Research on temperature drift compensation method of mems gyroscope based on neural network
-
摘要:
针对MEMS惯性传感器对温度变化非常敏感问题,提出了一种MEMS 陀螺仪实时温度补偿方法,并比较了此方法与传统使用的分段线性拟合方法的补偿效果. 首先,使用一个标定过的MEMS-IMU进行温度试验,并在试验过程中采集MEMS陀螺仪的输出；其次,提出了一个基于MEMS陀螺仪前一时刻温漂输出的实时补偿模型,并引入虚假最近邻算法来确定神经网络隐藏层的神经元个数；然后,采用了一种普通循环神经网络的变体,门循环单元循环神经网络,来辨识所提出模型的输入与输出之间的关系,从而得到完整的MEMS陀螺仪温漂实时补偿模型；最后,比较了所提出的方法和分段线性拟合方法的补偿效果. 为了得到定量的比较结果,分别在相同试验条件下和不同升温速率条件下比较了以上两种方法补偿前后MEMS陀螺仪的艾伦方差和零偏稳定性. 根据比较结果可知,在以上两种试验条件下,采用所提出的方法补偿后,MEMS陀螺仪的艾伦方差系数和零偏稳定性均显著降低,并且补偿效果均优于分段线性拟合方法.
Abstract:Aiming at the problem that MEMS inertial sensors are very sensitive to temperature changes, a real-time temperature compensation method for MEMS gyroscopes is proposed, and the compensation effect of this method is compared with the traditional piecewise linear fitting method. First, an in-house-designed MEMS-IMU for temperature effect tests, and the output of the MEMS gyroscope is collected during the test; Second, a real-time compensation model based on gyroscope output is proposed at the last moment, and the false nearest neighbor algorithm is introduced to determine the number of neurons in the hidden layer of the neural network. Then, a variant of the standard recurrent neural network, which is the gated recurrent neural network, is used to identify the relationship between the input and output of the proposed model, and a complete real-time compensation model for the temperature drift of the MEMS gyroscope is obtained. Finally, We compare the proposed and the piecewise linear fitting methods’ compensation effects. Toconditions and different heating rate conditions, respectively. According to the comparison results, under the above two test conditions, the Allen variance coefficient and zero bias stability of the MEMS gyroscope are significantly reduced after the compensation by the proposed method, and the compensation effect is better than the segmented linear fitting method. Therefore, the method proposed in this paper effectively compensates for the temperature drift of the MEMS gyroscope.
-
Key words:
- mems gyroscope /
- temperature drift compensation /
- neural network /
- false nearest neighbor
-
图 8 验证数据拟合结果-升温速率1℃/min
Figure 8. Fitting results of the validation data- heating rate 1℃/min
图 9 验证数据补偿结果-升温速率1℃/min
Figure 9. Compensation results of the validation data-heating rate 1℃/min
图 10 验证数据拟合结果-升温速率2℃/min
Figure 10. Fitting results of the validation data-heating rate 2℃/min
图 11 验证数据补偿结果-升温速率2℃/min
Figure 11. Compensation results of the validation data-heating rate 2℃/min
图 12 验证数据补偿前后艾伦方差对比-升温速率1℃/min
Figure 12. Comparison of Allen variance before and after compensation for validation data-heating rate 1℃/min
图 13 验证数据补偿前后艾伦方差对比-升温速率2℃/min
Figure 13. Comparison of Allen variance before and after compensation for validation data-heating rate 2℃/min
表 1 常温下MEMS-IMU性能指标−10 s平滑
Table 1. Performance index of MEMS-IMU at room temperature −10 s smoothing
X陀螺（°/h） Y陀螺（°/h） Z陀螺（°/h） X加计/
mgY加计/
mgZ加计/
mg零偏稳
定性9.10 11.31 7.61 0.25 0.21 0.29 零偏重
复性13.50 4.83 2.17 0.09 0.13 0.07 
下载: 导出CSV
表 2 FNN百分比随嵌入维数变化情况
Table 2. Percentage of FNN VS.embedding dimension
$ J $ $ l $ $P/{\%}$ 0 0 27.05 1 0 23.87 1 1 2.83 2 1 2.24 2 2 0.12 3 2 0.12 3 3 0.00
下载: 导出CSV
表 3 验证数据补偿前后艾伦方差-升温速率1℃/min
Table 3. Comparison of Allen variance before and after compensation for validation data-heating rate 1℃/min
验证数据 神经网络 提升百分比/% 分段线性拟合 提升百分比/% $ N(^{\circ}/{s}^{\frac{1}{2}}) $ 6.70e−3 1.17e−3 82.59 6.70e−3 −0.02 $ B(^{\circ}/s) $ 1.27e−3 1.17e−4 90.82 1.76e−3 −38.32 $ K(^{\circ}/{s}^{\frac{3}{2}}) $ 1.65e−4 6.88e−6 95.84 2.03e−4 −22.73 $ R(^{\circ}/{s}^{2}) $ 4.47e−6 2.04e−7 95.44 8.40e−6 −87.79
下载: 导出CSV
表 4 验证数据补偿前后艾伦方差-升温速率2℃/min
Table 4. Comparison of Allen variance before and after compensation for validation data-heating rate 2℃/min
验证数据 神经网络 提升百分比/% 分段线性拟合 提升百分比/% $ N(^{\circ}/{s}^{\frac{1}{2}}) $ 6.76e−3 1.76e−3 73.99 6.77e−3 −0.13 $ B(^{\circ}/s) $ 1.42e−3 1.05e−4 92.62 2.05e−3 −43.82 $ K(^{\circ}/{s}^{\frac{3}{2}}) $ 1.37e−4 7.59e−6 94.48 2.51e−4 −82.55 $ R(^{\circ}/{s}^{2}) $ 5.84e−6 2.45e−7 95.81 1.21e−5 −106.85
下载: 导出CSV
表 5 验证数据补偿前后全温下的零偏稳定性− 1 s平滑（°/h）
Table 5. Zero bias stability at full temperature before and after compensation for validation data −1 s smoothing
升温速率 验证数据 神经网络 提升百分比/% 分段线性拟合 提升百分比/% 1℃/min 37.64 27.67 26.47 40.42 −7.38 2℃/min 35.18 27.74 21.14 43.56 −23.81
下载: 导出CSV
-
[1] ZHANG X, ZHOU C, CHAO F, et al. Low-cost inertial measurement unit calibration with nonlinear scale factors[J]. IEEE Transactions on Industrial Informatics,2022,18(2):1028-1038. DOI: 10.1109/TII.2021.3077296. [2] ZHAO W L, CHENG Y X, ZHAO S H, et al. Navigation grade MEMS IMU for a satellite[J]. Micromachines,2021,12(2):151. DOI: 10.3390/mi12020151. [3] FU Z X, ZHANG G Y, LIN Y R, et al. Calibration and compensation of inertial sensor errors in portable applications—a review[C]//2016 UKACC 11th International Conference on Control (CONTROL). Belfast: IEEE, 2016: 1-4. [4] JOHNSTON W, O’REILLY M, ARGENT R, et al. Reliability, validity and utility of inertial sensor systems for postural control assessment in sport science and medicine applications: a systematic review[J]. Sports Medicine,2019,49(5):783-818. DOI: 10.1007/s40279-019-01095-9. [5] CHANG H C, HSU Y L, YANG S C, et al. A wearable inertial measurement system with complementary filter for gait analysis of patients with stroke or parkinson’s disease[J]. IEEE Access,2016,4:8442-8453. DOI: 10.1109/ACCESS.2016.2633304. [6] KOS M, KRAMBERGER I. A wearable device and system for movement and biometric data acquisition for sports applications[J]. IEEE Access,2017,5:6411-6420. DOI: 10.1109/ACCESS.2017.2675538. [7] CAO H L, LI H S. Investigation of a vacuum packaged MEMS gyroscope architecture’s temperature robustness[J]. International Journal of Applied Electromagnetics and Mechanics,2013,41(4):495-506. DOI: 10.3233/JAE-131668. [8] FU Q, DI X P, CHEN W P, et al. A temperature characteristic research and compensation design for micro-machined gyroscope[J]. Modern Physics Letters B,2017,31(6):1750064. DOI: 10.1142/S0217984917500646. [9] CUI M, HUANG Y, WANG W, et al. Mems gyroscope temperature compensation based on drive mode vibration characteristic control[J]. Micromachines,2019,10(4):248. DOI: 10.3390/mi10040248. [10] YANG W B, FANG B, TANG Y Y, et al. A temperature compensation model for low cost quartz accelerometers and its application in tilt sensing[J]. Mathematical Problems in Engineering, 2016,2016:2950376. DOI: 10.1155/2016/2950376. [11] FENG Y B, LI X S, ZHANG X J. An adaptive compensation algorithm for temperature drift of micro-electro-mechanical systems gyroscopes using a strong tracking Kalman filter[J]. Sensors,2015,15(5):11222-11238. DOI: 10.3390/s150511222. [12] SHEN C, SONG R, LI J, et al. Temperature drift modeling of MEMS gyroscope based on genetic-Elman neural network[J]. Mechanical Systems and Signal Processing,2016,72-73:897-905. DOI: 10.1016/j.ymssp.2015.11.004. [13] XING H F, HOU B, LIN Z H, et al. Modeling and compensation of random drift of MEMS gyroscopes based on least squares support vector machine optimized by chaotic particle swarm optimization[J]. Sensors,2017,17(10):2335. DOI: 10.3390/s17102335. [14] RHODES C, MORARI M. The false nearest neighbors algorithm: an overview[J]. Computers & Chemical Engineering,1997,21(S1):S1149-S1154. DOI: 10.1016/S0098-1354(97)87657-0. [15] WANG J S, HSU Y L, LIN H Y, et al. Minimal model dimension/order determination algorithms for recurrent neural networks[J]. Pattern Recognition Letters,2009,30(9):812-819. DOI: 10.1016/j.patrec.2008.05.007. [16] CHO K, VAN MERRIËNBOER B, BAHDANAU D, et al. On the properties of neural machine translation: encoder-decoder approaches[C]//Proceedings of SSST-8, Eighth Workshop on Syntax, Semantics and Structure in Statistical Translation. Doha: ACL, 2014: 103-111. [17] CHUNG J, GULCEHRE C, CHO K, et al. Empirical evaluation of gated recurrent neural networks on sequence modeling[Z]. arXiv: 1412.3555, 2014. [18] MIAO Z Y, SHEN F, XU D J, et al. Online estimation method of Allan variance coefficients for MEMS IMU[J]. Journal of Instrumentation,2014,9:P09001. DOI: 10.1088/1748-0221/9/09/P09001. -
下载:
点击查看大图
计量
- 文章访问数: 0
- HTML全文浏览量: 0
- PDF下载量: 4
- 被引次数: 0
