Next: Knowledge Acquisition Up: References Previous: Experts of the

Distributed Artificial Intelligence

Distributed Artificial Intelligence (DAI) systems can be defined as cooperative systems where a set of agents act together to solve a given problem. These agents are often heterogeneous (e.g., in Decision Support System, the interaction takes place between a human and an artificial problem solver).

Its metaphor of intelligence is based upon social behaviour (as opposed to the metaphor of individual human behavior in classical AI) and its emphasis is on actions and interactions, complementing knowledge representation and inference methods in classical AI.

This approach is well suited to face and solve large and complex problems, characterized by physically distributed reasoning, knowledge and data managing. In DAI, there is no universal definition of ``agent'', but Ferber's definition is quite appropriate for drawing a clear image of an agent: "An agent is a real or virtual entity which is emerged in an environment where it can take some actions, which is able to perceive and represent partially this environment, which is able to communicate with the other agents and which possesses an autonomous behaviour that is a consequence of its observations, its knowledge and its interactions with the other agents".

DAI systems are based on different technologies like, e.g., distributed expert systems, planning systems or blackboard systems. What is now new in the DAI community is the need for methodology for helping in the development and the maintenance of DAI systems. Part of the solution relies on the use of more abstract formalisms for representing essential DAI properties (in fact, in the software engineering community, the same problem led to the definition of specification languages).

Different languages are now advocated for describing DAI systems at a conceptual level. As such, DAI systems are therefore recognized as composite systems. A common feature of all these languages relies on the introduction of agents. Important aspects covered by researchers are the type of knowledge that agents has (about the domain, the self and the others, believes vs confidence, etc) and the nature of negotiations taking place between agents (arbitrator, reviewer, proposer, etc.) [Gas88] [Hew91] [Wer91]. Other important topics include conflict resolution, unified negotiation protocols and applications to various applications areas including among many others constraint satisfaction, planning and diagnosis.

Several comprehensive collections of papers [GH89][Huh87][Bond88b] and surveys [Dur89][LC87][Dec87] have been published on the subject. DAI may indeed be the focus point within ModelAge in providing a lot of formalisms, architectures and metaphors for other areas. It is our view that many of the results of DAI can be successfully applied in the other areas covered within ModelAge.

The investigation of these possible applications will be a major point for the DAI researchers in ModelAge, while on the other hand the analysis of application areas not traditionally covered in DAI (such as requirements engineering or federated data bases) are expected to lead to new insights into the question what kind of multi-agent models are required in more conventional applications.

Next: Knowledge Acquisition Up: References Previous: Experts of the

Pierre-Yves Schobbens
Fri Nov 25 10:58:38 WET 1994