Multi-sensor adaptive data fusion with colored measurement noise

Abstract

Listen

Translate article

As one of the key theories of target tracking, Multi-sensor data fusion(MSDF) is widely utilized in engineering practice. Generally, the research of MSDF is under the hypothesis that the measurement systems of sensors are only influenced by white noise and their priori knowledge is obtained, which is not appropriate for practical application. Considering these situations, a white noise model convention and adaptive measurement noise adjustment is added into the two proposed fusion methods called centralized measurement fusion(CMF) and decentralized state fusion(DSF). The equations of fusion estimation results are derived both with and without white noise convention. The method of adaptive covariance adjustment based on artificial neural network(ANN) is explored. The simulation results show that the measurement model is estimated more precisely by denoting the colored noise with white noise, the adaptive variance adjustment method contributes to overcoming the priori information deficiency of noise and keep the tracking precision, and the DSF has higher tracking accuracy than the local filter based on each sensor but lower than the CMF.