加性噪声下增广容积卡尔曼滤波及其目标跟踪应用

AUGMENTED CUBATURE KALMAN FILTER FOR ADDITIVE NOISE AND ITS APPLICATION TO OBJECT TRACKING

传统容积卡尔曼滤波(CKF)有良好的滤波精度和较低的计算复杂度,使其广泛被应用于目标跟踪系统。但在高维非线性和波动性大的目标跟踪系统中,3阶和高阶CKF分别存在滤波精度不足和稳定性低的问题。为提高CKF的滤波精度并保证稳定性,讨论和给出加性噪声下的增广容积卡尔曼滤波(ACKF)。在仿真中,将CKF、UKF和ACKF应用于5维高非线性目标跟踪,并分析比较三者的目标跟踪性能。研究结果表明,在高维非线性目标跟踪系统中,3阶ACKF可以获得更好目标跟踪精度和稳定性,以及可接受的计算复杂度。

Since the cubature Kalman filter （CKF） provides a good accuracy with low computational complexity, it is wildly applied in estimation and tracking systems. But for a tracking system involving high dimensionality and acute nonlinearity, 3-degree CKF and high-degree CKF encounter low accuracy and instability problems, respectively. To improve the the performance, augmented cubature Kalman filter for additive noise is discussed. In the simulation, CKF, UKF and ACKF are applied to 5-dimensional targets tracking system. Besides, their performances including accuracy, stability and complexity are compared by RMSEs. The results show that 3-degree ACKF can obtain better tracking accuracy and stability with acceptable computational complexity than UKF, 3-degree CKF and 5-degree CKF in highly nonlinear and dimensional systems.