Noise covariance estimation for networked linear systems under random access protocol scheduling

Abstract

This paper investigates the noise covariance estimation problem for the networked linear system under communication protocol. To reduce network burden and avoid data collisions, the data transmission from the sensor nodes to the filter is scheduled by the random access protocol (RAP), where only a sensor node is permitted to send data at each transmission instant. Within the framework, the correlations of the innovations data are utilized to construct a least-squares problem to estimate the noise covariances, where a novel estimation form of the auto-covariance from measurement data is given. Moreover, the two cases of the process noise correlated and uncorrelated with the measurement noise are both discussed. It can be proved that the estimated covariances are unbiased and asymptotically converge to the true values as the sample size increasing. Finally, the simulation results are given to demonstrate the effectiveness of the presented method.