Ryu et al.提出一种基于Kalman滤波和信号子空间的机动多目标角度跟踪算法,在该算法中需要计算信号子空间矩阵W的投影矩阵,因而需要计算N×N维复数逆矩阵,这主要是由于W的列向量间不正交。提出一种用Kalman滤波预测的角初始化W的方...Ryu et al.提出一种基于Kalman滤波和信号子空间的机动多目标角度跟踪算法,在该算法中需要计算信号子空间矩阵W的投影矩阵,因而需要计算N×N维复数逆矩阵,这主要是由于W的列向量间不正交。提出一种用Kalman滤波预测的角初始化W的方法,使得在用PASTd算法时W能够更快地收敛于列向量为正交向量的矩阵,从而避免了计算N×N维复数逆矩阵,既降低算法的运算量,同时跟踪性能也得到提高。展开更多
In this paper,we present a novel particle filter(PF)-based direct position tracking method utilizing multiple distributed observation stations.Traditional passive tracking methods are anchored on repetitive position e...In this paper,we present a novel particle filter(PF)-based direct position tracking method utilizing multiple distributed observation stations.Traditional passive tracking methods are anchored on repetitive position estimation,where the set of consecutive estimates provides the tracking trajectory,such as Two-step and direct position determination methods.However,duplicate estimates can be computationally expensive.In addition,these techniques suffer from data association problems.The PF algorithm is a tracking method that avoids these drawbacks,but the conventional PF algorithm is unable to construct a likelihood function from the received signals of multiple observatories to determine the weights of particles.Therefore,we developed an improved PF algorithm with the likelihood function modified by the projection approximation subspace tracking with deflation(PASTd)algorithm.The proposed algorithm uses the projection subspace and spectral function to replace the likelihood function of PF.Then,the weights of particles are calculated jointly by multiple likelihood functions.Finally,the tracking problem of multiple targets is solved by multiple sets of particles.Simulations demonstrate the effectiveness of the proposed method in terms of computational complexity and tracking accuracy.展开更多
文摘Ryu et al.提出一种基于Kalman滤波和信号子空间的机动多目标角度跟踪算法,在该算法中需要计算信号子空间矩阵W的投影矩阵,因而需要计算N×N维复数逆矩阵,这主要是由于W的列向量间不正交。提出一种用Kalman滤波预测的角初始化W的方法,使得在用PASTd算法时W能够更快地收敛于列向量为正交向量的矩阵,从而避免了计算N×N维复数逆矩阵,既降低算法的运算量,同时跟踪性能也得到提高。
基金supported by China NSF Grants(62371225,62371227)Postgraduate Research&Practice Innovation Program of Jiangsu Province(KYCX250590).
文摘In this paper,we present a novel particle filter(PF)-based direct position tracking method utilizing multiple distributed observation stations.Traditional passive tracking methods are anchored on repetitive position estimation,where the set of consecutive estimates provides the tracking trajectory,such as Two-step and direct position determination methods.However,duplicate estimates can be computationally expensive.In addition,these techniques suffer from data association problems.The PF algorithm is a tracking method that avoids these drawbacks,but the conventional PF algorithm is unable to construct a likelihood function from the received signals of multiple observatories to determine the weights of particles.Therefore,we developed an improved PF algorithm with the likelihood function modified by the projection approximation subspace tracking with deflation(PASTd)algorithm.The proposed algorithm uses the projection subspace and spectral function to replace the likelihood function of PF.Then,the weights of particles are calculated jointly by multiple likelihood functions.Finally,the tracking problem of multiple targets is solved by multiple sets of particles.Simulations demonstrate the effectiveness of the proposed method in terms of computational complexity and tracking accuracy.
