一种提高神经网络泛化能力的自适应UKF滤波算法

给出了利用EKF(extended Kalman)滤波和UKF(unscented Kalman)滤波提高神经网络泛化能力的方法.针对UKF参数选取随意性的问题,采用移动开窗估计法对状态噪声和观测噪声协方差矩阵进行自适应估计,提出了一种新的提高神经网络泛化能力的自适应UKF算法.利用检测样本进行了验证,结果表明,利用EKF、UKF和自适应UKF算法训练神经网络都能提高其泛化能力,其中自适应UKF算法优于其他几种算法.     

一种采样型平方根滤波及其应用

探讨了非线性系统的滤波问题，提出了将采样型平方根滤波SR-UKF（square root unscented Kalmanfilter）用于星载GPS卫星实时定轨。在滤波过程中，以协方差阵的平方根代替协方差阵参加递推运算，有效地提高了滤波算法的计算效率和数值稳定性。实例计算结果表明，SR-UKF的性能要优于推广卡尔曼滤波（extended Kalmanfilter）和Unscented卡尔曼滤波（unscented Kalmanfilter）。     

一种新型非线性滤波理论及其应用

研究一种新型的非线性滤波理论,即Unscented卡尔曼滤波（UKF）,同时为了获得更高的计算效率和确保协方差阵的非负定性,研究了平方根UKF。将UKF和平方根UKF应用到星载GPS卫星定轨中,实际算例表明UKF和平方根UKF的性能要优于常用的推广卡尔曼滤波的性能。     

修正扩维无迹卡尔曼滤波算法研究

在脉冲星导航中，角位置误差是主要的误差源之一。因此本文基于X射线脉冲星导航，提出了一种修正扩维无迹卡尔曼滤波（MASUKF）算法进行角位置误差的改进。MASUKF算法在原扩维无迹卡尔曼滤波（ASUKF）算法的基础上加入了Roamer延迟的高阶项，并将其作为误差项，将修改后的误差项代入状态方程与量测方程中，即可进行仿真分析。在仿真中，首先将ASUKF算法与UKF算法进行对比仿真模拟试验，结果显示ASUKF算法能显著地提高约45 m的定位精度，在X、Y、Z 3方向的速度误差估计精度约提高了20%；然后比较ASUKF与MASUKF算法，结果显示MASUKF算法较ASUKF算法的速度误差与位置误差的估计精度均提高2%以上。     

鲁棒无迹卡尔曼滤波组合导航定位方法

针对无迹卡尔曼滤波(UKF)在非高斯噪声或统计特性不准确时滤波精度会下降甚至发散等问题,提出了一种基于Huber-M估计的无迹卡尔曼滤波导航算法。首先采用奇异值分解(SVD)迭代计算代替协方差矩阵的迭代变换;然后将Huber方法用于UKF框架中,使先验信息和量测信息进行重构;最终以达到克服传统UKF滤波器稳定性差的问题,提高滤波抗差能力。对提出算法进行GPS/UWB组合导航仿真验证,并与EKF和UKF进行了比较。实验结果表明,加入M估计的SVD-UKF在噪声统计特性不准确时和加入随机观测异常状态下都可以将滤波器性能提高25%～40%,与其他两种算法相比,本文所提算法的定位误差能快速收敛,并保持较高滤波精度。     

用GPS观测值对推断数据进行非线性平滑的模型建立

通过对全球定位系统（GPS）与推断系统（DRS）的联合系统中影响DRS观测值的主要误差来源进行分析，建立了一个表示这些误差的数学模型。并将该模型用于建立一个在DRS观测值存在的情况下，状态向量估计的非线性平滑器。     

多传感器行人航位推算方法和UKF融合算法

针对利用惯性测量单元进行行人航位推算（PDR）时,其定位误差会随时间累积的问题,提出了一种基于多传感器融合的室内行人航位推算方法;对于智能移动设备的低成本多传感器,设计了基于无迹卡尔曼滤波（UKF）的初始对准,设定4种阈值条件进行步伐状态检测;在行走过程中,针对步长和航向角误差累积的问题,利用基于UKF的零速度更新（ZUPT）对速度误差进行修正,零角速率更新（ZARU）和磁力计融合对航向角误差进行修正,从而有效提高了行人最终的位置精度。试验结果表明：使用该方法可以有效提高PDR位置精度,平均位置偏差占总路程的1.5%左右。     

SPKF滤波方法在变形监测数据分析中的应用

讨论了SP(Sigma Point)变换算法的性质,给出了一种新的扩展型卡尔曼滤波方法SPKF(Sigma PointKalman Filter)。它不仅具有较高的精度,而且不必计算偏导数阵。结合变形监测数据处理进行仿真分析的结果表明,SPKF具有良好的状态估计性能,而且使用简便,适合于非线性系统状态估计。     

基于UKF的GPS非线性动态滤波算法

介绍了一种Unscented卡尔曼滤波算法，它通过确定性采样获得一组采样点，可获得更多的观测假设，对系统状态统计特性的估计更加准确，同时该算法无需对系统方程进行线性化，避免了传统的EKF算法由于线性化引入的误差。本文将UKF算法用于GPS非线性动态滤波技术中，建立了仿真模型并定义了仿真条件，与EKF算法的仿真结果相比，在系统状态统计特性未知的情况下，UKF算法对系统状态的估计更准确，定位精度更高。     

SPKF滤波方法在变形监测数据分析中的应用

讨论了SP(Sigma Point)变换算法的性质,给出了一种新的扩展型卡尔曼滤波方法SPKF(Sigma Point Kalman Filter).它不仅具有较高的精度,而且不必计算偏导数阵.结合变形监测数据处理进行仿真分析的结果表明,SPKF具有良好的状态估计性能,而且使用简便,适合于非线性系统状态估计.     

Application of Unscented Kalman Filter in Satellite Orbit Simulation

IntroductionAs is well known,the Kal manfilter(KF) is al-ways usedto deal withthe system whose dynam-ics and observation models are linear , and theextended Kal manfilter(EKF) is the most widelyused esti mator for nonlinear systems . In theEKFthe kal man …     

Application of unscented Kalman filter in satellite orbit simulation

A new estimate method is proposed, which takes advantage of the unscented transform method, thus the true mean and covariance are approximated more accurately. The new method can be applied to nonlinear systems without the linearization process necessary for the EKF, and it does not demand a Gaussian distribution of noise and what's more, its ease of implementation and more accurate estimation features enables it to demonstrate its good performance in the experiment of satellite orbit simulation. Numerical experiments show that the application of the unscented Kalman filter is more effective than the EKF.     

UKF算法及其在GPS卫星轨道短期预报中的应用

分析了GPS卫星预报星历,在比较分析EKF和UKF优缺点的基础上,将UKF引入GPS卫星轨道预报研究中.数值模拟和结果分析表明,UKF方法预报更稳定,能有效地提高轨道预报精度和稳定性.     

移动机器人SLAM位姿估计的改进四元数无迹卡尔曼滤波EI北大核心CSCD

在全自主运动控制的移动机器人系统中,自身位姿的估计和校正对于移动机器人的运动至关重要。卡尔曼滤波是解决移动机器人同步定位与地图构建(SLAM)常用方法。相较于卡尔曼滤波,无迹卡尔曼滤波(UKF)无须对复杂的非线性函数进行雅可比矩阵运算。本文基于无迹卡尔曼滤波,根据先验协方差的平方根选择sigma点,计算协方差以及加权均值。用四元数表示姿态,将四元数矢量转换为旋转空间进行矩阵运算,在此基础上设计了一种位姿估计算法——基于四元数平方根的无迹卡尔曼滤波(QSR-UKF)算法。试验将EKF、QSR-UKF、SR-UKFEKF 3种算法的位姿估计结果进行仿真分析,并通过相关定量指标进行了描述,验证了本文算法的有效性。     

Unscented Kalman filter with nonlinear dynamic process modeling for GPS navigation

This paper preliminarily investigates the application of unscented Kalman filter (UKF) approach with nonlinear dynamic process modeling for Global positioning system (GPS) navigation processing. Many estimation problems, including the GPS navigation, are actually nonlinear. Although it has been common that additional fictitious process noise can be added to the system model, however, the more suitable cure for non convergence caused by unmodeled states is to correct the model. For the nonlinear estimation problem, alternatives for the classical model-based extended Kalman filter (EKF) can be employed. The UKF is a nonlinear distribution approximation method, which uses a finite number of sigma points to propagate the probability of state distribution through the nonlinear dynamics of system. The UKF exhibits superior performance when compared with EKF since the series approximations in the EKF algorithm can lead to poor representations of the nonlinear functions and probability distributions of interest. GPS navigation processing using the proposed approach will be conducted to validate the effectiveness of the proposed strategy. The performance of the UKF with nonlinear dynamic process model will be assessed and compared to those of conventional EKF.     

Design and implementation of a model predictive observer for AHRS

A GPS-aided Inertial Navigation System (GAINS) is used to determine the orientation? position and velocity of ground and aerial vehicles. The data measured by Inertial Navigation System (INS) and GPS are commonly integrated through an Extended Kalman Filter (EKF). Since the EKF requires linearized models and complete knowledge of predefined stochastic noises? the estimation performance of this filter is attenuated by unmodeled nonlinearity and bias uncertainties of MEMS inertial sensors. The Attitude Heading Reference System (AHRS) is applied based on the quaternion and Euler angles methods. A moving horizon-based estimator such as Model Predictive Observer (MPO) enables us to approximate and estimate linear systems affected by unknown uncertainties. The main objective of this research is to present a new MPO method based on the duality principle between controller and observer of dynamic systems and its implementation in AHRS mode of a low-cost INS aided by a GPS. Asymptotic stability of the proposed MPO is proven by applying Lyapunov’s direct method. The field test of a GAINS is performed by a ground vehicle to assess the long-time performance of the MPO method compared with the EKF. Both the EKF and MPO estimators are applied in AHRS mode of the MEMS GAINS for the purpose of real-time performance comparison. Furthermore? we use flight test data of the GAINS for evaluation of the estimation filters. The proposed MPO based on both the Euler angles and quaternion methods yields better estimation performances compared to the classic EKF.     

基于SUT-EKF的DGPS/DR 组合定位算法

针对基于DGPS/DR的移动机器人组合定位问题,采用一种尺度无迹变换扩展卡尔曼滤波(SUT-EKF)算法,由于组合定位系统中的状态方程是非线性的,并且观测方程是线性的特点,将SUT预测移动机器人位姿,利用EKF融合最新观测值更新机器人位姿,该算法在状态预测阶段避免了计算Jacobian矩阵,从而有效地减小了线性化对非线性系统误差的影响。仿真结果表明,该算法具有较好的滤波精度和稳定性,为实现DGPS/DR组合定位系统提供了一种有效可靠的途径。     

IMU辅助X射线脉冲星的深空探测器精确定姿方法及仿真

为满足深空探测器的精确定姿需求,提出了一种惯性测量单元(IMU)辅助的X射线脉冲星定姿方法。该方法用IMU的速率陀螺来估计航天器短时姿态,观测两颗或多颗脉冲星的X射线辐射信号,将拟合得到的观测矢量作为滤波器信息输入,利用这两种测姿手段在时间和空间上的互补特性,提供一种全天候、抗干扰性强的定姿方法。仿真结果表明,相比于EKF,基于UKF的俯仰、横滚和偏航三姿态角的测量精度可提高21.9%、21.1%和31.7%;与仅使用脉冲星或IMU的定姿方法相比,组合定姿方法的俯仰角估计精度分别提高了32.5%和77.6%。