Robust Monitoring of Medical Cyber-Physical Systems
Jessica Aline Schmidt
Medical cyber-physical systems (MCPS) have become indispensable in modern medicine. They monitor patients’ endogenous data, warn them if abnormalities are found and might even decide on a treatment. An essential requirement of such health-critical systems is robustness, a metric that small changes in the input only cause slight changes in the output. In particular, measurement errors should influence the output only slightly and not lead to wrong decisions of the monitor. Using RTLola as a monitoring language, the thesis present an algorithm to analyze the robustness of an RTLola specification, which is used to generate the monitor. The algorithms are based on symbolic execution and constraint programming. They calculate the maximum output difference when the input is slightly mutated. Furthermore, we will interpret the outputs of the algorithms and evaluate them with respect to their practical application.