一种无迹粒子PHD滤波的多目标跟踪算法

针对粒子概率假设密度滤波(P-PHDF)算法估计精度低、滤波发散和粒子退化的缺陷,提出了一种无迹粒子PHD滤波(UP-PHDF)算法.该算法以UKF算法产生重要性函数并从中采样通过观测值更新粒子的权值,再用加权的粒子估计PHD函数,进而得到优化的状态估计均值和方差进行传播最后,对UP-PHDF算法进行了分析和实现,并将该算法和P-PHDF算法进行了比较.仿真结果表明,UP-PHDF算法不仅大大提高了滤波估计的精度,同时提高了跟踪系统的稳定性和鲁棒性.     

基于容积卡尔曼的粒子PHD多目标跟踪算法

标准粒子概率假设密度(standard particle probability hypothesis density, SP-PHD)滤波在预测粒子状态时没有考虑最新的观测信息,因而存在估计精度较低、粒子退化严重的问题,针对上述问题,提出基于容积卡尔曼的粒子概率假设密度(cubature Kalman particle probability hypothesis density, CP-PHD)滤波算法,该算法基于球面-径向容积数值积分准则,利用容积卡尔曼滤波(cubature Kalman filter, CKF)产生建议密度函数,并对其进行采样得到当前时刻的粒子状态,从而使粒子分布更接近于真实的多目标后验概率密度函数。同时,CP-PHD算法性能不受目标状态维数影响,与无迹卡尔曼粒子概率假设密度(unscented Kalman particle probability hypothesis density, UP-PHD)滤波相比,具有更强适应性和更好的跟踪性能。实验结果表明,CP-PHD算法的跟踪精度优于SP-PHD和UP-PHD。     

抗“飞点”的UKF GMPCPHD滤波算法

为实现被动测角目标状态和数目的实时估计,在高斯混合粒子（Gaussian mixture particle, GMP）的势化概率假设密度(cardinalized probability hypothesis density, CPHD) 滤波框架下,提出了基于抗“飞点”无迹卡尔曼滤波器（unscented Kalman filter, UKF）的GMPCPHD滤波算法,即抗“飞点”的UKF GMPCPHD滤波算法。在该算法中,粒子滤波的重要性采样函数由抗“飞点”UKF产生,粒子的预测与更新采用拟蒙特卡罗(quasi Monte Carlo, QMC)方式,目标状态的概率假设密度（probability hypothesis density, PHD）和势分布用一组高斯粒子滤波器（Gaussian particle filtering, GPF）近似。通过该算法与GMPCPHD、UKF-GMPPHD滤波算法的对比仿真,验证了该算法良好的跟踪性能。     

基于改进高斯粒子滤波器的目标跟踪算法

针对现有机动目标跟踪中粒子滤波算法的不足,提出了一种改进的粒子滤波方法。该方法在高斯粒子滤波的基础上通过利用当前时刻量测值对量测误差的分布参数进行实时的统计和更新,并以此得到粒子的权值,从而考虑到了量测值对估计值的影响,该方法适合于量测误差分布为高斯白噪声且状态量与量测误差相关条件下的非线性估计。仿真结果表明,与传统的自举粒子滤波(boot trap particle filter, BPF)、高斯粒子滤波（Gaussian particle filter, GPF)以及无迹粒子滤波（unscented particle filter, UPF)相比,该方法具有较高的精度和较少的计算量。     

Cubature粒子滤波

非线性非高斯下后验概率密度函数解析值无法获得,需设计合理的重要性密度函数进行逼近。传统粒子滤波(particle filter, PF)直接采用未含最新量测信息的状态转移先验分布函数作为重要性密度函数来逼近后验概率密度函数。针对PF缺乏量测信息的问题,提出一种基于Cubature卡尔曼滤波(Cubature Kalman filter, CKF)重采样的Cubature粒子滤波新算法(Cubature particle filter, CPF)。该算法在先验分布更新阶段融入了最新的观测数据,通过CKF设计重要性密度函数,使其更加接近系统状态后验概率密度。仿真表明CPF估计精度高于PF和扩展卡尔曼滤波(extended particle filter, EPF),与无轨迹粒子滤波(unscented particle filter, UPF)相比,其精度相当,但算法运行时间降低了约20%。     

基于无迹变换的概率假设密度滤波算法

基于有限集统计理论的概率假设密度滤波算法运用于多目标跟踪时,不再考虑数据关联问题,突破了传统的跟踪方法.但该滤波公式在非线性条件下没有解析解,在非线性高斯条件下提出了基于无迹变换的概率假设密度滤波算法,实现了算法在强杂波环境下的多目标跟踪.仿真实验比较了该算法与基于粒子滤波的概率假设密度滤波算法的跟踪性能,验证了该算法的跟踪性能和精度.同时分析指出了此算法的不足,以及结合无迹变换与粒子滤波的概率假设密度滤波算法的改进研究方向.     

基于IEK-PF的多传感器序贯融合跟踪

针对粒子滤波中得到优化的重要性密度函数比较困难的问题,将迭代扩展卡尔曼滤波和序贯融合与粒子滤波相结合,应于雷达和红外多传感器目标融合跟踪.利用基于迭代扩展卡尔曼滤波的序贯融合算法得到的系统状态更新矩阵和误差协方差矩阵来构造粒子滤波的重要性密度函数,使重要性密度函数能够融入最新观测信息的同时,更加符合真实状态的后验概率分布.仿真结果表明基于序贯融合的迭代扩展卡尔曼粒子滤波(IEK-PF)能提高状态估计的精度.     

基于无模型无迹粒子滤波的编队卫星相对运动估计

由于地球引力和大气阻力等因素造成的模型不确定性,使常规滤波方法用于卫星编队飞行相对运动估计时精度不高。为克服这种影响,提出了一种融合高斯过程回归(Gaussian process regression,GPR)的无模型无迹粒子滤波(model free unscented particle filter, MF-UPF)方法。对近圆轨道的双星编队问题,新方法通过高斯过程回归对已有的量测数据学习建立相对运动模型, 有效地避免了模型不确定性造成的滤波性能下降。仿真对比验证了无模型无迹粒子滤波在编队飞行相对运动估计中的优越性。     

平方根递推更新GMP-PHDF

传统高斯混合粒子概率假设密度滤波器(Gaussian mixture particle probability hypothesis density filter,GMP PHDF)采用先验状态转移概率密度作为重要性密度函数,会出现粒子退化问题。而递推更新高斯滤波器依据测量函数梯度渐进式地进行状态更新,可获得更为接近于真实分布的后验估计,但其协方差矩阵易非正定而导致递推中断。对此,本文首先分析平方根递推更新高斯滤波器(square-root recursive update Gaussian filter,SR-RUGF)的实现思路,并给出基于容积卡尔曼滤波(cubature Kalman filter,CKF)的SR RUGF实现步骤。在此基础上,利用SR RUGF为GMP PHDF构建重要性密度函数,进而提出基于平方根递推更新的GMP-PHDF(square-root recursive update GMP-PHDF, SRRU-GMP-PHDF)算法。仿真结果表明,算法可以很好地利用量测信息,获得更高精度的估计结果。     

高斯厄米特粒子PHD被动测角多目标跟踪算法

针对传统粒子概率假设密度（probability hypothesis density, PHD）滤波跟踪被动多目标时,估计精度不高,且存在粒子退化,容易导致滤波器发散的问题,提出一种新的被动多目标跟踪算法--高斯厄米特粒子PHD滤波算法。该算法采用一族高斯厄米特滤波产生的高斯分布拟合更优的重要性密度函数,充分考虑了当前时刻的最新量测,并将该方法融入高斯混合粒子PHD（Gaussian mixture particle PHD, GMP-PHD）滤波框架中,在解决观测非线性的同时,有效提高了被动多目标的跟踪精度。实验结果表明,该算法较传统的GMP PHD滤波算法具有更高的状态估计精度,且有效降低了目标的失跟率。     

Unscented extended Kalman filter for target tracking

A new method of unscented extended Kalman filter (UEKF) for nonlinear system is presented. This new method is a combination of the unscented transformation and the extended Kalman filter (EKF). The extended Kalman filter is similar to that in a conventional EKF. However, in every running step of the EKF the unscented transformation is running, the deterministic sample is caught by unscented transformation, then posterior mean of nonlinearity is caught by propagating, but the posterior covariance of nonlinearity is caught by linearizing. The accuracy of new method is a little better than that of the unscented Kalman filter (UKF), however, the computational time of the UEKF is much less than that of the UKF.     

MLP training in a self-organizing state space model using unscented Kalman particle filter

Many Bayesian learning approaches to the multi-layer perceptron (MLP) parameter optimization have been proposed such as the extended Kalman filter (EKF). This paper uses the unscented Kalman particle filter (UPF) to train the MLP in a selforganizing state space (SOSS) model. This involves forming augmented state vectors consisting of all parameters (the weights of the MLP) and outputs. The UPF is used to sequentially update the true system states and high dimensional parameters that are inherent to the SOSS model for the MLP simultaneously. Simulation results show that the new method performs better than traditional optimization methods.     

Optimization-based particle filter for state and parameter estimation

In recent years, the theory of particle filter has been developed and widely used for state and parameter estimation in nonlinear/non-Gaussian systems. Choosing good importance density is a critical issue in particle filter design. In order to improve the approximation of posterior distribution, this paper provides an optimization-based algorithm (the steepest descent method) to generate the proposal distribution and then sample particles from the distribution. This algorithm is applied in 1-D case, and the simulation results show that the proposed particle filter performs better than the extended Kalman filter (EKF), the standard particle filter (PF), the extended Kalman particle filter (PF-EKF) and the unscented particle filter (UPF) both in efficiency and in estimation precision.     

基于IUPF算法的三维无人机毫米波波束跟踪

由于无人机毫米波通信技术具有高速数据传输和广域网络覆盖能力, 因此在军用和民用领域中拥有广阔的应用前景。针对无人机毫米波通信需要进行精确的波束跟踪这一问题, 提出一种基于改进无迹卡尔曼粒子滤波算法的三维波束跟踪方法。该方法首先利用无迹卡尔曼滤波建立建议密度函数并更新采样粒子; 然后计算每一个采样粒子的权值, 并在归一化后再次对粒子进行重采样; 最后计算粒子均值, 得到波束跟踪角度。仿真结果表明, 该方法相较于以往毫米波波束跟踪方法大大降低了估计误差, 显著提高了波束的跟踪精度。     

基于强跟踪UKF的航天器自主导航间接量测滤波算法

针对广义卡尔曼滤波(extended Kalman filter, EKF)和无迹卡尔曼滤波(unscented Kalman filter, UKF)缺乏对系统异常的在线自适应调整能力、导致滤波器精度降低的问题,提出了一种将强跟踪滤波(strong tracking filter, STF)和UKF相结合的滤波算法,并进一步采用部分状态信息作为间接观测量,同时量测噪声方差阵实时调整,从而避免了对观测方程求取Jacobi矩阵的过程,使滤波器的设计得到简化。将该算法应用于航天器自主导航系统中,仿真结果表明,该算法在系统出现突变或缓变异常时,能够迅速检测出异常,在保证较高估计精度的同时,提高了系统的可靠性。     

UPF based autonomous navigation scheme for deep space probe

The autonomous ＂celestial navigation scheme＂ for deep space probe departing from the earth and the autonomous ＂optical navigation scheme＂ for encountering object celestial body are presented. Then, aiming at the conditions that large initial estimation errors and non-Gaussian distribution of state or measurement errors may exist in orbit determination process of the two phases, UPF （unscented particle filter） is introduced into the navigation schemes. By tackling nonlinear and non-Gaussian problems, UPF overcomes the accuracy influence brought by the traditional EKF （extended Kalman filter）, UKF （unscented Kalman filter）, and PF （particle filter） schemes in approximate treatment to nonlinear and non-Gaussian state model and measurement model. The numerical simulations demonstrate the feasibility and higher accuracy of the UPF navigation scheme.     

基于迭代无迹卡尔曼滤波的SLAM算法仿真研究

对迭代无迹卡尔曼滤波算法在SLAM问题中的应用进行仿真研究。通过仿真分析发现,与一般的无迹卡尔曼滤波算法相比,迭代的算法有时无法提高SLAM的精度,继而探讨了SLAM问题中选择采用迭代算法的条件;同时针对迭代算法的观测更新阶段,用阻尼的高斯-牛顿迭代方法改进完全高斯-牛顿迭代方法,从而提出一种改进的基于迭代无迹卡尔曼滤波的SLAM算法。仿真实验对提出的迭代条件进行了验证,仿真结果表明提出的SLAM算法与无迹卡尔曼滤波算法相比,可以进一步提高SLAM问题的估计精度。     

迭代无味卡尔曼滤波器的算法实现与应用评价

为了对各种迭代无味卡尔曼滤波(iterated unscented Kalman filter, IUKF)算法的应用及性能表现给出较为全面、客观的评价,分别导出并探讨了3种IUKF算法之间的内在联系。多种情况下的仿真应用表明,当观测噪声不太大,且该非线性系统状态的后验密度为可用高斯分布很好近似的单峰形式时,或者说是引起系统非线性的状态量是完全瞬时可观测时,选用恰当的IUKF算法,通过2~3次迭代,就可以在保持滤波一致性的条件下,进一步获得显著的精度收益;否则,IUKF相对于无味卡尔曼滤波(unscented Kalman filter, UKF)的迭代收益就难以保证。     

Modified unscented particle filter for nonlinear Bayesian tracking

A modified unscented particle filtering scheme for nonlinear tracking is proposed, in view of the potential drawbacks （such as, particle impoverishment and numerical sensitivity in calculating the prior） of the conventional unscented particle filter （UPF） confronted in practice. Specifically, a different derivation of the importance weight is presented in detail. The proposed method can avoid the calculation of the prior and reduce the effects of the impoverishment problem caused by sampling from the proposal distribution, Simulations have been performed using two illustrative examples and results have been provided to demonstrate the validity of the modified UPF as well as its improved performance over the conventional one.