Distributed Kalman Filter with minimum-time covariance computation

This paper considerably improves the well-known Distributed Kalman Filter (DKF) algorithm by Olfati-Saber (2007) by introducing a novel decentralised consensus value computation scheme, using only local observations of sensors. It has been shown that the state estimates obtained in [8] and [9] approaches those of the Central Kalman Filter (CKF) asymptotically. However, the convergence to the CKF can sometimes be too slow. This paper proposes an algorithm that enables every node in a sensor network to compute the global average consensus matrix of measurement noise covariance in minimum time without accessing global information. Compared with the algorithm in [8], our theoretical analysis and simulation results show that the new algorithm can offer improved performance in terms of time taken for the state estimates to converge to that of the CKF.