【作者】 罗勇；

【导师】 吴文启；

【作者基本信息】 国防科学技术大学 ， 控制科学与工程， 2012， 博士

【摘要】 独立的GNSS接收机始终会面临高动态适应性和抗干扰能力在带宽选择上的固有矛盾，从而制约了其应用范围。GNSS/INS深组合导航系统综合了惯性辅助接收机和矢量跟踪的概念，将接收机内部相关器输出的基带信息和惯性信息进行深层次融合，有效地解决了独立GNSS接收机所面临的问题，并且极大地拓展了其应用范围。综合当前研究文献和相关专利，深组合导航系统大体上可以分为集中式、非相干型级联式和相干型级联式三种结构，其中以非相干型级联式结构应用最为广泛，同时也是本文研究的重点。本文基于GPS L1频点和BD B3频点卫星信号，针对GNSS/INS深组合导航关键技术，开展了惯性信息与基带I/Q信息匹配问题研究、基带信号处理中的跟踪算法研究，论文的主要工作归纳如下：(1)在相干型或非相干型级联式GNSS/INS深组合导航系统中，包含有基带信号预处理滤波器、码/载波NCO控制量计算、卫星位置速度计算、GNSS/INS组合导航滤波器、惯导解算等信息处理单元，受自身物理特性或硬件条件限制，各信息处理单元的更新率通常不一致，将各信息处理单元与最高更新率模块实现软件同步是深组合导航系统正常工作的前提。通过深入分析，上述问题最终归结为惯导系统估计Doppler频率的在线插值问题。论文研究了一种基于外推和CIC滤波器的Doppler频率插值方法，仿真分析的结果表明：外推+CIC滤波器的插值方法将相对插值误差控制在0.3‰以内；对比Kalman滤波+CIC滤波器的插值方法，外推+CIC滤波器的插值方法能提高插值精度约40%。(2)在非相干型级联式GNSS/INS深组合导航系统中，选用码和载波鉴别器输出作为各通道基带信号预处理滤波器的观测量，利用经典的Kalman滤波实现码和载波跟踪，因为码跟踪误差要远大于载波跟踪误差，这一特性会影响基带预滤波器的载波跟踪性能，在一定条件下甚至导致预滤波器发散。此外，码和鉴别器破坏了基带I/Q信息测量噪声原有的独立性，从而不再满足Kalman滤波作为最优估计的前提。借助于误差分离的思想，论文提出了一种基于双滤波器结构的基带信号预处理滤波模型，该模型中利用两个独立的滤波器分别实现载波跟踪和码跟踪；为进一步提高载波跟踪性能，论文针对其中的载波跟踪滤波器，提出了一种修正的Kalman滤波算法，该滤波算法利用二象限反正切鉴相器输出值的条件概率密度，对载波跟踪滤波器观测噪声方差进行周期性修正。基于实际静态场景的仿真结果表明：相比传统的预滤波器模型实现的矢量接收机，双滤波器模型分别提高了Doppler频率跟踪精度、位置精度和速度精度约50%、20%和30%，同时采用修正的Kalman滤波算法分别提高约90%、48%和80%；基于GNSS/INS复合信号源生成动态场景的仿真结果表明：相比传统的预滤波器模型实现的矢量接收机，双滤波器模型分别提高了Doppler频率跟踪精度、位置精度和速度精度约35%、20%和37%，修正的Kalman滤波算法分别提高约85%、53%和71%。(3)对于传统的的基带信号预处理滤波器，其状态变量通常包括归一化信号幅值、码跟踪误差、载波相位跟踪误差、载波频率跟踪误差和载波频率变化率跟踪误差，论文在深入分析传统模型的基础上，分析了一种状态变量只包含载波相位跟踪误差、载波频率跟踪误差和载波频率变化率跟踪误差简化的预处理滤波器模型，并利用该简化模型构建了深组合导航系统。基于GNSS/INS仿真平台，对简化的GPS/INS和BD/INS深组合导航系统的性能进行了评估，测试结果表明：与传统的深组合导航系统相比，简化的深组合导航系统有着几乎相同的跟踪性能和导航性能，但后者的运算复杂度降低约67.8%。(4)当载噪比低于一定阈值时，码或载波鉴别器的输出噪声会急剧增大，对于非相干型深组合导航系统，将无法得到有用的观测信息。针对上述问题，论文研究了一种基于参数优化的跟踪算法，该算法可以在避免使用鉴别器的同时实现码和载波跟踪。仿真结果显示：与基于Kalman滤波的跟踪方法相比，通过基于参数优化的跟踪方法， GPS和BD的载波跟踪灵敏度分别提高了约2.7dB和2.6dB。论文虽然以GPS和BD系统的典型频点展开的，但是相关结论同样适用于其它的频点(如GPS L5、BD B1和B2等)或卫星导航系统(如GLONASS、GALLIEO等)。更多还原

【Abstract】 As independent GNSS receivers always face the dilemma in choosing a bandwidthto satisfy both anti-jamming capability and dynamics adaptation, so the applicationrange of which have been limited。 GNSS/INS deeply integrated navigation systeminherits the INS-aiding and vector-tracking concepts, and fuses the basebandinformation from GNSS receiver’s correlators and INS information in a deep level,which was actually proposed to solve the problems in independent GNSS receivers andgreatly extend the application range of them. Generally, the GNSS/INS deepintegration can be divided into central architecture, coherent federated architecture andnon-coherent federated architecture, the non-coherent federated architecture was mostcommonly investigated, which was also the focus of this dissertation. In thisdissertation, GPS L1frequency and BD B3frequency signals were considered,baseband I/Q information and INS information matching problem, baseband signalprefilter algorithm and deep integration architecture problem have been investigated,main conclusions are summarized as follows:(1) Different information processing modules, such as baseband signal prefilter,code/carrier NCO control information computation, satellite position and velocitycomputation, GNSS/INS integrated navigation filter, INS, etc., are included in coherentor non-coherent federated GNSS/INS deeply integrated navigation system, the updatedfrequency of which are different from each other, however, the synchronization ofwhich is the necessary condition for implementation. After a profound analysis ofrelationships between each two modules, the synchronization problem was boiled downto INS information estimated Doppler frequency online interpolation or up-samplingproblem, and an extrapolation plus CIC(Cascaded Integrate Comb) filter method wasactually proposed to solve this problem, the simulation results showed that:extrapolation improved the precision of INS information estimated Doppler frequency;the single-cascade CIC was sufficient for Doppler frequency interpolation requirementin deeply integrated navigation system.(2) In non-coherent federated GNSS/INS deeply integrated navigation system, thecode and carrier discriminators outputs are considered as measurement information, themeasurement noise of the two kinds of discriminators would not be independent anylonger thereby violating the a priori condition of the Kalman filter. In this dissertation,we have proposed a double-filter based prefilter model, a4-dimension state filter isresponsible for code tracking whereas a3-dimension state filter is responsible for carriertracking, and a modified Kalman filter algorithm has been investigated to furtherimprove the tracking performance of the carrier-tracking-filter, which is obtained bymodifying the measurement noise variance of the carrier-tracking-filter based on the conditional joint probability density function of normalized I/Q measurements andcarrier to noise-density ratio estimation periodically. Simulation results of static fieldtest showed that, compared with traditional prefilter model implemented vector-trackingbased receiver, the double-filter based prefilter model improved the Doppler frequencytracking precision, position precison and velocity precision by about50%,20%and30%respectively, the model with modified Kalman filter alogorithm improved about90%,48%and80%respectively. Simulation results with complex GNSS/INS signalsimulator showed that, compared with traditional prefilter model implementedvector-tracking based receiver, the double-filter based prefilter model improved theDoppler frequency tracking precision, position precison and velocity precision by about35%,20%and37%respectively, while the modified Kalman filter alogorithm improvedabout85%,53%and71%respectively.(3) Generally, the normalized signal amplitude, carrier phase tracking error, carrierfrequency tracking error, carrier frequency rate tracking error and code phase trackingerror are included in the state space of traditional prefilter. In this dissertation, wehave put forward a simplified prefilter model to replace the traditional one, the statespace of this simplified prefilter model consists of only carrier phase tracking error,carrier frequency tracking error and carrier frequency rate tracking error. Simulationresults showed a more or less identical tracking and navigation performance of this twoprefilter implemented GPS/INS or BD/INS deeply integrated navigation systems,however, the simplified one has been reduced the computational complexity by67.8%.(4) The measurement errors of discriminators increase very rapidly whencarrier-to-signal, correspondently, the measurement information of non-coherent deeplyintegrated navigation system would become invalid. An optimization based trackingmethod was investigated to solve this problem in this dissertation, which implementedcarrier and code tracking without using any discriminators. Simulation results showedabout2.7dB GPS carrier tracking and2.6dB BD carrier tracking sensitivityimprovement over traditional Kalman filter based tracking method.Although the discussions in this dissertation were based on two classical frequencies ofGPS and BD signals, the corresponding algorithms or models would be applicable toother frequencies(such as GPS L5, BD B1and B2, etc.) or GNSS(such as GLONASS,GALLIEO, etc.).更多还原