Distributed Widely Linear Complex Kalman Filtering

Abstract

We introduce cooperative sequential state space estimation in the domain of augmented complex statistics, whereby nodes in a network collaborate locally to estimate noncircular complex signals. For rigour, a distributed augmented (widely linear) complex Kalman filter (D-ACKF) suited to the generality of complex signals is introduced, allowing for unified treatment of both proper (rotation invariant) and improper (rotation dependent) signal distributions. Its duality with the bivariate real-valued distributed Kalman filter, along with several issues of implementation are also illuminated. The analysis and simulations show that unlike existing distributed Kalman filter solutions, the D-ACKF caters for both the improper data and the correlations between nodal observation noises, thus providing enhanced performance in real-world scenarios.