Dependable multi-agent systems: Layered reference architecture and representative mechanisms

Abstract

Layered architectures are a proven principle for the design of software systems and components. The paper introduces a layered reference architecture for multi-agent systems which assigns each agent property to select layers. It demonstrates how the same reference architecture provides a framework for a dependability model that associates the sources of failures and the ensuing error handling with a specific layer, thus integrating dependability directly into the design of agents. The architectural approach is illustrated by several dependability mechanisms that assume transactional conversations. © 2009 Springer Berlin Heidelberg.