Quantitative Assurance and Synthesis of Controllers from Activity Diagrams
Kangfeng Ye,Fang Yan,Simos Gerasimou
2024-03-01
Abstract:Probabilistic model checking is a widely used formal verification technique to automatically verify qualitative and quantitative properties for probabilistic models. However, capturing such systems, writing corresponding properties, and verifying them require domain knowledge. This makes it not accessible for researchers and engineers who may not have the required knowledge. Previous studies have extended UML activity diagrams (ADs), developed transformations, and implemented accompanying tools for automation. The research, however, is incomprehensive and not fully open, which makes it hard to be evaluated, extended, adapted, and accessed. In this paper, we propose a comprehensive verification framework for ADs, including a new profile for probability, time, and quality annotations, a semantics interpretation of ADs in three Markov models, and a set of transformation rules from activity diagrams to the PRISM language, supported by PRISM and Storm. Most importantly, we developed algorithms for transformation and implemented them in a tool, called QASCAD, using model-based techniques, for fully automated verification. We evaluated one case study where multiple robots are used for delivery in a hospital and further evaluated six other examples from the literature. With all these together, this work makes noteworthy contributions to the verification of ADs by improving evaluation, extensibility, adaptability, and accessibility.
Logic in Computer Science,Formal Languages and Automata Theory,Software Engineering