基于自适应UKF微型航姿系统噪声在线估计.docVIP

基金项目：国家自然科学基金资助项目(No.; MEMS振动传感与微姿态组合测井技术国际联合研究中心 科技平台与基地建设 (cstc2014gjhz0038)； 作者简介：刘宇(1972-),男(汉)，重庆人，硕士生导师，教授。长期从事振动惯性传感器，光纤传感器，惯性导航及传感器件的研究。 刘琼(1989-),男(汉)，湖南人，硕士研究生，学生。主要研究振动惯性传感器，惯性导航及传感器件的研究。 基于自适应UKF微型航姿系统噪声在线估计 刘宇 刘琼 周帆 李云梅 向磊 （重庆邮电大学 光电信息感测与传输技术重庆市重点实验室，重庆 400065） 摘 要：针对先验噪声与系统真实噪声不符引起标准无迹卡尔曼（UKF）性能退化的情况，提出了一种应用于非线性时变状态和参数联合估计的自适应UKF（AUKF）算法。该算法首先基于新的协方差矩阵与相应估计值之间存在的误差，构建成本函数。然后为了最小化成本函数，采用梯度下降法进行在线预估。最后对噪声的协方差进行在线更新，将更新后的协方差反馈给标准的UKF，使更新后的理论协方差接近真实噪声。通过实验仿真分析，自适应UKF相比标准UKF，静态精度提高了40%，动态精度提高了30%，验证了自适应UKF噪声在线估计模型的准确性和可行性。在实际测试系统中，对于时变噪声协方差不确定时，自适应UKF噪声在线估计的鲁棒性得到明显改善。 关键词：UKF ；AUKF；联合估计；成本函数；梯度下降算法；鲁棒性 中图分类号：TN212 文献标识码：A Noise On-line Estimation of Mini-AHRS Based On Adaptive UKF Algorithm Yu LIU, Qiong LIU， （Chongqing Municipal Level Key Laboratory of Photoelectronic Information Sensing Transmitting Technology, Chongqing University of Posts and Telecommunications, Chongqing, 400065, China） Abstract: Considering that the prior noise of a normal unscented Kalman filter does not agree with its real behavior，an adaptive unscented Kalman filter algorithm applied to nonlinear joint estimation of both time-varying states and parameters is proposed. Firstly, a cost function is built based on the error between the covariance matrices of innovation and their corresponding estimations. Then in order to minimize the cost function, the gradient descent method for on-line forecast is used. Finally, the noise covariance is online updated, the updated covariance feedback to the standard UKF, which makes the updated theoretical covariance closer to the true noise. Through experimental simulation analysis, compared adaptive UKF with standard UKF, static precision is improved by 40%, dynamic precision is improved by 30%, the accuracy and feasibility of online adaptive UKF noise estimation model is verified. In the actual test system, for time-varying noise covariance is uncertain, adaptive UKF online noi