Co-Ordination in Artificial Agent Societies : Social Structures and Its Implications for Autonomous Problem-Solving Agents. Dieser Artikel ist nicht verfügbar.Co-Ordination in Artificial Agent Societies : Social Structures and Its Implications for Autonomous Problem-Solving Agents

Advances in Computer Science often arise from new ideas and concepts, that prove to be advantageous for the design of complex software systems. The con ception of multi agent systems is particularly attractive, as it prommodul ises arity based on the conceptual speciality of an agent, as well as flexibility in their inte gration through appropriate interaction models. While early systems drew upon co operative agents, recent developments have realised the importance of the notion of autonomy in the design of agent based applications. The emergence of systems of autonomous problem solving agents paves the way for complex Artificial Intelligence applications that allow fosca r lability and at the same time foster the reusability of their components. In consequence, an intelligent multi agent application can be seen as a collec tion of autonomous agents, usually specialised in different tasks, together with a social model of their interactions. This approach implies a dynamic generation of complex relational structures, that agents need to be knowledgeable of in order to successfully achieve their goals. Therefore, a multi agent system designer needs to think carefully about conceptualisation, representation and enactment of the different types of knowledge that its agents rely on, for individual problem solving as well as for mutual co ordination.

This monograph provides a comprehensive survey of the different approaches to coordination in societies of artificial an human agents. Setting out from a critical assessment of the state of the art, the author develops a method of structuring multi-agent applications with a mechanism called structural cooperation. Agents are equipped with expertise about their environment in order to detect and overcome specific types of problem, they make use of their social knowledge to mutually adjust their activities, and they are coerced toward coherent collective behavior through normative rules. The proposed model is formalized theoretically within game theory and realized by means of an agent architecture. It is assessed experimentally by building a prototype of a distributed decision support system for road traffic management and compared to an alternative model based on a centralized architecture.
A valuable feature of the work is that it not only promotes a well-founded formal model of coordination in artificial agent societies but also applies it in an operational software architecture organized as a society of intelligent agents to solve real-world problems.