Enhancing UPPAAL’s Explanatory Power using Static Zeno Run Analysis

This thesis discusses Zeno runs and their detection in timed automata networks. Specifically, we explore Zeno runs in the context of model checking with the tool UPPAAL. Zeno runs are transition sequences in the system that can execute arbitrarily fast, which conflicts with the real world experience where execution always takes time. Accordingly, most of the time the presence of Zeno runs in a model is unintentional. Also, Zeno runs can influence the model behavior in a negative way, for example timelocks, an analogy to deadlocks regarding time, may occur due to them, and therefore their detection is a desirable goal. Gómez applied and extended Tripakis’ strong non-Zenoness property concept to detect Zeno runs in timed automata systems. In this thesis we enhance Gómez’ method by eliminating false positives that the approach yields in some circumstances. For this purpose we modify the way synchronization in the system is exploited by not only ensuring that valid synchronization partners exist but also that their amount is correct. Additionally, we incorporate two data variable heuristics because in most models certain Zeno runs can not occur due to constraints on data variables. We implemented Gómez’ algorithm and our extensions in an analysis tool named ZenoTool to validate the theoretical improvements in the detection accuracy due to our enhancements and also measured its run-time performance. We applied our tool to several real world case studies and the experiments show that the static analysis approach generally performs well. Positive evidence was found for our theoretical assumptions on the accuracy improvements and the tool’s run time appears to be no issue. Thus, we conclude that static Zeno run detection can effectively contribute to ensure safety of timed automata network specifications and therefore enhances the explanatory power of model checkers like UPPAAL.