Agent research within the Information Systems group of the Eindhoven University 
of Technology

People involved:
Prof. dr. P.M.E. De Bra, 
dr. ir. G.J. Houben,
dr. F. Dignum,
dr. A. Aerts
e-mail: {debra,houben,dignum,wsinatma}@win.tue.nl

Within the Information Systems group at the Eindhoven University of Technology
the following two agent projects are executed.


1. Task based Information Filtering (TIF):
  Finding Information that is Right for the Job


The TIF project aims at improving the quality of information retrieval
tools in professional situations, where the task being performed by the
user is an important factor in deciding whether documents are relevant
and appropriate or not. In production workflow systems the task may
uniquely determine the needed information, but in less strictly defined
processes there is a need for information agents or brokers which take
the task, (a model of) the user, and the possible content-based searches
on the information base into account. The TIF project aims at developing
such an architecture consisting of task-agents, user-agents and content-agents.

The task agent is normally part of, or linked to a Workflow engine.
Based on the specific task a user performs, in the entire workflow chain,
it can give helpful hints to the user's agent, as to which elements to
include in queries to the content agents, and also hints as to which content
agents to contact first.
The user agent learns about user-preferences, regarding media types,
level of sophistication and user skills, by monitoring how the user selects
or rejects suggested documents. It asks questions to one or more content
agents, based on its knowledge of which agent knows about which subject,
and based on the helpful hints from the task agent.
The content agents are information brokers that know about certain
subjects, and that actively search for new information on their topic.
An example of this active part of a content agent is the FishNet system,
developed at the TUE.

The focus of the TIF project is not on developing new types of agents,
but on the architecture that enables the agents to work together in
finding the information the user needs for completing a task.
The TIF prototypes (to be built) will be compared
to pure content-based search engines, and to the content and user based
brokers such as Pamela,
in order to be able to identify which aspects of task-based information is
most helpful in improving retrieval performance.
This evaluation will be performed in close collaboration with the TUE IPO
institute, which specializes in user-centered design and in human-computer
interaction.

The task of the professional user often involves the generation of new
information (and documents). These documents need to be stored and managed
in such a way that the content agents can easily find them, and storing
the documents must be communicated to the task agents as well, in order to
enable the generation of the appropriate search hints to find the documents
for (possibly different) subsequent tasks. The combination of document
repository, content classification by the content-agents and task classification
by the task agents turns the TIF system into a knowledge management system.

2. Mobile Agents Framework - project

In the MAF-project mobile agent architectures are researched with a focus on 
infrastructures for distributed environments that make it possible to provide 
support for business processes that span multiple locations. Mobile agents 
posses the capability to keep functioning in environments in which components 
and communication links may disappear (by going off-line or because of failure)
or become available. In such environments it is never certain which locations 
an agent has to visit to acquire the services needed to carry out its task. 

The present version of the framework provides transportation, authentication 
and trading (yellow pages) services. Current research is focussed on 
reliability of transportation, security and intelligence in trading.

The Mobile Agent Framework is a joint project of Baan Labs and the Computing 
Science Department of the Eindhoven University of Technology, that started in 
1996 and so far has involved prof. dr. ing. D.K. Hammer, dr. A.T.M. Aerts, 
ir. M. Dalmeijer, ir. E. Rietjens, ir. M. Soede, ir. R. Caelers, 
ir. K.J. Dijkzeul.