Proceedings Abstracts of the Twenty-Third International Joint Conference on Artificial Intelligence

Behavioral Diagnosis of LTL Specifications at Operator Level / 1053
Ingo Pill, Thomas Quaritsch

Product defects and rework efforts due to flawed specifications represent major issues for a project's performance, so that there is a high motivation for providing effective means that assist designers in assessing and ensuring a specification's quality. Recent research in the context of formal specifications, e.g. on coverage and vacuity, offers important means to tackle related issues. In the currently underrepresented research direction of diagnostic reasoning on a specification, we propose a scenario-based diagnosis at a specification's operator level using weak or strong fault models. Drawing on efficient SAT encodings, we show in this paper how to achieve that effectively for specifications in LTL. Our experimental results illustrate our approach's validity and attractiveness.