|
Project: Distributed Design
|
Project: AI and Distributed Design
Keywords:
distributed design, collaborative design, computer supported, design review
Last updated: September 1st, 2003
Introduction
The Intelligent Interactive Distributed Systems (IIDS) group aims to
support the development of flexible, adaptable architectures for
intelligent interactive distributed systems. The group designs and
builds the AgentScape middleware for large scale agent systems.
Design, and especially distributed design, is a typical application of
these dynamic agent systems, involving a highly interactive, heterogeneous
agent population in which co-operation is of great importance. The focus of
the design research at IIDS is on distributed design of dynamic artefacts,
with an emphasis on management of co-ordination processes, researching both
theory and implementation. The IIDS group is located at the Faculty of
Sciences, Vrije Universiteit Amsterdam, and combines expertise from Artifical
Intelligence and Computer Systems (http://www.iids.org).
Design
The IIDS group builds on a decade of theoretical and prototypical
research in AI & Design. The Generic Design Model, introduced in 1994,
is based on the premise that design involves exploration and
reflection. Within this model, manipulation of the design object is
explicitly separated from manipulation of sets of qualified
requirements and co-ordination of the overall design process: a
structure which facilitates acquisition and explicitation of design
domain and design process knowledge, including strategic knowledge, design
rationale and conflict management. The Generic Design Model has been
formalised and tested in a number of domains, including aircart
emergency exit design, environmental inventory model design, aircraft
toilet unit design, and elevator configuration, yielding a number of
prototype implementations.
By combining the Generic Design Model with the Generic Agent Model, a
prototype of a generic design agent (2001a) resulted: a basis for
studying distributed design. The design agent model explicitly
distinguishes communication, service interactions, and self-management of
its design capabilities, providing a structure to model aspects
of (human) designers involved in co-operative design tasks.
Design and the Internet
The Internet and related supportive technologies, such as AgentScape
and the Semantic Web, form a dynamic environment in which multitudes
of software agents and web services appear, disappear, roam, interact
and co-operate. This ever-changing environment requires agents and
web services to be flexible and adaptive; at this scale an unmanagable
process for humans. The agent factory and software configuration
services are researched by our group to perform unsupervised automated
re-design of agents (2001b) and (re-)configuration of web services
(2003); an application domain involving the study of dynamic
artefacts which are designed to be re-designed.
In this approach, agents and web services each have their own
blueprint, in which both their conceptual model and operational
architecture are described as a configuration of (conceptual or
operational) components. Adaptation of an agent or web service
entails adaptation of their blueprint, on the basis of needs for
adaptation provided by the agent, the web service, or other entities.
The adaptation process involves a re-design process separate from an
assembly process.
Automated adaptation of software agents offers side-benefits which
greatly influence the design and development of Internet applications:
mobile agents may be dynamically optimised to run on a new location
(including changing an agent's code from, e.g., Java to Python);
agents can be regularly sent for a 'check-up'; agents can be re-built
using trusted software components improving security and reducing the
danger of virusses; and novel solutions and creativity may emerge.
Distributed Design
One of the greatest benifits of the Internet concerns the ease of
communication between (human or automated) agents: a natural
environment to investigate automated support of design teams (both
human and automated) on the basis of an understanding of distributed
design processes. Distributed design involves many participants, each
with their own expertise, experience, and goals, requiring
participants to deliberate about co-ordination (1997), trust,
reflection (2001c), and the design process at hand. As such, the area
of distributed design provides a research setting in which results
from studies of teams of human designers can be applied to, and
structured by, models of automated design agents. Possible benefits
include a better insight in intangible aspects such as creativity and
novelty.
Our research on distributed design focusses on the application of
distributed design to dynamic artefacts, which not only include agents
and web services, but also agent platforms (i.e., the middleware which
supports agents, web services, and their interactions). Of special
interest, related to security aspects, is the role of trust in
distributed design. Each designer has his or her own view of the
world and other agents, and their environments, including assessments
of their expertise, reliability, experience, et cetera. Information
acquired from different participants may be valued differently in
terms of accuracy and trustworthiness, depending on the context in
which it is acquired. While often implicit, this knowledge does
influence distributed design processes, as it determines the way in which
members of a design team assess and incorporate each others’ designs,
objectives, and evaluations (e.g., see this trace on the role of trust).
These trust relations need to be made
explicit to be able to incorporate them into design agents.
Design Challenges
The current overall progress of Design Research is promising, yet
needs more effort dedicated to the research areas of distributed
design and unsupervised automated design. The Internet turns out to be
both a suitable domain to demonstrate theories and investigate
emergent effects, and to be a real-life 'customer' in desperate need
of support from the Design Community.
References
-
Brazier F.M.T., Jonker C.M., Treur J. (1997) Formalization of a
cooperation model based on joint intentions in Mueller J.P.,
Wooldridge M.J., Jennings N.R. (eds) Intelligent Agents III,
Proceedings of the Third International Workshop on Agent Theories,
Architectures and Languages (ATAL'96), Lecture Notes in AI, Vol.
1193, pp. 141-155
-
Brazier F.M.T., Jonker C.M., Treur J., Wijngaards, N.J.E. (2001a)
Compositional Design of a Generic Design Agent in Design Studies
journal, 22(5):439-471
-
Brazier F.M.T., Jonker C.M., Treur J., Wijngaards N.J.E. (2001b)
Deliberative Evolution in Multi-Agent Systems in International
Journal of Software Engineering and Knowledge Engineering,
11(5):559-581
-
Brazier F.M.T., Langen P.H.G. van, Ruttkay Z.S., Treur J. (1994) On
formal specification of design tasks in Gero J.S., Sudweeks F.
(eds), Proceedings Artificial Intelligence in Design (AID'94),
Kluwer Academic Publishers
-
Brazier F.M.T., Moshkina L.V., Wijngaards N.J.E. (2001c) Knowledge
level model of an individual designer agent in collaborative
distributed design, Journal of Artificial Intelligence in
Engineering, 15(2):137-152
-
Splunter S. van, Sabou M., Brazier F.M.T., Richards D. (2003)
Configuring Web Services using Structurings and Techniques from
Agent Configuration in Proceedings of the International Conference
on Web Intelligence (WI 2003), (to appear), October
Page last modified: September 1st, 2003
