Sequential fusion filtering for networked multi-sensor systems based on noise estimation

In networked multi-sensor systems, the measurements sampled by sensors are transmitted to the fusion center through communication network with various time-varying delay phenomenons. The existing methods on the filtering problems of these systems either lost the real time property of the filtering process or lost the optimality of the filtering accuracy. In this paper, a real-time recursive optimal sequential fusion filter is proposed in the sense of linear minimum mean square error (LMMSE), based on the relationship of system states at different sampled instants and a novel noise estimation method. Firstly, based on the relationship of system states at different sampled instants, the measurement received by the fusion center at different time, is re-modeled as a pseudo measurement of the current state. Secondly, a noise estimation method is presented to estimate the gain noises in the pseudo measurement and solve the filtering gain matrix in the filtering process. Thirdly, the optimal estimate of the current state is obtained based on the re-modeled pseudo measurement and the solved filtering gain matrix. A real-time recursive optimal sequential fusion filter is obtained to deal with all the received measurement in the current fusion period according to the above proposed method. Finally, a simulation example is exploited to show the effectiveness of the proposed method.