量测缺失确定采样组合导航估计理论研究

GNSS/INS integrated navigation system is the essential and pivotal subsystem in precision-guided weapons, and in the complex and interference environment, it takes on nonlinearity and missing measurement problems, which are coupled with eath other. The traditional hardware solutions for the above problem, including enhanced GNSS reveiver anti-jamming technology or additional introduction of other navigation equipment measurement information, are difficult to be achieved due to the load limits of the missile body, or have shortcomings of hardware complexity, high cost, poor reliability and poor real-time even if the above solutions are implemented. This topic tries to treat the GNSS/INS integrated navigation and position problem as the nonlinear estimation one with missing measurement by measurement equation modeling software approach, which is expected to overcome the above shortcomings in traditional hardware solutions. Firstly, the nonlinear optimal estimation (filtering and smoothing) frameworks with missing measurement are derived in Gaussian sinificance. Then the suboptimal deterministic sampling estimators (filter and smoother) are designed by applying the sampling strategies, including the unscented transformation (UT), the polynomial interpolation and the spherical-radial rule, for approximating the nonlinear state posterior distribution in proposed frameworks. Finally, through constructing the cost function and MIT update criteria, the deterministic sampling estimation and fusion algorithms with missing measurement and adaptive selection of parameter and sampling strategies are studies, which provides theoretical guidance and technical support for improving the accuracy, reliability and real-time of precision-guided weapons navigation.

GNSS/INS组合导航系统是精确制导武器必不可少的关键子系统，其在复杂干扰环境下呈现非线性及量测缺失的耦合问题，而传统的增强GNSS接收机抗干扰技术或额外引入其他导航设备量测信息等硬件解决方案，要么受弹体载荷限制难以实现要么即使实现但存在硬件复杂、成本昂贵、可靠性差、实时性不佳的缺点。为此，本课题通过对量测缺失进行建模的软件方法，首次将GNSS/INS组合导航与定位问题表征为带量测数据缺失的非线性系统估计问题，有望克服传统方法的上述缺点。首先推导高斯意义下带量测缺失的非线性最优估计(滤波和平滑)框架，接着采用UT变换、多项式插值及球面径向规则等采样策略来近似框架中的非线性状态后验分布，进而设计量测缺失下确定采样型次优估计器，最后通过代价函数构造和MIT更新准则，研究参数和采样策略自适应的量测缺失下确定采样型估计与融合算法，为提高精确制导武器导航精度、可靠性及实时性提供理论指导和技术支撑。

项目组面向组合导航、目标跟踪的工程应用需求，研究了非线性与量测随机异常耦合下的系统建模、估计、辨识等问题，取得如下创新性成果：.1.在研究思路创新方面，提出了高斯估计的一般性和通用性理论，建立了基于解析计算和非线性积分迭代递推计算的估计框架，设计了多种逼近非线性积分的数值采样策略，满足了工程中根据快速性和高精度要求自适应选取多种采样方案的实际需求；通过提出巧妙的高斯假设，使复杂特性和非线性的耦合会共存于非线性积分中，切断了耦合与数值采样策略之间的关联性，突破了国际上传统方法必须对复杂特性和非线性进行解耦的局限性，为设计思路在不同滤波器上的互通互用搭建了“桥梁”。.2.在理论创新方面，沿着高斯估计的设计思路，提出了两回路并行执行的量测一步随机延时高斯滤波，通过增加量测噪声估计回路，实现了状态与量测噪声的解相关，系统论证了高斯估计框架结构在解决量测随机延时问题上具有可行性，拓展了高斯估计的适用范围；提出了一种两通道的量测一步随机延时高斯平滑，揭示了前向通道为状态一步平滑而非标准量测下的状态滤波，发现了量测随机延时下高斯平滑器可自动退化为标准高斯平滑，给出了量测随机延时高斯和平滑框架的权值更新公式；提出了一种带有色噪声的非线性高斯滤波框架，通过引入量测差分的方法，避免了系统动态方程与量测方程的嵌套耦合，实现了量测噪声和状态的解相关，突破了传统扩维法计算量大、因缺少量测噪声方差而极易发散的局限性，具有很强的工程应用价值；提出了一种非扩维状态时滞高斯滤波，实现了状态估计以单状态为载体而非扩维状态的迭代执行，突破了扩维策略计算复杂度高、执行困难的理论局限性；提出了基于EM的联合高斯估计与量测延时概率辨识算法，揭示了一类量测随机异常下全数据似然函数等价分解规律，实现了延时概率的次优解析辨识。.目前考虑非线性、量测随机异常、有色噪声、状态时滞的高斯估计、确定采样估计及EM辨识等理论研究结果，国内外尚未见相关文献报道，上述研究成果已发表学术文章11篇，其中国际SCI源期刊10篇， JCR一区3篇，JCR二区6篇, 特别是在国际控制理论领域公认顶级期刊《IEEE Transactions on Automatic Control》、《Automatica》上连续发表论文5篇，荣获陕西省高等学校科学技术一等奖1项（排名4）和陕西省科学技术二等奖1项（排名5）。