Extended Kalman filtering based parameter estimation and drift compensation for a MEMS rate integrating gyroscope
Extended Kalman filtering based parameter estimation and drift compensation for a MEMS rate integrating gyroscope
复制标题
DOI：
10.1016/j.sna.2016.09.019
复制
发表时间：
2016-10-15
影响因子：
4.6
通讯作者：
Gallacher, Barry
中科院分区：
文献类型：
--
作者：
Hu, Zhongxu;Gallacher, Barry
提出了一种基于离线扩展卡尔曼滤波的MEMS陀螺参数辨识和漂移补偿方法。阻尼和刚度缺陷是MEMS振动陀螺仪的主要误差源。在速率积分操作模式中，其中输出角度而不是角速度，如在速率陀螺仪的情况下，参数识别是任何反馈控制和补偿算法的必要先决条件，以使角度漂移和其他误差最小化。EKF方法提供了谐振器直流回路增益的五个估计值，以及与非比例阻尼和各向异性弹性相关的四个参数。该方法是基于轨道根数表示的慢变平均动力学模型。平均方法提供了重要的优势，类似的尝试直接基于动态模型表示的快速时变位移和振动速度。首先，由于在计算轨道要素时采用了窄带解调过程，观测到的测量值受到的噪声水平明显较低。其次，由于增广状态的缓慢变化性质，EKF需要低得多的更新速率。这些优点导致更精确的估计、改进的稳定性能和EKF的真实的时间实现的可能性。数值仿真和离线实现的EKF使用实验陀螺仪的操作数据，以验证所提出的方法。此外，识别的阻尼缺陷已被用于基于DSP的真实的时间速率积分陀螺控制系统的漂移补偿控制。最终，最大角漂移已减少到每秒1度。频谱分析表明，角漂误差主要是由传统陀螺仪模型中未考虑的动力学引起的四次谐波。(C)© 2016 Elsevier B.V.版权所有。
This paper presents an offline extended Kalman filtering based parameter identification and drift compensation for a MEMS ring vibratory gyroscope. Damping and stiffness imperfections are the major error sources in MEMS vibratory gyroscopes. In the rate integrating operation mode, where angle is output instead of angular velocity as in the case of the rate gyroscope, parameter identification is an essential prerequisite for any feedback control and compensation algorithm to minimize angle drift and other errors. The proposed EKF method provides five estimates for the resonator DC loop gain, and another four parameters related to the non-proportional damping and aniso-elasticities. The method is based on the slowly varying averaged dynamic model expressed in terms of orbital elements. The averaging methodology offers important advantages over similar attempts based directly on the dynamic model expressed in terms of fast time varying displacement and velocity of vibration. Firstly, the observed measurements are subjected to significantly lower levels of noise as a consequence of the narrowband demodulation process employed in the calculation of the orbital elements. Secondly, the EKF requires much lower update rate due to the slowly varying nature of the augmented states. These advantages result in a more accurate estimation, improved stability performance and the possibility for real time implementation of the EKF. Numerical simulation and offline implementation of the EKF using experimental gyroscope operation data are provided to validate the proposed method. Moreover, the identified damping imperfections have been used in the drift compensation control in a DSP based real time rate integrating gyroscope control system. Ultimately, the maximum angular drift has been reduced to 1 degrees per second. Spectrum analysis shows the angle drift error is dominated by 4th harmonics caused by dynamics not included in the conventional gyroscope model. (C) 2016 Elsevier B.V. All rights reserved.