Decomposition-based monitoring alignment with RV-LTL semantics

Identify and formalise a decomposition-based monitoring strategy that splits any LTL property into LTL3-monitorable and non-monitorable components and combines the component verdicts into a four-valued output, and determine whether there exists a configuration of this approach whose verdicts coincide with RV-LTL semantics for all finite traces.

Background

RV-LTL augments LTL3 with presumptive verdicts (true/false presumptively) based on Pnueli’s finite-trace LTL when definitive LTL3 verdicts are unavailable, yielding four truth values. The authors note sensitivities and limitations of RV-LTL related to where an observation ends.

They propose an alternative path: decompose properties into LTL3-monitorable and non-monitorable components and combine the monitored answers. They conjecture that some configuration of this decomposition and combination could reproduce RV-LTL’s semantics.

References

We conjecture that there will be some configuration of this approach whose answers coincide with RV-LTL, but it will be interesting future work to explore the design space here.

— Semantics for Linear-time Temporal Logic with Finite Observations  (2411.14581 - Amjad et al., 2024) in Section 6.3 (RV-LTL)