A formal approach for identifying assurance deficits in unmanned aerial vehicle software

Abstract

While formal methods have proved to be unfeasible for large scale systems, argument-based safety cases offer a plausible alternative basis for certification of critical software. Our proposed method for increasing safety combines formal methods with argumentation-based reasoning. In a first step, we provide a formal representation of the the argumentative-based Goal Structuring Notation (GSN) standard used in industry. In a second step, our solution exploits reasoning in description logic to identify assurance deficits in the GSN model. The identified flaws are given to a hybrid logic-based model checker to be validated against a Kripke model. The method is illustrated for an unmanned aerial vehicle software, with reasoning performed in RacerPro engine and the HLMC model checker based on hybrid logic.