[Modeling] AIP and actions

Renato Levy rlevy@i-a-i.com
Thu, 19 Aug 2004 10:12:54 -0400

At 09:00 AM 8/19/2004 -0500, Marian Nodine wrote:
>We also took the state machine approach in InfoSleuth. There was one state 
>machine per participant in the conversation, and the structure of the 
>state machine was tailored to the role in the conversation. Transitions 
>not only are actions, but may also include other things such as timeouts.
>The only real issue that I have with this approach is that I think agents 
>need to move to a situation where conversations are more dynamic, and not 
>actually preselected from a set of protocols. Renato, I know that the DIVA 
>tool links these, I was wondering how it would deal with something more 
>flexible, e.g., composable conversation fragments?
>-- Misty

We are in the process to implement an overview diagram that allows to 
compose multiple protocols into into one. Runtime composability must still 
be done by hand (Cybele supports it, DIVA does not), mostly because we have 
been able to find a CLEAN way to express it.

>Renato Levy wrote:
>>In our tool (DIVA) we have linked AIPs with state diagrams, activity 
>>diagrams are not used. Each state diagrams represent a "machine" for one 
>>role. There is way more detail to it, if you are interested.
>>At 01:23 PM 8/16/2004 +0200, Marc-Philippe Huget wrote:
>>>Hello everybody,
>>>I have several times the same question during AAMAS: we have AIP but 
>>>what about actions associated to the AIP? How to represent them? I 
>>>proposed in the first draft to decorate the AIP with actions but this is 
>>>certainly too simple. I think we need to consider this question and find 
>>>an answer to everybody who wants to design and **implement** AIPs.
>>>First idea: linking Interaction diagrams to Activity diagrams: each 
>>>action in Interaction diagrams is related to a "real" implementation in 
>>>activity diagrams. What's your opinion?
>>>Marc-Philippe HUGET, PhD           MAGMA-INPG/Leibniz
>>>Laboratoire Leibniz
>>>46, Avenue Felix Viallet
>>>Modeling mailing list
>>Renato Levy
>>Principal Scientist
>>Intelligent Automation, Inc.
>>phone: +1 (301) 294-5241
>>fax:      +1 (301) 294-5201
>>Modeling mailing list

Renato Levy
Principal Scientist
Intelligent Automation, Inc.
phone: +1 (301) 294-5241
fax:      +1 (301) 294-5201