An analysis of the divergence problem in the Kalman filter

The effect of modeling errors in a linear discrete stochastic system upon the Kalman filter state estimates is investigated. Errors in both plant dynamics and noise covariances are permitted. The errors are characterized in such a manner that a linear recursion relation for the actual estimation error covariances can be derived. Conditions which guarantee that the covariance matrix remains bounded are described in terms of the asymptotic stability of the homogeneous part of the covariance equation and the boundedness of the forcing terms in the inhomogeneous equation.     

Explicit solution to the singular discrete-time stationary linear filtering problem

A closed form solution to the stationary discrete-time linear filtering problem is obtained explicitly in terms of the system state-space matrices in the limiting singular case where the measurement noise tends to zero. Simple expressions, in closed form, are obtained for the Kalman gain matrix both for uniform and nonuniform rank systems and the explicit eigenstructure of the Kalman filter closed loop matrix is derived. The minimum error covariance matrices of the a priori and a posteriori filtered estimates are obtained using this special eigenstructure, and a remarkably different behavior of the solution in the minimum- and nonminimum-phase cases is found.     

基于旋转矩阵KF的低成本MEMS姿态解算

针对低成本MEMS器件组合的姿态检测系统在运动加速度干扰下姿态估计精度较差等问题,提出了一种基于旋转矩阵卡尔曼滤波器(KF)的姿态解算方法.为了克服四元数法观测方程为非线性的缺点,该方法以旋转矩阵部分元素建立状态方程,并对量测加速度采用状态反馈估计的运动加速度进行补偿,减小了外部加速度的干扰,然后通过构造水平观测向量降低了计算复杂度,并给出了量测噪声协方差的推导.最后设计了卡尔曼滤波器对量测信息实现融合.动静态测试表明,该方法消除了累计误差,与无迹卡尔曼滤波(UKF)相比,提高了在运动加速度干扰下的姿态估计精度.     

Performance evaluation of UKF-based nonlinear filtering

The performance of the modified unscented Kalman filter (UKF) for nonlinear stochastic discrete-time system with linear measurement equation is investigated. It is proved that under certain conditions, the estimation error of the UKF remains bounded. Furthermore, it is shown that the design of noise covariance matrix plays an important role in improving the stability of the algorithm. Error behavior of the UKF is then derived in terms of mean square error (MSE), and the Cramér-Rao lower bound (CRLB) is introduced as a performance measure. The modified UKF is found to approach the CRLB if the difference between the real noise covariance matrix and the selected one is small enough. These results are verified by using Monte Carlo simulations on two example systems.     

基于改进卡尔曼滤波的陀螺仪误差补偿算法

针对基于卡尔曼滤波的MEMS陀螺仪误差补偿算法中量测噪声方差选取不准确的问题,提出一种基于改进卡尔曼滤波的陀螺仪误差补偿算法.卡尔曼滤波通常采用统计特性估计得到固定的量测噪声方差,无法自适应地估计不同环境下陀螺仪噪声特性.该算法将卡尔曼滤波与神经网络相融合,使用卡尔曼滤波新息矩阵作为神经网络输入,通过神经网络得到新息协方差矩阵,以此自适应地估计卡尔曼滤波量测噪声方差.将该算法应用到陀螺仪信号误差补偿中,使用Allan方差分析法对原始信号以及误差补偿后的陀螺仪信号进行分析,实验结果表明该算法能够有效地抑制陀螺仪随机误差,提高MEMS陀螺仪的精度.     

有色噪声下的平方根UKF在天文自主导航中的应用

针对由星敏感器和光学导航相机组成的卫星天文自主导航系统, 传统的平方根UKF不能很好地解决测量噪声为有色噪声情况下的非线性滤波问题, 导致导航系统的精度下降. 为此, 提出了一种有色噪声情况下的平方根UKF方法. 同时, 为了避免在数值计算的过程中, 由于舍入误差而破坏误差协方差矩阵的正定性和对称性, 在整个递推计算过程中, 借鉴平方根Kalman滤波理论, 采用协方差矩阵平方根进行递推计算, 改善滤波算法的稳定性, 协方差矩阵的平方根更新用cholesky分解和qr分解来计算. 将该方法应用于卫星天文自主导航系统中, 实验仿真结果表明, 相对于传统的平方根UKF算法, 所设计的平方根UKF算法能够很好地解决测量噪声为有色噪声情况下估计精度低问题.     

The Cramér-Rao estimation error lower bound computation for deterministic nonlinear systems

For continuous-time nonlinear deterministic system models with discrete nonlinear measurements in additive Ganssian white noise, the extended Kalman filter (EKF) convariance propagation equations linearized about the true unknown trajectory provide the Cramér-Rao lower bound to the estimation error covariance matrix. A useful application is establishing the optimum filter performance for a given nonlinear estimation problem by developing a simulation of the nonlinear system and an EKF linearized about the true trajectory.     

On the nice behaviour of the Gaussian projection filter with small observation noise

When projecting on the manifold of Gaussian densities, the projection filter has been shown to be equal to a McShane-Fisk-Stratonovich (MFS) derivation of the Gaussian assumed density filter. Starting from this point, we study the asymptotic behaviour of the Gaussian projection filter when the covariance of the observation noise tends to zero. We prove that the mean square difference between the true state of the system and the estimate given by the projection filter is bounded by a constant which is proportional to the magnitude of the observation noise.     

Analysis of discrete-time Kalman filtering under incorrect noisecovariances

Analysis tools are developed that can be effectively used to study the performance degradation of a filter when incorrect models of the state and measurement noise covariances are used. For a linear time-variant system with stationary noise processes, it is shown that under certain stability conditions on the system model, the one-step prediction error covariance matrix will converge to a steady-state solution even when the filter gain is not optimal. On the other hand, if the state transition matrix has an unreachable mode outside a unit circle, then the modeling errors in the noise covariances may cause the filter to diverge. Bounds on the asymptotic filter performance are computed when the range of errors in the noise covariance matrices are known. Using simple examples, insights into the behavior of a Kalman filter under nonideal conditions are provided     

改进的自适应Kalman滤波在SINS/GPS组合导航中的应用

全球定位系统（GPS）因信号受到遮挡和干扰而产生观测量突然失准,使捷联式惯性导航系统（SINS）/GPS组合导航的卡尔曼滤波器性能急剧下降。针对上述问题,提出了一种改进的自适应卡尔曼滤波的方法,通过失准时的新息对先验状态均方误差阵进行自适应调节,解决了新息协方差与实际严重不符的问题。仿真实验中,对比了传统的卡尔曼滤波、自适应卡尔曼滤波、自适应抗差卡尔曼滤波和改进的自适应卡尔曼滤波的估计性能,证明了所提出算法的有效性。     

Stochastic stability of the unscented Kalman filter with intermittent observations

In this paper, the stochastic stability of the discrete-time unscented Kalman filter for general nonlinear stochastic systems with intermittent observations is proposed. It is shown that the estimation error remains bounded if the system satisfies some assumptions. And the statistical convergence property of the estimation error covariance is studied, showing the existence of a critical value for the arrival rate of the observations. An upper bound on this expected state error covariance is given. A numerical example is given to illustrate the effectiveness of the techniques developed.     

随机逼近自适应滤波在捷联惯导系统初始对准中的应用

对于测量噪声方差未知的捷联惯导系统(SINS),采用常规Kalman滤波进行初始对准会造成较大状态估计误差,甚至使滤波器发散。为了解决系统测量噪声方差未知或不确切知道时SINS的误差估计问题,提出一种基于随机逼近的自适应滤波方法。该方法将Robbins-Monro算法与Kalman滤波相结合,通过简化求逆运算,解决了系统观测噪声特性未知情况下SINS的误差估计问题,并提高了算法的数值稳定性。仿真结果表明,该方法能在系统测量噪声方差未知情况下有效实现SINS初始对准。     

基于最大互相关熵UKF的传感网目标状态和系统偏差联合估计算法

针对空天地一体化传感网络中传感器观测目标时观测噪声具有重尾或突变性质的问题以及系统偏差对目标状态估计的影响,提出一种基于最大互相关熵无迹卡尔曼滤波（MCUKF）的目标状态和系统偏差联合估计（ASMCUKF）算法。MCUKF算法首先通过无迹变换（UT）获得预测状态估计值和协方差矩阵,然后使用基于最大互相关熵准则（MCC）的非线性回归方法重新构建观测信息,增强了UKF对重尾噪声的鲁棒性。ASMCUKF算法通过目标状态向量扩维的方法建立状态方程和带有系统误差的非线性观测方程,根据估计的系统偏差进行偏差配准,改善了系统偏差对目标状态估计的影响。仿真结果表明,ASMCUKF在重尾非高斯观测噪声的环境下对通信目标状态和系统偏差的估计效果比传统方法更好。     

Robust unscented Kalman filter with adaptation of process and measurement noise covariances

Unscented Kalman filter (UKF) has been extensively used for state estimation of nonlinear stochastic systems, which suffers from performance degradation and even divergence when the noise distribution used in the UKF and the truth in a real system are mismatched. For state estimation of nonlinear stochastic systems with non-Gaussian measurement noise, the Masreliez–Martin extended Kalman filter (EKF) gives better state estimates in relation to the standard EKF. However, the process noise and the measurement noise covariance matrices should be known, which is impractical in applications. This paper presents a robust Masreliez–Martin UKF which can provide reliable state estimates in the presence of both unknown process noise and measurement noise covariance matrices. Two numerical examples involving relative navigation of spacecrafts demonstrate that the proposed filter can provide improved state estimation performance over existing robust filtering approaches. Vision-aided robot arm tracking experiments are also provided to show the effectiveness of the proposed approach.     

Discrete-time filtering for nonlinear polynomial systems over linear observations

This paper designs a discrete-time filter for nonlinear polynomial systems driven by additive white Gaussian noises over linear observations. The solution is obtained by computing the time-update and measurement-update equations for the state estimate and the error covariance matrix. A closed form of this filter is obtained by expressing the conditional expectations of polynomial terms as functions of the estimate and the error covariance. As a particular case, a third-degree polynomial is considered to obtain the finite-dimensional filtering equations. Numerical simulations are performed for a third-degree polynomial system and an induction motor model. Performance of the designed filter is compared with the extended Kalman one to verify its effectiveness.     

An interlaced extended Kalman filter

An estimation algorithm for a class of discrete time nonlinear systems is proposed. The system structure we deal with is partitionable into in subsystems, each affine w.r.t. the corresponding part of the state vector. The algorithm consists of a bank of m interlaced Kalman filters, and each of them estimates a part of the state, considering the remaining parts as known time-varying parameters whose values are evaluated by the other filters at the previous step. The procedure neglects the subsystem coupling terms in the covariance matrix of the state estimation error and counteracts the errors so introduced by suitably “increasing” the noise covariance matrices. Comparisons through numerical simulations with the extended Kalman filter and its modified versions proposed in the literature illustrate the good trade-off provided by the algorithm between the reduction of the computational load and the estimation accuracy     

Stochastic stability of the discrete-time extended Kalman filter

The authors analyze the error behavior for the discrete-time extended Kalman filter for general nonlinear systems in a stochastic framework. In particular, it is shown that the estimation error remains bounded if the system satisfies the nonlinear observability rank condition and the initial estimation error as well as the disturbing noise terms are small enough. This result is verified by numerical simulations for an example system     

Noise covariance identification for time-varying and nonlinear systems

Kalman-based state estimators assume a priori knowledge of the covariance matrices of the process and observation noise. However, in most practical situations, noise statistics and initial conditions are often unknown and need to be estimated from measurement data. This paper presents an auto-covariance least-squares-based algorithm for noise and initial state error covariance estimation of large-scale linear time-varying (LTV) and nonlinear systems. Compared to existing auto-covariance least-squares based-algorithms, our method does not involve any approximations for LTV systems, has fewer parameters to determine and is more memory/computationally efficient for large-scale systems. For nonlinear systems, our algorithm uses full information estimation/moving horizon estimation instead of the extended Kalman filter, so that the stability and accuracy of noise covariance estimation for nonlinear systems can be guaranteed or improved, respectively.     

基于GP-SRCDKF的初始对准技术研究

随着对惯性导航系统中对准时间要求的不断提高,初始对准需要在大方位失准角条件下进行,此时需采用非线性滤波方法来实现初始对准。基于此,提出高斯过程回归平方根中心差分卡尔曼滤波算法(GP-SRCDKF)。将高斯过程回归融入到SRCDKF算法中,利用高斯过程得到系统回归模型及噪声协方差,用回归模型代替状态方程和观测方程,对相应的噪声协方差进行实时自适应调整。该算法不仅克服了扩展卡尔曼滤波滤波精度低、需要计算雅可比矩阵的不足,而且可解决传统滤波容易受系统动态模型不确定和噪声协方差不准确的限制。仿真实验结果验证了该算法的有效性和优越性。