have to 
> consider that these diagrams will be developed later and there is a 
> big gap between agent theory and agent development.

> I propose, in developing AUML diagrams, to consider how agent specific

> aspects will be developed.

see Jim's e-mail,

Best regards,
> Best regards,
> --
> Mellouli Sehl
> Computer Science Engineer
> Information Systems Administration MBA
> Computer Science Phd Student
> Université Laval, Québec, P.Q, Canada
> Tél: bur (418) 656-2131 (4704)
> Home page: http://www.ift.ulaval.ca/~mellouli

Modeling mailing list

Prof. Dr. Bernhard Bauer 
Institute of Computer Science 
Programming of Distributed Systems 
University of Augsburg 
Universitätsstraße 14 
D-86135 Augsburg 

e-mail: bernhard.bauer@informatik.uni-augsburg.de 

Tel. +49 821 598 2118 
     +49 821 598 2174 (Secretary) 
Fax. +49 821 598 2175 

Office Munich: 
Tel. +49 89 612 09 026 
Fax. +49 89 612 09 427