Specifying and Executing Open Multi-agent Systems

Abstract

An ‘open’ Multi-Agent System (MAS) is a system in which the member agents are developed by different parties and serve different, often competing interests. In open MAS, the behaviour of an agent cannot be predicted in advance (Hewitt 1991). Furthermore, an agent may choose not to conform to the MAS specification in order to achieve its individual goals, or it may fail to conform to the MAS specification due to, say, a bug in its code (Sergot 2004). Agents may also fail to behave as intended because of factors beyond their control. This is commonplace when open MAS are deployed on distributed environments with unreliable communication channels. A few examples of this type of MAS are electronic marketplaces, virtual organisations and digital media rights management applications.