A Generic Approach to Quantitative Verification

TL;DR

This thesis develops a comprehensive, distance-based framework for quantitative verification of systems, enabling robust, compositional analysis independent of specific distance measures, and extends to quantitative specifications.

Contribution

It introduces a general, distance-agnostic theory of quantitative verification using game-based distances and extends to quantitative modal transition systems.

Findings

Defines various distances like bisimulation and simulation distances.

Constructs a quantitative generalization of the linear-time--branching-time spectrum.

Extends the theory to quantitative behavioral specifications.

Abstract

This thesis is concerned with quantitative verification, that is, the verification of quantitative properties of quantitative systems. These systems are found in numerous applications, and their quantitative verification is important, but also rather challenging. In particular, given that most systems found in applications are rather big, compositionality and incrementality of verification methods are essential. In order to ensure robustness of verification, we replace the Boolean yes-no answers of standard verification with distances. Depending on the application context, many different types of distances are being employed in quantitative verification. Consequently, there is a need for a general theory of system distances which abstracts away from the concrete distances and develops quantitative verification at a level independent of the distance. It is our view that in a theory of quantitative verification, the quantitative aspects should be treated just as much as input to a verification problem as the qualitative aspects are. In this work we develop such a general theory of quantitative verification. We assume as input a distance between traces, or executions, and then employ the theory of games with quantitative objectives to define distances between quantitative systems. Different versions of the quantitative bisimulation game give rise to different types of distances, viz.~bisimulation distance, simulation distance, trace equivalence distance, etc., enabling us to construct a quantitative generalization of van Glabbeek's linear-time--branching-time spectrum. We also extend our general theory of quantitative verification to a theory of quantitative specifications. For this we use modal transition systems, and we develop the quantitative properties of the usual operators for behavioral specification theories.