Dynamic Event/Fault Tree Analysis of Multi-agent Systems Using Galileo

Agent-oriented software engineering (AOSE) has the potential to revolutionize critical software systems by employing autonomy, adaptability, and robustness mechanisms advantageous to essential infrastructure. Yet, requisite with critical software is the need for dependability techniques to provide high assurance that the multi-agent system (MAS) will function as intended. In this paper we investigate how dynamic event/fault trees (DEFT) can be utilized for AOSE as a mechanism to analyze and reason about agent behavior and failures. Specifically, we: (1) adapt DEFT for MAS to identify and analyze necessary safety responsibilities; (2) show how the use of Galileo, the tool for DEFT probabilistic risk assessment, assists in developing a more robust MAS; and, (3) identify necessary AOSE-specific enhancements for DEFT needed to better support the dependability analysis of MAS. We illustrate this work through an application to a proposed NASA agent-based satellite swarm.