Parallel fusion distributed unscented information filter algorithm for sparse dynamic wireless sensor network

Sparsity and stochastic dynamic change are two kinds of instability factors of communication topology, which universal joint exist in real wireless sensor network (WSN). The practicability of distributed unscented information filter (DUIF) will be improved vastly if making it applicable to the sparse dynamic WSN. For this purpose, a parallel fusion DUIF (PF-DUIF) is proposed. In the PF-DUIF algorithm, the local unscented information filter (LUIF) and weighted average consensus filter (WACF) can be implemented parallelly by applying the real local posterior estimated mean and covariance to generat the sigma points. And then the consensus tracking errors caused by stochastic dynamic communication topolagy can be avoided effectively. Meanwhile, by implementing average consensus filter on the unbias local information matrices and vectors output by the LUIF respectively in the WACF, the distributed posterior estimated results without average consensus error can be got. Furthermore, employing the real-time update mechanism to avoid the problem of filter asynchronization in PF-DUIF algorithm caused by stochastic dynamic communication topolagy. And meanwhile, on the basis of the average network model of the sparse dynamic WSN, the convergence rate of the WACF is modified under the condition of limited communication energy consumption, so as to improve the global efficiency of PF-DUIF algorithm. The simulation results show that PF-DUIF algorithm can efficiently track the target in sparse dynamic WSN. © 2016, Editorial Department of Control Theory & Applications South China University of Technology. All right reserved.