Squared-root cubature information consensus filter for non-linear decentralised state estimation in sensor networks
Distributed analysis of target kinematics captured by a large network of sensors has received significant attention lately. Tracking moving targets in nonlinear systems is one of the most fundamental tasks in this regard and information-type consensus filters (ICFs) have been applied to this problem. To improve the estimate performance, a squared-root cubature information filter which can avoid numerically sensitive matrix operations such as matrix square-rooting and inversion has been developed firstly. And then, based on this filter, a decentralised information filtering algorithm is proposed in an improved consensus framework. Specifically, consensus update at each time-cycle in the modified consensus scheme is computed in two steps, first towards the predicted value and then towards the final information estimate update, which can improve average estimation accuracy and speed the average consensus. Besides the basic merits of the traditional ICFs, the resulting algorithm is more scalable and robust. Simulation results clearly show the advantage of the proposed algorithm compared with existing ICFs in the considered application scenario.