A study of estimation and communication tradeoff using an event-based approach

In this paper, we consider estimating the state of a linear time-invariant system over a network subject to limited sensor communications. A sensor locally computes the state estimate for the system from its observations and send it to a remote estimator under the constraint that the total transmission times are no more than a pre-specified value. The sensor needs to decide when to send the local estimate in order to minimize the average estimation error covariance at the remote estimator. Offline scheduling and online scheduling policies are two typical solutions. The main contribution of this paper is that we propose a novel form of hybrid scheduling policies, which combine the two conventional ones and demonstrate that the estimator performance is improved when compared with the optimal offline schedule while the computation complexity is reduced when compared with the optimal online schedule. Therefore, the proposed schedules provide a trade-off between the two classic approaches in terms of estimation quality and computation complexity.