【作者】 熊剑；

【导师】 刘建业；

【作者基本信息】 南京航空航天大学 ， 导航、制导与控制， 2010， 博士

【摘要】 粒子滤波技术是非线性滤波技术中的一种,理论上它适用于任意非线性系统的状态估计,可有效解决光纤捷联惯性导航系统中的非线性滤波问题。本文针对光纤捷联惯性导航系统的特性,对以粒子滤波为代表的非线性滤波基本理论及其在惯性器件误差补偿、惯性导航系统快速初始对准、惯性/卫星组合导航系统中的应用展开了深入研究。充分发挥和挖掘粒子滤波的优势,解决光纤捷联惯性导航系统中的非线性滤波问题,提高光纤捷联惯性导航系统的精度,为进一步提高武器装备的精确打击能力奠定基础。本文首先针对粒子滤波基本原理,深入研究了粒子滤波在实际应用中出现退化现象的解决方案,从选取更优重要性函数和增加粒子多样性两个方面,对粒子滤波改进算法进行了研究。论文首先从设计更优重要性函数的角度出发,结合马尓柯夫链蒙特卡罗移动方法，设计了基于二阶插值滤波器的重要性函数,研究并提出了一种基于二阶插值滤波的粒子滤波改进算法;从获取更具效率粒子、增加粒子多样性的角度出发,分析了不同近似方法的特性,研究了基于网格的粒子滤波算法;最后针对网格近似过于平均化的问题,引入分层近似的策略,进一步研究并提出了采用分层近似策略的粒子滤波改进算法。为了提高光纤陀螺的使用精度,本文对粒子滤波技术在光纤陀螺ARIMA降噪中的应用进行了研究。在光纤陀螺ARIMA建模方法的基础上,将光纤陀螺的ARIMA模型辨识和状态估计相结合,构建了光纤陀螺的ARIMA非线性滤波模型,研究并提出了基于粒子滤波的光纤陀螺ARIMA模型辨识方法。静态、动态情况下光纤陀螺实际数据验证结果表明,该方法可根据光纤陀螺输出对ARIMA模型参数进行估计,从而提高光纤陀螺状态估计的精度,能够适应不同运动状态下的光纤陀螺信号降噪。为了提高光纤捷联惯性导航系统初始对准速度和精度,本文研究并提出了两种基于粒子滤波的快速初始对准方法。在大失准角对准非线性滤波模型基础上,提出了引入等效东向陀螺信息辅助的快速初始对准方案,研究并提出了基于marginalized粒子滤波的快速初始对准方法。通过对常规二位置初始对准方法的深入分析,推导了二位置情况下的姿态角计算和滤波非线性模型,研究并提出了一种适用于光纤捷联惯性导航系统的粒子滤波二位置快速初始对准方法,有效提高光纤捷联惯性导航系统的初始对准精度和速度,具有较好的工程应用价值。最后,根据光纤捷联惯导系统/GPS组合导航系统的非线性特性,论文将粒子滤波与其他非线性滤波方法有效结合,充分发挥各种滤波方法的特点与优势,研究并提出了基于粒子滤波的光纤捷联惯导系统/GPS松组合导航技术方案。该方案在保证组合导航精度的前提条件下,有效降低粒子滤波计算量,从而提高了粒子滤波应用于光纤捷联惯导系统/GPS组合导航系统的实时性能。本文紧密结合工程应用需求,对粒子滤波技术及其在光纤捷联惯性导航系统中的应用进行了深入和系统的研究,论文研究成果对于非线性粒子滤波技术的工程应用和推广具有重要参考价值。更多还原

【Abstract】 Particle filtering is a key technique in nonlinear filtering. It can be applied to state estimation for any nonlinear system and effectively solve the nonlinear filtering problem in fiber optic strap-down inertial navigation system. Based on the property of fiber optic strap-down inertial navigation system, this paper makes an in-depth research on the basic theory of particle filter and its application to inertial instrument error compensation, to fast initial alignment of inertial navigation system and to inertial / satellite integrated navigation system. The advantages of particle filter are brought into full play to address the non-linear filtering problem in fiber optic strap-down inertial navigation system, to improve the system precision and thus lay the foundation for the enhancement of accurate targeting and attacking of weapons and equipments.Firstly, based on the basic theory of particle filtering, a thorough study is made on the problem of particle degeneracy. The improved particle filtering algorithm is investigated from two aspects: the selection of better importance density and the increase of particle diversity. Started from the point of finding a better importance density, combined with MCMC transfer method, this research first designs the importance density grounded on 2-Order interpolation filters. And then brings forward an improved particle filtering algorithm based on 2-Order interpolation filtering. In terms of particle efficiency and particle diversity, the research first analyzes the characteristics of different approximation methods, studies particle filtering algorithm based on grid, then introduces stratified approximation to solve the problem of even grid approxiamtion, and finally proposes an improved particle filter algorithm using stratified approximation.In order to improve the applied precision of FOG, this paper has studied noise removal performance of particle filtering in fiber optic gyro ARIMA. Based on FOG ARIMA modeling method, the research combines ARIMA model identification with FOG state estimation, constructs a nonlinear filtering FOG ARIMA model and puts forward a FOG ARIMA model identification method based on particle filtering. Data collected from FOG in both static and dynamic states experiments indicates that the proposed method can estimate ARIMA model parameters in response to the outputs of FOG, improve the accuracy of FOG state estimation and reduce signal noise of FOG in different motion states.In view of initial alignment speed and accuracy for fiber optic SINS, this study initiates two fast initial alignment methods based on particle filter. On the basis of large misalignment angle alignment in nonlinear filtering model, this research introduces the measurement information from the equivalent of east gyro and comes up with a fast initial alignment algorithm based on marginalized particle filter. In addition, after a thorough analysis on conventional two-position initial alignment method is made, the nonlinear attitude angle calculation and filter modeling is carefully deduced. Finally, a two-position fast initial alignment algorithm for particle filter feasible to fiber optic strap-down inertial navigation system is proposed. The new approach can effectively improve the accuracy and speed of initial alignment in fiber optic strap-down inertial navigation system, and thus is of great engineering application value.At last, considering the nonlinear characteristics of fiber optic strap-down inertial / GPS integrated navigation system, this research effectively combines particle filtering with other nonlinear filtering method to make the best use of the traits and advantages of all the filtering methods. A fiber optic strap-down inertial / GPS loose integrated navigation scheme based on particle filter is brought forward. With the precision of integrated navigation assured, this scheme can effectively reduce the computational workload of particle filter, and thus improve the real time performance of particle filter used in fiber optic strap-down inertial / GPS integrated navigation system.Closely linked to engineering applications, this dissertation has carried out a thorough and systematic research on particle filtering technique and its applications to fiber optic strap-down inertial navigation system. Achievements of the study will be valuable reference to the engineering application and promotion of nonlinear particle filter technique.更多还原