From: Gerhard Kraetzschmar
Subject: Request for info on DAI tools and testbeds
Date:
Message-ID: <1pf0shEaj4@uni-erlangen.de>
Hello netters and news readers!
I am in the process of doing a survey in Distributed AI
regarding two specific areas of DAI software/systems:
- tools
- testbeds
As AI people's understanding of tools and testbeds can be
quite different (experience :-)), I give a characterization of
the kind of things that in my opinion qualify for these
two areas below (after the signature).
I am very grateful for every hint I get. For both kinds of
systems please specify in every case
- the name of the system
- a contact (author, implementer, provider of further info)
and, if possible
- an abstract of its functionality
- implementation language and state (in use, in development))
- availability (commercial/proprietary/public domain)
- how to use it
- how to integrate it in DAI sw systems
- other relevant info (ftp source: host, Internet-Id, dir)
Please respond via email (do not post to this list or newsgroup) to
···@forwiss.uni-erlangen.de
I will collect the responses and mail/post a summary to all the
newsgroups/lists I posted this article to (probably around End of April).
Also, most likely a report from our research institute will be available
on the topic.
Thanks in advance for your cooperation!
So long, Gerhard
---------------------------------------------------------------------------
Gerhard K. Kraetzschmar FORWISS
Email: ············@forwiss.uni-erlangen.de Am Weichselgarten 7
Telephone: intl.+49-9131-691-193 8520 Erlangen-Tennenlohe
Telefax: intl.+49-9131-691-185 Federal Republic of Germany
---------------------------------------------------------------------------
Things that qualify for DAI tools:
----------------------------------
The general idea here is to substantially reduce the effort needed for
implementing DAI systems, e.g. by providing a framework for multiple agents
communicating via a specific protocol, etc. That is, I look for
software systems and utilities that implement specific DAI concepts,
technologies or architectures, such as (just a few examples)
- an implementation of the contract net protocol
- an implementation of a distributed TMS
- an system that allows for easy configuration and programming of
multi-agent systems with a specific architecture
- the implementation of a mechanism for enabling/ensuring cooperative
behavior between agents
- the implementation of a mechanism for resolving conflicts between
competing/non-competing agents
Note, that the software should provide significantly more support
than e.g. CommonLisp + C interface + TCP-IP + (lightweight) processes,
and one or more DAI technologies should be more or less directly
applicable.
Things that qualify for DAI testbeds:
------------------------------------
Software environments, that allow testing and evaluation of
DAI systems, such as multi-agent systems (MAS) and (cooperative,)
distributed problem solvers ((C)DPS). They must provide functionality
to define and simulate an environment for the DAI system to be evaluated in
as well as an interface to the DAI system. Certain restrictions, like
"can interface only via TCP/IP" or "good only for LISP-implemented agents"
are acceptable. The testbed must provide functionality to perform
experiments with a DAI system under varying environment conditions,
record internal data of the DAI system during experiments for later evaluation,
and enable the experimenter to repeat experiments and reconstruct
critical or interesting situations.
An essential criterion is that the testbed should allow (and actually has
been used) to test and evaluate more than one specific DAI system or
architecture. Probably there are more testbeds around that do NOT meet
this essential criterion; if this is the case, please send info,
but do clearly state that testbed X has been developed to evaluate
DAI system Y, but (in theory/with some work/easy but not done) could
be used to evaluate DAI system Z as well.