Stability of Parameter Estimates for a Gaussian Process

Maximum-likelihood estimates for the levels of the mean value function and the covariance function of a Gaussian random process are investigated. The stability of these estimates is examined as the actual covariance function of the process deviates from the form assumed in the estimators. It is found that the time-bandwidth product for stationary processes represents an upper bound on the number of estimator terms that can be safely used when estimating with uncertainty about the process covariance function. This result is consistent with other interpretations of the time-bandwidth product and tempers the conclusion that, in principle, an infinite number of estimator terms can be used to obtain a perfect estimate of the covariance level. In practice, the estimate of the level can never be perfect, and the accuracy of the estimate depends on the observation interval. Finally, conditions are established to ensure asymptotic stability of the estimates and physical interpretations are presented.     

Reduced state estimator for systems with parametric inputs

A reduced state estimator is derived for systems with bounded parameters as inputs. Optimal filter gains are derived for minimizing the total covariance of the estimation error due to measurement noise and parameter uncertainty. It is shown that these filter gains for a two-state system with a Gaussian parameter satisfy the Kalata relation in steady state. Equations are also derived for optimally filtering measurements in arbitrary time order. This reduced state estimator offers novelties over a traditional Kalman filter in its application to the class of problems considered. The total error covariance, which is minimized, makes no use of plant noise. Furthermore, the filter is easier to optimize in high dimensional and multiple sensor applications as well as in processing out-of-sequence measurements.     

Cubature Kalman filter with closed-loop covariance feedback control for integrated INS/GNSS navigation

Cubature Kalman Filter (CKF) offers a promising solution to handle the data fusion of integrated nonlinear INS/GNSS (Inertial Navigation System/Global Navigation Satellite System) navigation. However, its accuracy is degraded by inaccurate kinematic noise statistics which originate from disturbances of system dynamics. This paper develops a method of closed-loop feedback covariance control to address the above problem of CKF. In this method, the posterior state and its covariance are fed back to the filtering process to constitute a closed-loop structure for CKF covariance propagation. Subsequently, based on the maximum likelihood principle, a control scheme of the prior state covariance is established by using the feedback state and covariance within an estimation window and further adopting a proportional coefficient to amplify the feedback terms in recent time steps for the full use of new information to reflect actual system characteristics. Since it does not directly use kinematic noise covariance, the proposed method can effectively avoid the adverse impact of inaccurate kinematic noise statistics on filtering solutions. Further, it can also guarantee the prior state covariance to be positive semi-definite without involving extra measures. The efficacy of the proposed method is validated by simulations and experiments for integrated INS/GNSS navigation.     

基于约束最小冗余线阵与干扰对消的测向方法

针对有源干扰背景下信号源和干扰源的个数超过线阵的自由度而产生线阵饱和现象,提出一种将约束最小冗余线阵与干扰对消技术相结合的测向方法。通过将无源状态和有源状态下线阵输出数据的协方差矩阵进行对消运算去除有源干扰和噪声分量,并对约束最小冗余线阵的波达方向(DOA)估计算法进行改进,构造了新的协方差Toeplitz矩阵,有效抑制了由阵列非均匀性导致的伪峰,提高了阵列的DOA估计性能。仿真结果表明:该算法在低信噪比背景下具有抗有源干扰能力,扩展了阵列孔径,并具有较高的测向精度和鲁棒性。     

鲁棒EKF在脉冲星导航系统中的应用

针对脉冲星导航系统的滤波问题,传统的扩展卡尔曼滤波(EKF)算法存在不能克服系统模型存在不确定性参数以及乘性噪声等缺陷,提出一种鲁棒EKF算法。首先,分析了状态预测误差方程和估计误差方程,利用统计学原理,得到了状态预测方差矩阵和状态估计方差矩阵计算等式。由于系统模型存在不确定性参数,状态预测协方差矩阵和状态估计协方差矩阵无法计算;因此,利用4个重要矩阵不等式,分析并找到预测方差矩阵和状态估计方差矩阵的上界。最后,利用状态估计误差协方差矩阵上界设计状态增益矩阵,使得状态估计协方差矩阵的迹最小。将该算法对脉冲星导航系统进行仿真,仿真结果验证了所提算法的有效性。     

多渐消因子Kalman滤波器在SINS初始对准中的应用(英文)

针对系统模型和统计信息不能精确已知的条件下Kalman滤波无法给出最优解这一问题,单一渐消因子Kalman滤波算法对于简单的系统是有效的,但是对于复杂的多变量系统,仅仅利用单个的渐消因子是不够的。本文提出了一种多渐消因子滤波算法,通过利用开窗法计算新息序列协方差的无偏估计获得渐消因子矩阵。利用渐消因子矩阵调节一步预测均方误差矩阵k|k1P,对不同的滤波通道提供不同的渐消速率。将该方法应用于SINS的初始对准中,仿真和试验结果表明:当真实系统噪声统计特性同设定参数不一致时,对准精度明显高于其他滤波算法。其对不确定性噪声具有较低的敏感度,对系统参数具有较好的滤波效果。因而,在实际应用中具有重要的参考价值。     

确定时间序列协方差函数的方法

提出一种确定时间序列协方差函数的方法,它首先根据(多元)时间序列构造其互协方差函数随机序列、互相关函数随机序列或自协方差函数随机序列、自相关函数随机序列,然后采用谱分析和多点平均方法对互协方差函数随机序列、互相关函数随机序列或自协方差函数随机序列、自相关函数随机序列的趋势项进行分离,分别求得其周期项和非周期项的函数表达式,再综合给出整个趋势项函数。从而得到原时间序列的互协方差函数、互相关函数或自协方差函数、自相关函数的函数形式,并通过最小二乘方法确定其中的待定参数。该方法可用于时间序列协方差函数的建模、分析和预测,并且计算简单易行、精度高,便于实际应用。      

Systematic methods for knowledge acquisition and expert systemdevelopment [for combat aircraft]

Nine cooperating rule-based systems, collectively called AUTOCREW which were designed to automate functions and decisions associated with a combat aircraft's subsystems, are discussed. The organization of tasks within each system is described; performance metrics were developed to evaluate the workload of each rule base and to assess the cooperation between the rule bases. Simulation and comparative workload results for two mission scenarios are given. The scenarios are inbound surface-to-air-missile attack on the aircraft and pilot incapacitation. The methodology used to develop the AUTOCREW knowledge bases is summarized. Issues involved in designing the navigation sensor selection expert in AUTOCREW's NAVIGATOR knowledge base are discussed in detail. The performance of seven navigation systems aiding a medium-accuracy inertial navigation system (INS) was investigated using Kalman filter covariance analyses. A navigation sensor management (NSM) expert system was formulated from covariance simulation data using the analysis of variance (ANOVA) method and the ID3 algorithm     

Kalman filtering for matrix estimation

A general discrete-time Kalman filter (KF) for state matrix estimation using matrix measurements is presented. The new algorithm evaluates the state matrix estimate and the estimation error covariance matrix in terms of the original system matrices. The proposed algorithm naturally fits systems which are most conveniently described by matrix process and measurement equations. Its formulation uses a compact notation for aiding both intuition and mathematical manipulation. It is a straightforward extension of the classical KF, and includes as special cases other matrix filters that were developed in the past. Beyond the analytical value of the matrix filter, it is shown through various examples arising in engineering problems that this filter can be computationally more efficient than its vectorized version.     

基于稀疏表示的双基地MIMO雷达多目标定位及幅相误差估计

 基于稀疏表示理论,提出一种新的双基地多输入多输出(MIMO)雷达收发角度及幅相误差估计算法。利用接收数据,分别构造发射和接收协方差矩阵,并以列向量化后的发射和接收协方差矩阵为量测信号建立2个一维稀疏线性模型,构造模型求解的 L₂-L₁ 混合范数优化目标函数,通过交替迭代寻优获得目标角度估计和幅相误差估计,最后给出了本文算法的收敛性分析。与现有算法相比,该算法充分利用了目标发射和接收空域的稀疏特性,且能够通过对噪声功率的预估计来抑制噪声。仿真结果表明:在低信噪比(SNR)条件下,本文算法仍能够得到较好的估计精度,且对幅相误差具有一定的稳健性。     

基于新息异常检测的改进抗差自适应卡尔曼滤波算法

在工程应用中,量测异常及量测噪声统计特性的时变是引起标准卡尔曼滤波振荡甚至发散的主要原因。经典抗差Sage-Husa自适应滤波方案,对量测中的孤立型异常有所抵抗,并可在线估计量测噪声统计特性改善滤波效果,但当连续型异常值出现时,其滤波效果不佳。针对现有抗差Sage-Husa自适应滤波方案的不足,提出了新的改进滤波方法。在改进算法中,当检测到量测异常时采用模值更大的先验预测方差阵代替原算法中的后验估计方差阵,在估计量测噪声方差时起到放大作用,以降低异常量测权重,提高滤波精度;采用IGG方案构造了新的权函数,可在抑制异常影响的同时调节估计方差阵,以免连续异常时新息持续置零引起的滤波发散;采用标准卡尔曼滤波新息辅助异常检测的双重检测策略,避免了因量测噪声方差阵的调节引起检测阈值变化而导致的漏检率增高。仿真实验表明,与常规抗差自适应滤波算法相比,该方案可更加有效地抑制量测异常值的影响。     

A Stochastic Algorithm for Sensitivity Analysis

A direct stochastic sensitivity analysis algorithm is developed for linear dynamical systems having incompletely known input statistics. The new algorithm extends previous results by applying covariance propagation concepts which utilize as a forcing function the sensitivity covariance matrix associated with the uncertainty in the elements of the system input covariance matrix itself. The developed algorithm is evaluated in the context of a generalized sensitivity analysis formulation involving nonlinear transformations on the input signals. Numerical results are provided to demonstrate the usefulness of the new algorithm.     

Efficient L-D factorization algorithms for PDA, IMM, and IMMPDAfilters

Efficient algorithms exist for the square-root probabilistic data association filter (PDAF). The same approach is extended to develop square-root versions of the interacting multiple model (IMM) Kalman filter and the IMMPDAF algorithms. The computational efficiency of the method stems from the fact that the terms needed in the overall covariance updates of PDAF, IMM, and IMMPDAF can be obtained as part of the square-root covariance update of an ordinary Kalman filter. In addition, a new square-root covariance prediction algorithm that is substantially faster than the usual modified weighted Gram-Schmidt (MWG-S) algorithm, whenever the process noise covariance matrix is time invariant, is proposed     

Discrete-time state estimation of analog double integrators

The problem of minimum variance discrete-time state estimation of a continuous-time double integrator via noisy continuous-time measurements is considered. The error covariance matrices of this estimation are calculated and analyzed. The relations between these covariance matrices and the error covariance matrix of the optimal continuous-time filter are obtained, and a way for determining the required sampling period is proposed. A commonly used approximated model is investigated; it is shown to be inappropriate unless a specific improvement is introduced in the model     

噪声协方差估计新算法在惯性导航系统中的仿真研究

研究一种在动态系统常值误差未知的情况下对线性时变随机系统误差协方差进行估计的新方法。该方法通过构造一个新的时间序列,其协方差由未知参数的线性组合组成,然后用递推最小二乘法来计算新序列的协方差,该方法不需要任何关于噪声的先验知识。从仿真结果来看达到了较好的效果。     

Coordinate Conversion and Tracking for Very Long Range Radars

The problem of tracking with very long range radars is studied in this paper. First, the measurement conversion from a radar's r-u-v coordinate system to the Cartesian coordinate system is discussed. Although the nonlinearity of this coordinate transformation appears insignificant based on the evaluation of the bias of the converted measurements, it is shown that this nonlinearity can cause significant covariance inconsistency in the conventionally converted measurements (CM1). Since data association depends critically on filter consistency, this issue is very important. Following this, it is shown that a suitably corrected conversion (CM2) eliminates the inconsistency. Then, initialized with the converted measurements (using CM2), four Cartesian filters are evaluated. It is shown that, among these filters, the converted measurement Kalman filter with second order Taylor expansion (CM2KF) is the only one that is consistent for very long range tracking scenarios. Another two approaches, the range-direction-cosine extended Kalman filter (ruvEKF) and the unscented Kalman filter (UKF) are also evaluated and shown to suffer from consistency problems. However, the CM2KF has the disadvantage of reduced accuracy in the range direction. To fix this problem, a consistency-based modification for the standard extended Kalman filter (E1KF) is proposed. This leads to a new filtering approach, designated as measurement covariance adaptive extended Kalman filter (MCAEKF). For very long range tracking scenarios, the MCAEKF is shown to produce consistent filtering results and be able to avoid the loss of accuracy in the range direction. It is also shown that the MCAEKF meets the posterior Carmer-Rao lower bound for the scenarios considered.     

GLRT subspace detection for range and Doppler distributed targets

A generalized likelihood ratio test (GLRT) is derived for adaptive detection of range and Doppler-distributed targets. The clutter is modeled as a spherically invariant random process (SIRP) and its texture component is range dependent (heterogeneous clutter). We suppose here that the speckle component covariance matrix is known or estimated thanks to a secondary data set. Thus, unknown parameters to be estimated are local texture values, the complex amplitudes and Doppler frequencies of all scattering centers. To do so, we use superresolution methods. The proposed detector assumes a priori knowledge on the spatial distribution of the target and has the precious property of having a constant false alarm rate (CFAR) with the assumption of a known speckle covariance matrix or by the use of frequency agility.     

Statistical tests for higher order analysis of radar clutter: their application to L-band measured data

We design three statistical tests to ascertain whether radar data comply with the hypotheses of multivariate Gaussianity, spatial homogeneity, and covariance persymmetry, respectively. For the first issue we develop a statistical procedure based on quadratic distributional distances, which exploits the representation of Gaussian vectors in generalized spherical coordinates. As to the spatial homogeneity we propose a technique, based on the Kolmogorov-Smirnov (KS) test, relying on the properties of quadratic forms constructed from Gaussian vectors and Wishart distributed matrices. Finally, in order to analyze the persymmetry property of the disturbance covariance matrix, we design a testing procedure based on the generalized likelihood ratio test (GLRT). We thus apply the new tests to L-band experimentally measured clutter data, collected by the MIT Lincoln Laboratory Phase One radar, at the Katahdin Hill site. The results show that the multivariate Gaussian hypothesis for the considered data file is reasonable. On the contrary the assumption of spatial homogeneity can be done only within small clutter regions which, in general, exhibit also a persymmetric covariance matrix.     

Fast converging adaptive processor or a structured covariance matrix

The use of adaptive linear techniques to solve signal processing problems is needed particularly when the interference environment external to the signal processor (such as for a radar or communication system) is not known a priori. Due to this lack of knowledge of an external environment, adaptive techniques require a certain amount of data to cancel the external interference. The number of statistically independent samples per input sensor required so that the performance of the adaptive processor is close (nominally within 3 dB) to the optimum is called the convergence measure of effectiveness (MOE) of the processor. The minimization of the convergence MOE is important since in many environments the external interference changes rapidly with time. Although there are heuristic techniques in the literature that provide fast convergence for particular problems, there is currently not a general solution for arbitrary interference that is derived via classical theory. A maximum likelihood (ML) solution (under the assumption that the input interference is Gaussian) is derived here for a structured covariance matrix that has the form of the identity matrix plus an unknown positive semi-definite Hermitian (PSDH) matrix. This covariance matrix form is often valid in realistic interference scenarios for radar and communication systems. Using this ML estimate, simulation results are given that show that the convergence is much faster than the often-used sample matrix inversion method. In addition, the ML solution for a structured covariance matrix that has the aforementioned form where the scale factor on the identity matrix is arbitrarily lower-bounded, is derived. Finally, an efficient implementation is presented.     

Asymptotic distribution of the conditional signal-to-noise ratio in an eigenanalysis-based adaptive array

The statistical characterization of the conditioned signal-to-noise ratio (SNR) of the sample matrix inversion (SMI) method has been known for some time. An eigenanalysis-based detection method, referred to as the eigencanceler, has been shown to be a useful alternative to SMI, when the interference has low rank. In this work, the density function of the conditioned SNR is developed for the eigencanceler. The development is based on the asymptotic expansion of the distribution of the principal components of the covariance matrix. It is shown that, unlike the SMI method, the eigencanceler yields a conditional SNR distribution that is dependent on the covariance matrix, It is further shown that simpler, covariance matrix-independent approximations can be found for the large interference-to-noise case. The new distribution is shown to be in good agreement with the numerical data obtained from simulations.