We propose an architecture, TOMAS (Transaction Oriented Multi Agent System), as foundation for reliable multi-agent systems. TOMAS adopts the distributed nested transaction paradigm as computational model for the well known BDI (Belief, Desire, Intention) agent architecture.
In this talk, we give an overview of a simple formalism that we developed to specify the operational semantics of BDI agents, and present the nested transaction model. We then discuss TOMAS, focusing on some of the properties which have been proven with our formalism and comparing it with the conventional BDI architecture. We conclude with a short presentation of the prototype of TOMAS, developed in Java and available from the author's Web page.
Currently, he is a postgraduate student of the Department of Computer Science of the University of Melbourne. His research addresses the issues of robustness of software for Intelligent Agents. He also cooperates with Agent Oriented Software, Melbourne, on consultancy and training in agent systems.