Transmission scheduling for multi-process multi-sensor remote estimation via approximate dynamic programming (Q2063834)

The authors consider multiple sensors multiple processes transmission scheduling problem where a set of N sensors communicate, may be simultaneously, to the estimator through a noisy wireless network. The purpose is to minimize the expected sum of the traces of the error covariances across the sensors. The authors formulate the considered problem as a time-homogeneous Markov decision processes with the associated stochastic dynamic programming framework. It is supposed that each sensor is equipped with a local Kalman filter and the remote estimator can decode multiple sensor transmissions while a received signal-to-interference-and-noise ratios are above a certain threshold. Enhanced-exploration greedy least squares temporal difference algorithm is proposed for the model to approximate the objective function. The convergence properties of the algorithm with the errors estimations are proved. The work of the algorithm is demonstrated on numerical examples.