Stochastic approximation algorithm with application to event-triggered filtering

This paper is focused on the problem of event-triggered filtering which has various applications such as sensor networks and data sampling/acquisition. An estimator may get a “sparse” sequence of observations: the observations may arrive only when some events trigger the sensor. In this paper, a series of stopping times is used to model the times when the sensors are triggered. Based on this model, a filtering problem is formulated as to estimate the true state of the dynamic system using the information from both the new observations and their corresponding stopping times. This filtering problem is numerically solved by a stochastic approximation algorithm which uses a Markov chain to approximate the evolution of the system.