惯性导航辅助的无缝定位改进模型研究

发布时间：2018-06-05 04:50

本文选题：组合导航 + 抗差自适应模型　；参考：《中国矿业大学》2014年博士论文

【摘要】：惯性导航系统是一种无源导航设备，相对于其他导航系统而言，具有自主性强、短时精度高，可以连续输出导航信息，在军事、民用中都有巨大应用价值。本文围绕无缝定位的惯性导航模型改正方法中的关键技术开展研究，重点涵盖惯导元件随机误差辨识、异常检测与改正抗差自适应滤波模型、机器学习辅助遮蔽区智能导航算法，惯性辅助的行人航迹推算与零速修正室内导航定位，主要研究成果如下： (1)针对常规Allan方差计算量庞大，基于最小二乘拟合随机误差参数时无法修正系数矩阵，提出一种基于WTLS的Allan方差简化估计算法。经实测数据验证，表明该算法可实现大幅降低计算量、加快运算速度并保持Allan方差分析的准确性。 (2)针对室外遮蔽区卫星失锁，提出一种改进径向基神经网络结合自适应滤波辅助的组合系统导航模型。采用遗传算法参数寻优和最近邻聚类学习算法改进径向基神经网络，通过预测出伪观测值与其对应的协方差，实现了卫星失锁情况下短时可靠的导航算法。 (3)提出一种改进抗差非线性滤波模型，通过判断矩阵病态性自主选取抗差策略；针对松组合系统观测无冗余，无法区分观测异常和状态异常，提出一种支持向量回归辅助的组合导航抗差自适应模型，实现智能区分观测值异常和动力学模型异常，保证组合导航精度，实现抗差精度与可靠度的统一。 (4)提出一种LS-SVR辅助的改进多重渐消自适应SVD-UKF算法，利用奇异值分解抑制UKF中先验协方差矩阵负定性变化，采用LS-SVR算法削弱观测异常对残差序列高斯白噪声分布特性的影响，拓展了多重渐消因子的应用范围，为多变量复杂系统提供了一种可行的先进滤波模型。 (5)基于时频变换分析惯性传感器用于室内定位的噪声特征，提出了基于FIR设计滤波器、磁力计数据改进步态检测和姿态计算的行人航迹推算；采用广义似然比辨识零速时刻，，基于可靠观测构建自适应滤波模型的零速度修正导航模型。削弱观测值中的噪声信息，提高定向稳定性，增强零速检测的可靠性，改进惯性辅助室内行人导航精度。
[Abstract]:Inertial navigation system is a kind of passive navigation equipment. Compared with other navigation systems, inertial navigation system has strong autonomy, high accuracy in short time, and can continuously output navigation information. It has great application value in military and civilian applications. This paper focuses on the key techniques of the inertial navigation model correction method for seamless positioning, focusing on the random error identification of inertial navigation elements, anomaly detection and correction robust adaptive filtering model. Machine learning aided shelter area intelligent navigation algorithm, inertial aided footpath reckoning and zero speed correction indoor navigation positioning. The main research results are as follows: In view of the large amount of calculation of Allan variance and the inability to modify the coefficient matrix when least square fitting random error parameters, a simplified Allan variance estimation algorithm based on WTLS is proposed. The experimental results show that the algorithm can greatly reduce the computational complexity, speed up the calculation and maintain the accuracy of Allan variance analysis. A new integrated system navigation model based on improved radial basis function neural network and adaptive filter is proposed. The radial basis function neural network is improved by using genetic algorithm parameter optimization and nearest neighbor clustering learning algorithm. By predicting the pseudo-observation value and its corresponding covariance, the short-time and reliable navigation algorithm is realized in the case of satellite lost lock. (3) an improved robust nonlinear filtering model is proposed, in which the robust strategy is chosen by the ill-conditioned judgment matrix, and the observation of the loose composite system is not redundant, so it is impossible to distinguish the observed anomalies from the state anomalies. An adaptive robust model of integrated navigation aided by support vector regression is proposed, which can intelligently distinguish the anomaly of observation value from the anomaly of dynamic model, ensure the accuracy of integrated navigation, and realize the unity of robust accuracy and reliability. In this paper, an improved multi-fading adaptive SVD-UKF algorithm assisted by LS-SVR is proposed. The singular value decomposition (SVD) is used to suppress the negative qualitative change of the prior covariance matrix in UKF, and the LS-SVR algorithm is used to weaken the influence of the observed anomalies on the white noise distribution of the residual Gao Si sequence. The application range of multiple fading factors is extended and a feasible advanced filtering model for multivariable complex systems is provided. Based on time-frequency transform (TFT) analysis of noise characteristics of inertial sensors for indoor positioning, a new method is proposed, which is based on FIR filter, magnetometer data to improve gait detection and attitude calculation, and the generalized likelihood ratio is used to identify the zero-speed time. A zero-velocity modified navigation model based on reliable observation is constructed for adaptive filtering model. The noise information in observation value is weakened, the directional stability is improved, the reliability of zero velocity detection is enhanced, and the precision of inertial assistant indoor pedestrian navigation is improved.
【学位授予单位】：中国矿业大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN96

【参考文献】