Behavioral and Temporal Rule Checking for Gaussian Random Process a Kalman Filter Example
Doron Drusinsky (Naval Postgraduate School, USA)
Abstract: This paper describes a behavioral and temporal pattern detection technique for state-space systems whose state is a random variable such as the state estimated using a Kalman filter. Our novel behavioral and temporal pattern detection technique uses diagrammatic, intuitive, yet formal specifications based on a dialect of the UML of the kind used to monitor or formally verify the correctness of deterministic systems. Combining these formal specifications with a special code generator, extends the deterministic pattern detection technique to the domain of stochastic processes.
We demonstrate the technique using a Ballistic trajectory Kalman filter tracking example in which a pattern-rule of interest is not flagged when observing the sequence of mean track position values but is flagged with a reasonable probability using the proposed technique.
Keywords: Kalman Filter, UML, monitoring, patterns, random process, statecharts
Categories: D.2.4, F.1.1, F.4.1