System Identification with Non-Uniformly Sampled and Intermittent Data in the Internet of Things

This paper describes a numerical approach to the system identification problem when the available time series data is non-uniformly sampled and intermittent in time. This includes instances where short packets of data are available at irregular intervals, e.g., data transmitted via the internet for cyber-physical systems or IoT, or when data over certain time intervals is missing or unreliable. The novel approach is a matrix-based discretization of the state-space equations, whereby the nature of the intermittent data leads to a natural matrix partitioning of the available and unavailable data. The method is tested with a real-world example of a pendulum on a cart, where the intermittent data effect is simulated so that the results can be compared with the ground-truth full-data case.