A decorrelation method in asynchronous arithmetic average fusion of GM-PHD filters

We present a decorrelation arithmetic average (AA) fusion algorithm of Gaussian mixture probability hypothesis density (GM-PHD) filters to ameliorate the multi-target tracking accuracy in asynchronous scenarios. First, the correlations in single-target asynchronous scenarios and the Bayesian optimal decorrelation fusion method are derived. Then, the derived single-target decorrelation fusion is employed to merge estimates of the same target. As required by the derived decorrelation method, a measurement extraction technique is developed to acquire measurements contained in locally filtered estimates, and a hierarchical structure involving a master filter is designed to provide prior estimates automatically. Simulations verify that our algorithm inherits the derived single-target fusion’s Bayesian optimality in handling delays. Meanwhile, extending our algorithm to track multiple maneuvering targets also exhibits certain potential.