Applying Product-Line Fault Tree Analysis to Build Safer Multi-Agent Systems

Software fault tree analysis (SFTA) is a widely used technique for safety analysis of a system as it helps identify the causal relationship between failures and hazards. Yet, SFTA is difficult to apply to dynamic systems with shifting run-time configurations (eg., autonomous multi-agent systems (MAS)). This paper illustrates the application and use of a product-line approach to SFTA applied to a highly-autonomous, distributed MAS. Using this approach, engineers can evaluate safety properties and the dependability of a system early in the development lifecycle. Further, this approach utilizes a technique in which the assets produced are reusable (i.e., applicable towards all configurations of the system) with little additional work.