Robust Multi-Agent Systems: The Transactional Conversation Approach?

Before industry can adopt Multi-Agent System (MAS) critical issues like safety and security must be solved. The first part of this paper introduces an implementation architecture that serves as a framework to discuss the various aspects of MAS safety, and to classify the many heterogeneous approaches in this field. In the second part of the paper we present our own approach to increase MAS safety. Its key idea is to use transaction-based robustness mechanisms known from database management systems and to incorporate them into a robust FIPA-compliant MAS development framework. In particular, agent conversations are treated as distributed transactions, leading to the so-called transactional conversations. We discuss, how an appropriate transactional semantics can be derived from the formal semantics of the conversations’ speech acts.