Component-Based Design of Intelligent Multi-Agent Systems
In today's complex business processes, most often information, knowledge and other resources are inherently distributed. Distributed factory and office automation, computer supported co-operative work, intelligent information management, large scale transaction processing, and wide-area networking mandate flexible, distributed and scaleable IT support. Interaction and decentralised control are of crucial importance to new emerging phenomena such as Electronic Commerce, in which agent technology, the World Wide Web and Internet are indispensable ingredients.
Agents are autonomous, conceptual entities capable of reactively and/or pro-actively influencing their dynamic environment. This environment almost necessarily includes other agents.
The component-based system design method DESIRE explicitly models such autonomous entities, their environment, and their interaction, during different phases of design. Different types of reasoning (such as strategic and reflective reasoning) and different forms of co-operation are modelled and specified from the start.
In DESIRE, multi-agent systems are modelled and specified as component-based or compositional systems: agents are modelled as interacting complex components. In this course, the constructs needed to model flexible, adaptive behaviour are introduced and applied to a number of case studies.
DESIRE adheres to many of the principles of modern approaches to component-based system design. Information hiding, for example, is of utmost importance in the design of compositional architectures: the level of abstraction at which a component is defined determines which information is 'visible' to other components at that level. Explicit (declarative) representation of control knowledge within components is expressed in terms of the knowledge that is 'known' at the related level within a composition. This control knowledge can specify both parallel and sequential activation of components and information flow, and provides a means to model event-driven and decentralised control.
Reuse of generic models and instantiated components is inherent to the modelling approach. Generic models for different types of agents and tasks provide a means to structure system design. Such models support task analysis and conceptual modelling. Domain specific knowledge is added to instantiate generic components; this knowledge may be represented in many ways, depending on the application at hand, and reused when appropriate.
In this course, participants learn to construct transparent, structured conceptual designs of multi-agent systems, to specify conceptual and detailed designs, and to transform specifications into implementations. In addition, participants learn to use graphical tools designed to support modelling, specification, implementation and testing of multi-agent systems.
Participants are expected to be interested in the design and development of multi-agent systems, and to have had some experience in the development of information systems and/or knowledge-based systems.
DESIRE is the result of ten years of research by the Agent Systems Group of the Department of Artificial Intelligence at the Vrije Universiteit Amsterdam in interaction with industry. It has been tested in various domains of application, such as distributed project co-ordination and agenda scheduling, co-operative information acquisition, chemical process control, negotiation for energy load balancing, distributed diagnosis and repair tasks, and design. The 10 members of the Agent Systems Group still focus their research on further development of the DESIRE method and support, its logical foundation, and applications. Dissemination of research results is one of the aims of the group. Courses are a means to this end. From 1996 the DESIRE method is the basis of a series of annual courses.
The course will be held at the Vrije Universiteit Amsterdam, The Netherlands, from Wednesday May 29 to Tuesday June 4, 2002. Course material and instructions are in English.
To acquire knowledge and skills in modelling and specifying multi-agent systems using the method DESIRE, active participation is required. For each of the five days the programme includes lectures, working sessions and computer-supported practical sessions. The main themes addressed in the programme are:
Wednesday May 29
Introduction: Basic agent
concepts and agent behaviour. Design principles, such as compositionality,
re-use, information and process hiding. Different types of reactive and
pro-active agent behaviour.
Practical session: a simple process control agent.
Thursday May 30
Compositional design: process
composition and knowledge composition. Social agent behaviour. Information
Practical session: simulation of pro-active and reactive behaviour.
Friday May 31
Interaction and dynamics.
Autonomy and control of behaviour. Generic models of agents.
Practical session: Information Broker agent.
Monday June 3
Informational and motivational
agents. Knowledge and reasoning capabilities. agents with explicit beliefs,
desires and intentions. Reflective agents.
Practical session: co-operative information gathering agents.
Tuesday June 4
Reusable models and applications.
Models for monitoring diagnosis and control. Co-operation and project
co-ordination. Negotiating agents. Societies of agents.
Practical session: negotiating agents.
The course fee is 1250 EURO, which is approximately USD 1085 as of Feb. 5,
2002. See, e.g., http://www.oanda.com
for the latest rates. The form should be returned, and the course fee received,
before April, 27 2002. A 50% reduction is available for a restricted number of
Ph.D. students. The total number of participants is limited. The Vrije
Universiteit Amsterdam reserves the right to cancel or postpone the course.
The organisation cooperates with the Dutch Research School for Information and Knowledge systems (SIKS). The Free University Amsterdam is the administrating university of this research school. The course fits into the Advanced Components stage of SIKS' educational program for Ph.D.-students. SIKS-Ph.D.-students, working in the field of Agent Technology, who are interested in taking this course, are requested to contact email@example.com for information on admission. Deadline: April 15, 2002
For registration, complete the form and return it to the address given.
If registration is cancelled before April 27 2002, the course fee may be refunded. After April 27 2002, no reimbursement is possible.
For more information, contact:dr. C.M. Jonker
Last updated: August 24, 2001Maintained by: Catholijn Jonker