Stochastic online sensor scheduler for remote state estimation

In this paper, a remote state estimation problem where a sensor measures the state of a linear discrete-time process in an infinite time horizon is considered. We aim to minimize the average estimation error subject to a limited sensor-estimator communication rate. We propose a stochastic online sensor schedule: whether or not the sensor sends data is based on its measurements and a stochastic holding time between the present and the most recent sensor-estimator communication instance. This decision process is formulated as a generalized geometric programming (GGP) optimization problem. It can be solved with a tractable computational complexity and provides a better performance compared with the optimal offline schedule. Numerical example is provided to illustrate main results.