High-order moment multi-sensor fusion filter design of Markov jump linear systems

Abstract

: To solve the problem of high-order moment Gaussian distribution (HGD) noise in state estimation, a fusion filter for Markov jump linear systems (MJLSs) with high-order moment information obtained from sensor data is designed. To obtain high-order moment information, the multi-sensor MJLS is converted to a single-mode system composed of high-order moment components by using a cumulant generating function. Next, a filter design based on Bayesian theory is established to achieve state estimation with a high-order moment information form according to the transformed single-mode deterministic system. Subsequently, a high-order moment fusion technique based on entropy theory is proposed to obtain a more accurate estimation result of the state by using the high-order moment information obtained from various sensors. Comparing the first- and second-order moment information obtained by traditional Gaussian distribution, the HGD introduces higher-order moment information and makes the fusion process more reasonable. In this way, a more precise and reasonable performance of the state estimation is achieved, depending on the sensor fusion technique. To confirm the effectiveness and advantages of the proposed method, a numerical simulation example is provided with various fusion methods. Thus, the performance of the proposed fusion filter design is verified.