Answering Temporal Conjunctive Queries over Description Logic Ontologies for Situation Recognition in Complex Operational Domains

Abstract

AbstractFor developing safe automated systems, recognizing safety-critical situations in data from their complex operational domain is imperative. This capability is, for example, essential when evaluating the system’s conformance to specified requirements in test run data. The requirements involve a temporal dimension, as the system operates over time. Moreover, the generated data are usually relational and require additional background knowledge about the domain for correctly recognizing the situation. This fact makes propositional temporal logics, an established tool, unsuitable for the task. We address this issue by developing a tailored temporal logic to query for situations in relational data over complex domains. Our language combines mission-time linear temporal logic with conjunctive queries to access time-stamped data with background knowledge formulated in an expressive description logic. Currently, however, no tools exist for answering queries in such settings. We hence also contribute an implementation in the logic reasoner Openllet, leveraging the efficacy of well-established conjunctive query answering. Moreover, we present a benchmark generator in the setting of automated driving and demonstrate that our tool performs well when tasked with recognizing safety-critical situations in road traffic.