Effective Sensor Scheduling Schemes in a Sensor Network by Employing Feedback in the Communication Loop

In this paper, we consider a state estimation problem over a bandwidth limited network. A sensor network consisting of N sensors is used to observe the states of M plants, but only p ≤ N sensors can transmit their measurements to a centralized estimator at each time. Therefore a suitable scheme that schedules the proper sensors to access the network at each time so that the total estimation error is minimized is required. We propose four different sensor scheduling schemes. The static and stochastic schemes assume no feedback from the estimator to the scheduler, while the two dynamic schemes, Maximum Error First (MEF) and Maximum Deduction First (MDF) assume such feedback is available. We compare the four schemes via some examples and show MEF and MDF schemes perform better than the static and stochastic schemes, which demonstrates that feedback can play an important role in this remote state estimation problem. We also show that MDF performs better than MEF as MDF considers the total estimation error while MEF considers the individual estimation error.