A Fault Injection Framework to Support STPA-Driven Hazard Assessment in a UAM Simulator

摘要

In this paper, we present a framework for creating and performing fault injection campaigns within an urban air mobility (UAM). Fault injection is a technique by which a process or data stream is intentionally altered with the aim of exposing some hazard in a system. Our process begins with System-Theoretic Process Analysis (STPA), passes the uncovered causes of failures for injection into communication streams as faults, and stores the data collected from the system. The fault injection framework presented here leverages and integrates with prior work on a simulation platform for activities supporting validation of runtime monitoring efforts in the UAM setting. Robot Operating System 2 (ROS 2) is the communication layer in the simulation testbed at which faults are introduced to the communication streams that are the inputs to software monitors. Results from two fault injection experiments are presented that demonstrate the ability of our framework to provide insight into monitors’ behavior and the boundaries of their correct operation.