Self-calibrating Monitoring of Hybrid Systems

Physical systems are by nature continuous, but their behaviors are often abstracted as a sequence of continuous behaviors over time intervals and discontinuous changes (hybrid systems). Model-based monitoring relies on a comparison between the predicted behavior of a model and the observed behavior of a physical system. A drawback of many monitoring approaches is the use of a single precise model. However, even if the system is behaving properly, precise parameter values and functional relationships are often not known. In our approach, the monitoring system starts with a coarse description of the physical systems and uses measurements to refine the behavior prediction, its underlying model and the time uncertainty of any discontinuous transition. This self-calibrating monitoring system is grounded on the framework of qualitative and semi-quantitative simulation and facilitates the treatment of incomplete and imprecise knowledge of the physical system as well as noisy measurements.