Practical Automated Partial Verification of Multi-Paradigm Real-Time Models

This article introduces a fully automated verification technique that permits to analyze real-time systems described using a continuous notion of time and a mixture of operational (i.e., automata-based) and descriptive (i.e., logic-based) formalisms. The technique relies on the reduction, under reasonable assumptions, of the continuous-time verification problem to its discrete-time counterpart. This reconciles in a viable and effective way the dense/discrete and operational/descriptive dichotomies that are often encountered in practice when it comes to specifying and analyzing complex critical systems. The article investigates the applicability of the technique through a significant example centered on a communication protocol. More precisely, concurrent runs of the protocol are formalized by parallel instances of a Timed Automaton, while the synchronization rules between these instances are specified through Metric Temporal Logic formulas, thus creating a multi-paradigm model. Verification tests run on this model using a bounded validity checker implementing the technique show consistent results and interesting performances.

💡 Research Summary

The paper tackles a long‑standing challenge in real‑time system verification: how to automatically verify models that are expressed in dense (continuous) time while also allowing a mixture of operational formalisms (such as Timed Automata, TA) and descriptive formalisms (Metric Temporal Logic, MTL). The authors propose a “partial verification” technique that reduces the dense‑time verification problem to a discrete‑time one under a set of reasonable assumptions, thereby reconciling the dense/discrete and operational/descriptive dichotomies that often arise in practice.

Key ideas and assumptions
The approach rests on the notion of δ‑non‑Berkeley behaviors. A behavior is δ‑non‑Berkeley if, for any instant of time, there exists a closed interval of length δ that contains the instant and on which the system’s state does not change. This bounded‑variability assumption guarantees that the system does not exhibit arbitrarily fast switching, which would otherwise break any discretization. Under this assumption the authors can safely replace continuous time with a uniform sampling grid of step δ.

From TA to MTL (axiomatization)
The operational part of the model—Timed Automata—is translated into a set of MTL formulas. The translation captures locations, clock constraints, resets, and invariants. The authors note that a naïve translation would lead to overly coarse approximations after discretization, so they redesign the axioms to be “discretization‑friendly”. The resulting axiomatization consists of two families of formulas:

1. Under‑approximation (safe) – if the discrete‑time MTL model is satisfied, then the original dense‑time model is guaranteed to satisfy the property. This direction is used for safety verification.
2. Over‑approximation (complete) – if the discrete‑time model violates the property, then the dense‑time model also violates it. This direction is used to obtain completeness when possible.

Both families are derived from the same underlying TA‑to‑MTL translation but differ in how interval operators are handled (e.g., using ☐_