A pseudo measurement-level arithmetic average fusion in asynchronous networks

Abstract

Association operation decomposes the multi-target arithmetic average (AA) fusion into multiple groups of single-target merging. The B1 fusion is highly close to the optimal fusion of delayed measurements by computing and handling correlations. Accordingly, a pseudo measurement-level AA (PML-AA) fusion algorithm of Gaussian mixture probability hypothesis density (GM-PHD) filters is proposed to ameliorate the tracking accuracy of asynchronous data by applying the superior B1 fusion as the required merging method. Since the B1 fusion belongs to measurement-level methods, a specified technique is developed to extract measurements contained in locally filtered estimates. As required by the B1 fusion and our measurement extraction technique, a master filter that only operates prediction is introduced to provide indispensable prior estimates. To accommodate this master filter, a hierarchical structure involving a master filter and several local filters is designed. Simulations demonstrate that by virtue of the superiority of the B1 method in fusing delayed data and accurate association, the proposed PML-AA fusion outperforms the existing AA fusion in tracking accuracy within various tracking scenarios.