【作者】 吉宇人；

【导师】 周卫东；

【作者基本信息】 哈尔滨工程大学 ， 导航、制导及控制， 2013， 博士

【摘要】 传递对准技术是为运动中的载体提供准确的初始基准。在瞬息万变的现代战争中，传递对准技术已成功应用于舰载、机载战术制导武器。而在舰载机传递对准中，由于载机指向不定，此时极有可能出现方位失准角为大角的情况；同时，海浪冲击、载舰机动将引起船体的挠曲变形；而机载惯导与舰载惯导之间的距离较远，挠曲变形和杆臂效应的耦合将使误差补偿问题变得更加复杂。针对上述情况，本文在比较和分析各类误差模型、匹配方法、滤波算法及误差补偿方法的基础上，对大方位失准角传递对准和误差补偿技术展开研究。主要研究内容如下：首先针对传递对准的三个主要误差源：惯性测量元件误差、挠曲变形误差和杆臂误差建立相应的误差模型，在此基础上分别推导了经典传递对准误差模型和快速传递对准误差模型。从速度误差和姿态失准角的定义出发，分析了两种误差模型中参考坐标系之间的相互关系，通过深入研究，给出了经典传递对准误差模型和快速传递对准误差模型的一致性条件与适用范围，为后面的匹配方法研究和四元数传递对准算法打下基础。匹配方法是传递对准技术的重要组成部分，本文深入研究速度匹配和姿态匹配。速度匹配不受到惯导类型的影响，且具备较为突出的综合性能。针对速度匹配的可观测性问题，在分析研究速度匹配量测与状态变量之间关系的基础上，提出一种更为直观的可观测性解析方法，并通过推导匀速直线运动下各状态与量测多阶微分的关系式验证了该方法的有效性。当主、子惯导都为捷联惯导时，姿态匹配成为一个重要的选择。针对姿态匹配量测与速度误差耦合较弱的问题，从量测量的获取和算法复杂度出发，深入研究了姿态角匹配、姿态矩阵匹配和量测失准角匹配。在不同机动条件下对速度匹配和速度+姿态匹配进行了仿真验证。深入研究大失准角传递对准中的非线性滤波算法。针对系统状态维数较高的问题，在比较分析两种单形采样策略的基础上选取超球体采样策略，降低了sigma点数量，提高算法实时性。同时为提高非线性滤波算法的计算稳定性，结合平方根UKF滤波算法，提出了超球体采样SRUKF对大失准角传递对准下的非线性系统进行估计。并通过舰载机大方位失准角传递对准仿真分析验证所提算法的有效性。针对欧拉角法在大失准角传递对准问题中存在的奇异性、运算繁琐等问题，引入非奇异、精度高和计算简单的姿态四元数法。在定义姿态误差四元数的基础上，首先推导得到四元数经典传递对准误差模型和四元数快速传递对准误差模型。与“速度+姿态”匹配相对应，深入研究四元数量测失准角匹配并给出量测方程。在该匹配方法中，量测量由主、子惯导的即时修正四元数相乘得到。得到四元数非线性误差模型的基础上，对四元数传递对准UKF算法进行研究。其中，针对四元数UKF中的四元数协方差阵奇异问题，利用误差四元数中非冗余的状态误差矢量作协方差阵运算，并根据UT变换作适当改进。同时，针对四元数加权均值的规范性问题，利用姿态参数切换和特征值向量求解的方法进行解决。后者在构造四元数代价函数的基础上，通过特征向量求解获得MMSE下的四元数加权均值，避免了姿态参数切换引起的实时性和数值稳定性下降问题。最后针对扩维滤波中采样点数量增加和高阶误差方差阵的平方根运算问题，引入超球体采样和平方根滤波思想，提出了四元数SSRUKF滤波。大方位失准角下的仿真实验证明了四元数传递对准算法的有效性。针对挠曲变形噪声补偿法中噪声方差难以确定的问题，深入研究描述挠曲变形的Markov过程的统计特性，并给出计算方法。在此基础上对挠曲变形角与杆臂效应的耦合问题进行定量分析，并将其统一等价为噪声补偿问题。针对噪声补偿法无法实时跟踪外界变化的问题，引入极大似然法及强跟踪滤波算法。其中，Sage-Husa自适应滤波对初值较为敏感，且滤波不稳定。对此，融入强跟踪滤波在线调整协方差阵和增益矩阵的思想，并在强跟踪滤波中增加量测噪声协方差阵的权重，重新定义渐消因子，从而使其更好发挥当前量测的作用。最后结合极大似然法，给出改进Sage-Husa自适应滤波算法。通过仿真实验验证在补偿噪声方差初值小于真实值的情况下改进Sage-Husa自适应滤波的有效性。更多还原

【Abstract】 Transfer alignment technology provides accurate initial reference for vehicles in motion.In rapidly changing modern war，transfer alignment technology has been applied to shipborneand airborne tactical guided weapons successfully. But in the transfer alignment ofcarrier-based aircrafts, the situation of carrier aircraft’s direction cannot be determined willlead to the large azimuth misalignment. Meanwhile, the shock of waves and maneuvering ofcarrier ship will cause flexural deformation, and high distance between shipborne inertialnavigation system(INS) and airborne INS give rise to the coupling of flexural deformationand lever arm effect, which make the error compensation problem more complicated. Aimingat above problems, the transfer alignment technology under large misalignments and errorcompensation methods will be studied in this paper through comparing and analysising ofdifferent error models, matching methods, filtering algorithms and error compensationmethods.The main error sources of transfer alignment like errors of inertial measurementunit(imu), errors of flexural deformation and lever arm errors are modeled firstly, then classictransfer alignment error models and rapid transfer alignment error models are deducedrespectively. According to the definition of velocity and attitude errors, the relations ofreference coordinate systems of two error models are analyzed. On the basis of above job, theconsistency and scope of the application of two error models are deeply studied, which laysthe foundations for the following matching methods and quaternion transfer alignmentalgorithm.Matching methods are important parts of transfer alignment technology, and the velocitymatching and attitude matching are mainly studied. Velocity matching is not subjected to theINS type, and have good comprehensive performance. Aiming at the observability problem ofvelocity matching, a more intuitive analytical method is proposed based on the analysis of therelation between measurements in velocity matching and states. The effectiveness is verifiedby deriving the relationship between states and measurement of multi-order differential underthe situation of uniform linear motion. When main INS and slave INS are both strapdowninertial navigation system(SDINS), attitude matching method becomes an important option.Because of the weak coupling relationship between measurements in attitude matching andvelocity errors, the attitude angle matching、 attitude matrix matching and measuredmisalignment matching are studied and compared according to the measurement obtainingand the complexity of the measurement equations, which resultes in the “velocity+attitude” matching, whose state and measurement equations are given finally. The velocity matchingand“velocity+attitude” matching are simulated under different maneuvering.Aiming at transfer alignment of large misalignment, the nonlinear filtering algorithm ismainly studied. To deal with the problem of high state dimension, the spherical sampling ischosen by comparing two simplex sampling strategy, so that the number of sigma point canbe reduced and real-time of algorithm can be achieved. Meanwhile, the square rootUKF(SRUKF) is used to improve the computational stability of nonlinear filtering algorithm.Finally the spherical sampling SRUKF(SSRUKF) is proposed. The effeteness of the methodis verified by the simulation experiment of transfer alignment of carrier-based aircraft underlarge azimuth misalignment.Aiming at the problems of singularity and high computational burden of euler anglemethod in transfer alignment under large misalignment, the nonsingular，precise and easyattitude quaternion is introduced. On the basis of defining error attitude quaternion, thequaternion classic transfer alignment error models and rapid transfer alignment error modelsare derived firstly. Corresponding to “velocity+attitude” matching, the quaternion measuredmisalignment matching method is studied deeply and the measurement equations are given,where the measurements are obtained by multiplying the immediate correction quaternion ofMINS and SINS. On the basis of above job, the quaternion UKF is studied. Thenon-redundant vectors of error quaternion are used to compute the covariance, so that thesingularity of quaternion covariance can be avoid. The covariance equation is adjustedappropriately by UT transformation. The method based on solving eigenvalue vector is usedto calculate the weighted mean of normal quaternion which is based on the construction of thecost function. Finally, the augmented UKF is given as a result of the multiplicative noise. Theeffeteness of proposed method is verified by the simulation experiment of transfer alignmentof carrier-based aircraft under large azimuth misalignment.The statistical properties of markov process which describe the flexural deformation arestudied and the equations are given to determine the noise variance for flexural deformation.On the basis of above job, the coupling of flexural deformation and lever arm effect isquantitatively analyzed, and equaled to the problem of noise compensation. In order to solvethe problem that the compensated noise cannot follow the external changes in real time,maximum likelihood method and strong tracking filtering are introduced. The idea ofadjusting covariance and gain matrix online is integrated to solve the problem that theSage-Husa adaptive filtering is unstable and sensitive to the initial value of filtering. To takefull advantage of current measurements, the weight of measurement noise covariance is increased, and the fading factors are redefined in strong tracking filtering. By combiningmaximum likelihood method, an improved Sage-Husa adaptive filtering is given finally. Theeffectiveness of Sage-Husa adaptive filtering is verified by the simulation experiment, wherethe variance of compensated noise is set smaller than the true value.更多还原