A Case Study on Runtime Monitoring of an Autonomous Research Vehicle (ARV) System

Runtime monitoring is a versatile technique for detecting property violations in safety-critical (SC) systems. Although instrumentation of the system under monitoring is a common approach for obtaining the events relevant for checking the desired properties, the current trend of using black-box commercial-off-the-shelf components in SC system development makes these systems unamenable to instrumentation. In this paper we develop an online runtime monitoring approach targeting an autonomous research vehicle (ARV) system and recount our experience with it. To avoid instrumentation we passively monitor the target system by generating atomic propositions from the observed network state. We then develop an efficient runtime monitoring algorithm, EgMon, that eagerly checks for violations of desired properties written in future-bounded, propositional metric temporal logic. We show the efficacy of EgMon by implementing and empirically evaluating it against logs obtained from the testing of an ARV system. EgMon was able to detect violations of several safety requirements.