高精度惯性导航系统不易受到外界环境干扰，可自主、连续地确定运载体的位置、速度和姿态，对国防现代化具有重要意义。本文针对现阶段空间稳定系统性能提升的主要制约因素，从重力补偿、极区导航和精对准及标定优化三方面展开研究，实现了系统的精度提升、工作范围扩展和启动加速。研究内容包括： 明确了空间稳定系统中重力扰动矢量引起的时域位置误差，主要为垂线偏差在水平通道中1:1传递的低频分量和其中频波动激起的舒拉振荡。建立了补偿精度与补偿间隔和计算时间的理论关系，得出航海条件下单点计算应快于10 s，补偿间隔建议在100s以内。设计了导航解算与重力计算并行的实时补偿方案，从计算流程和系数文件结构方面进行了球谐模型计算的优化设计，使单点重力矢量计算时间降低至1 s以内，可同时满足航空和航海条件下的实时性要求。动态试验数据的重力补偿结果显示，地形变化剧烈区域频率大于 的误差成分最大降低了80.82%（纬度）和76.09%（经度）。补偿后DOV的影响基本被消除，导航精度得到显著提升。 分析了空间稳定系统极区误差振荡特性，得出极点附近水平速度阻尼将趋于失效。提出了一种适用于空间稳定系统的混合全球导航方案，并在极区导航模式采用横向和格网两种输出模式的情况下，推导了系统误差模型、误差传播特性、阻尼方案和长时导航误差，补充了空间稳定系统极区导航方法和极区误差理论的空白。建立了横向导航与格网导航的联系，得出横向地理坐标系与格网地理坐标系只相差绕D轴的90°转角，该结论有助于统一极区导航坐标系。基于该理论提出了一种将格网机械编排统一到横向导航的改进方案，使格网导航可完全脱离传统经纬度，计算更简单。设计跨极点半实物仿真方法验证了极区导航理论。 设计了三种精对准与标定优化方案：方案一将热稳定后期与精对准及标定滤波器前期收敛阶段重合实现时间复用，可将启动时间缩短11 h；方案二采用考虑加速度计误差影响的17维卡尔曼滤波器解决估计值长周期波动的问题，可将启动时间缩短12 h；方案三在采用17维滤波器的基础上，同时进行热稳定与精对准及标定阶段的时间复用，可将启动时间缩短23 h。利用三种工况的试验数据验证了3种方案均不影响导航精度，其中方案三对系统启动时间优化效果显著。
High-precision inertial navigation system (INS) is not easily interfered by the external environment and is able to determine the vehicle’s position, velocity and attitude independently and continuously, which is of great significance to the modernization of national defense. This dissertation aims at the main constraints in performance enhancement of the space-stable INS and focuses on the gravity compensation, polar navigation and the optimization of inertial alignment and calibration in order to improve the navigation precision, expand the working area and speed up the initialization. Research of this dissertation includes the following aspects:Gravity induced position error of the space-stable INS is clarified in the time domain, which is mainly the 1:1 propagation of the low-frequency component of deflection of vertical into the horizontal channel and the Schuler oscillation excited by its medium-frequency component. The theoretical relationship between compensation accuracy and compensation time interval, gravity computing time is established. Under the shipborne condition, the calculation time of single-point gravity should be shorter than 10 s and the compensation interval is better to be less than 100 s. A concurrent real-time gravity compensation scheme is designed, and the computing of the spherical harmonic model is optimized in the computing process and the structure of the coefficient file. The optimization shortens the single-point computing time to less than 1s, which satisfies the real-time requirement under both the aerial and shipborne conditions. Gravity compensation result of dynamic test data shows that position error components with a frequency higher than have been maximally reduced by 80.82%(latitude) and 76.09%(longitude) in the areas with drastic topographic changes. After compensation, the influence of DOV is basically removed and the navigation accuracy is significantly improved.Analysis of the change of the polar error oscillation character is conducted, which indicates that the horizontal damping network tends to be invalid near the pole. A hybrid global navigation scheme appropriate for the space-stable INS is designed. In the two different cases of hybrid transverse and grid polar navigation modes, a thorough study is conducted on the error model, error propagation, damping technique and long-term navigation error, which fills in the blank in polar navigation of the space-stable INS. The relationship between transverse navigation and grid navigation is established. Based on the geometrical relationship of the transverse and the grid frame, it is found out that the only difference between the two coordinates is a 90° rotation angle around the D axis. This conclusion benefits the unification of the polar navigation coordinate. Modified grid navigation is proposed to make it united with the transverse navigation, which has no need for traditional latitude and longitude and has a simpler calculation form. A semi-physical simulation method is designed, and the theoretical analysis on polar navigation is validated based on this method.Three optimization schemes of initial alignment and calibration are proposed: The first scheme overlaps the later stage of thermos-stabilization and earlier stage of initial alignment and calibration to achieve time multiplexing, which can shorten the initialization time by 11 h; The second scheme takes the un-modeled accelerometer error into account and uses a 17-dimension Kalman filter to solve the long-period fluctuation of the estimated values, which can shorten the initialization time by 12 h; The third scheme uses the 17-dimension Kalman filter and adopts time multiplexing of the later stage of thermos-stabilization and earlier stage of initial alignment and calibration, which can shorten the initialization time by 23 h. Test data from three different working conditions validate that all of the 3 schemes do no harm to the navigation precision, among which the third scheme significantly speeds up the initialization of the system.
