BP神经网络结合粒子群优化卡尔曼滤波的MEMS陀螺随机误差补偿方法

doi:10.12382/bgxb.2022.0110

摘要/Abstract

摘要：

针对微机电系统(MEMS)陀螺仪随机误差相对较大、影响其精度这一问题,提出一种基于BP神经网络结合具有量子行为的粒子群优化(QPSO)算法优化卡尔曼滤波(KF)的补偿方法。采集MEMS陀螺和转台数据作为样本,采用BP神经网络进行训练,建立误差模型;利用训练好的模型对MEMS陀螺进行误差补偿;利用QPSO算法优化KF,以达到更好的降噪效果。实验结果表明,该方法较BP神经网络优化KF、QPSO优化KF与变分模态分解结合小波阈值去噪等方法去噪处理后的平均绝对误差(MAE)和均方误差(MSE)更小,具有更好的降噪效果。

关键词: MEMS陀螺, BP神经网络, 量子粒子群优化, 卡尔曼滤波

Abstract:

To deal with the large random error of the micro-electro-mechanical-system (MEMS) gyroscope that affects its accuracy, an error compensation method based on BP neural network combined with Quantum-behaved Particle Swarm Optimization (QPSO)-optimized Kalman Filter (KF) is proposed. First, the MEMS gyroscope and turntable data are collected as samples, and the BP neural network is employed for training to establish the error model; then the error of the MEMS gyroscope is compensated by the model; finally, the QPSO algorithm is used to optimize KF to achieve better noise reduction effect. The experimental results show that compared with other methods like BP-KF, QPSO-KF and VMD-WTD, this method has better denoising effect, and the MAE and MSE values of the denoised data are smaller.

Key words: micro-electro-mechanical-system gyroscope, BP neural network, quantum-behaved particle swarm optimization, Kalman filter

中图分类号:

V241.5

万芯炜, 王晶, 杨辉, 李毅, 张远再, 王路. BP神经网络结合粒子群优化卡尔曼滤波的MEMS陀螺随机误差补偿方法[J]. 兵工学报, 2023, 44(2): 556-565.

WAN Xinwei, WANG Jing, YANG Hui, LI Yi, ZHANG Yuanzai, WANG Lu. A Random Error Compensation Method of MEMS Gyroscope Based on BP Neural Network Combined with PSO-Optimized Kalman Filter[J]. Acta Armamentarii, 2023, 44(2): 556-565.

图/表 10

参考文献 23

[1]	邱林茂, 高文冀, 李磊, 等. 导引头中MEMS陀螺漂移补偿算法研究[J]. 火控雷达技术, 2013, 42(2):10-14.
	QIU L M, GAO W J, LI L, et al. Study on algorithm for MEMS gyro drift compensation in seeker[J]. Fire Control Radar Technology, 2013, 42(2):10-14. (in Chinese)
[2]	熊必凤. 低成本MEMS陀螺仪随机漂移误差的建模及修正[D]. 重庆: 西南大学, 2017.
	XIONG B F. Research on the modeling and correction technology of random drift error of low cost MEMS gyroscope[D]. Chongqing: Southwest University, 2017. (in Chinese)
[3]	刘孝博, 陈光武, 王迪, 等. MEMS陀螺仪漂移和噪声的分析和补偿[J]. 传感技术学报, 2018, 31(3):368-373.
	LIU X B, CHEN G W, WANG D, et al. Analysis and compensation of drift and noise in MEMS gyroscope[J]. Chinese Journal of Sensors and Actuators, 2018, 31(3):368-373. (in Chinese)
[4]	刘文超, 郑小兵, 王荣颖, 等. 基于角速度估计的MEMS陀螺随机误差动态滤波方法[J]. 电光与控制, 2021, 28(5):79-84.
	LIU W C, DENG X B, WANG R Y, et al. MEMS gyroscope random error dynamic filtering method based on angular velocity estimation[J]. Electronics Optics & Control, 2021, 28(5):79-84. (in Chinese)
[5]	段志强, 刘洁瑜, 汪立新, 等. BPNN辅助KF的MEMS陀螺仪数据处理方法[J]. 压电与声光, 2020, 42(2):284-288.
	DUAN Z Q, LIU J Y, WANG L X, et al. Research on data processing method of MEMS gyroscope based on BPNN assisted Kalman filter[J]. Piezoelectric & Acoustooptics, 2020, 42(2):284-288. (in Chinese)
[6]	程章, 许江宁, 许微, 等. 基于神经网络的MEMS陀螺标定与补偿[J]. 压电与声光, 2018, 40(1):111-114.
	CHENG Z, XU J N, XU W, et al. Calibration and compensation of MEMS gyroscope based on neural network[J]. Piezoelectric & Acoustooptics, 2018, 40(1):111-114. (in Chinese)
[7]	BHATT D, AGGARWAL P, BHATTACHARYA P, et al. An enhanced MEMS error modeling approach based on nu-support vector regression[J]. Sensors, 2012, 12(7):9448-9466. doi: 10.3390/s120709448 pmid: 23012552
[8]	DRAGOMIRETSKIY K, ZOSSO D. Variational mode decomposition[J]. IEEE Transactions on Signal Processing, 2014, 62(3):531-544.
[9]	DING M K, SHI Z Y, DU B H, et al. A signal de-noising method for a MEMS gyroscope based on improved VMD-WTD[J]. Measurement Science and Technology, 2021, 32:095112. doi: 10.1088/1361-6501/abfe33
[10]	刘宇, 付乐乐, 邹新海, 等. 基于RBF神经网络的MEMS惯性传感器误差补偿方法[J]. 重庆理工大学学报(自然科学), 2021, 35(1):197-202.
	LIU Y, FU L L, ZOU X H, et al. Error compensation method for MEMS inertial sensor based on RBF neural networks[J]. Journal of Chongqing University of Technology(Natural Science), 2021, 35(1):197-202. (in Chinese)
[11]	孙伟, 段顺利, 文剑, 等. 阈值去噪与RBF神经网络在MEMS陀螺仪误差补偿中的应用[J]. 传感技术学报, 2017, 30(1):115-119.
	SUN W, DUAN S L, WEN J, et al. Application of threshold denoising and RBF neural network in the error compensating of MEMS gyro[J]. Chinese Journal of Sensors and Actuators, 2017, 30(1):115-119. (in Chinese)
[12]	郭宏伟, 侯宏录, 李光耀. 基于Kalman滤波和神经网络的MEMS陀螺温度漂移补偿[J]. 自动化与仪表, 2020, 35(1):1-4,9.
	GUO H W, HOU H L, LI G Y. Temperature drift compensation of MEMS gyroscope based on Kalman filter and neural network[J]. Process Automation Instrumentation, 2020, 35(1):1-4,9. (in Chinese)
[13]	禹忠, 黄一超, 郭畅. 卡尔曼滤波结合神经网络在MEMS传感器上的应用[J]. 传感器与微系统, 2021, 40(11):154-156,160.
	YU Z, HUANG Y C, GUO C. Application of Kalman filtering combines with neural network in Mems sensors[J]. Transducer and Microsystem Technologies, 2021, 40(11):154-156,160. (in Chinese)
[14]	ZHU C H, CAI S, YANG Y F, et al. A combined method for MEMS gyroscope error compensation using a long short-term memory network and Kalman filter in random vibration environments[J]. Sensors, 2021, 21(4):1181. doi: 10.3390/s21041181 URL
[15]	孙俊. 量子行为粒子群优化算法研究[D]. 无锡: 江南大学, 2009.
	SUN J. Particle swarm optimization with particles having quantum behavior[D]. Wuxi: Jiangnan University, 2009. (in Chinese)
[16]	陈一鸣. 基于卡尔曼滤波的陀螺仪随机误差分析[J]. 电子测量技术, 2020, 43(17):97-100.
	CHEN Y M. Analysis of gyro random error based on Kalman filter[J]. Electronic Measurement Technology, 2020, 43(17):97-100. (in Chinese)
[17]	周克良, 王威, 郭春燕. 基于NAR-KF的心音信号仿真研究[J]. 现代电子技术, 2021, 44(21):35-38.
	ZHOU K L, WANG W, GUO C Y. Research on heart sound signal denoising based on NAR and Kalman filtering[J]. Modern Electronic Technology, 2021, 44(21):35-38. (in Chinese)
[18]	张敏, 李凯, 韩焱, 等. 基于卡尔曼滤波的陀螺仪降噪处理[J]. 传感技术学报, 2018, 31(2):223-227.
	ZHANG M, LI K, HAN Y, et al. The noise reduction of gyroscope based on Kalman filter[J]. Chinese Journal of Sensors and Actuators, 2018, 31(2):223-227. (in Chinese)
[19]	谷鹏, 王大龙, 张世仓. 基于粒子群优化的扩展卡尔曼滤波方法研究[J]. 工业控制计算机, 2019, 32(11):80-82.
	GU P, WANG D L, ZHANG S C. Extended Kalman filter method based on particle swarm optimization[J]. Industrial Control Computer, 2019, 32(11):80-82. (in Chinese)
[20]	杨辉, 杨川, 姜湖海, 等. 基于改进极限学习机的MEMS陀螺随机误差补偿方法研究[J]. 传感技术学报, 2021, 34(2):244-248.
	YANG H, YANG C, JIANG H H, et al. Research on compensation method of MEMS gyroscope random error based on improved extreme learning machine[J]. Chinese Journal of Sensors and Actuators, 2021, 34(2):244-248. (in Chinese)
[21]	WANG X L, CAO H L, JIAO Y Z, et al. Research on novel denoising method of variational mode decomposition in MEMS gyroscope[J]. Measurement Science Review, 2021, 21(1):19-24. (in Chinese) doi: 10.2478/msr-2021-0003 URL
[22]	吴保锋, 夏林元, 吴东金, 等. MEMS陀螺仪随机误差分析与小波阈值去噪研究[J]. 电子设计工程, 2020, 28(15):1-5.
	WU B F, XIA L Y, ZHU H Y, et al. Random error analysis and wavelet threshold denoising of MEMS gyroscope[J]. Electronic Design Engineering, 2020, 28(15):1-5. (in Chinese)
[23]	李世银, 张楠, 武中文, 等. 基于小波分析的MEMS加速度计去噪优化算法[J]. 传感技术学报, 2018, 31(5):705-709.
	LI S Y, ZHANG N, WU Z W, et al. An optimized MEMS accelerometer de-noising algorithm based on wavelet analysis[J]. Chinese Journal of Sensors and Actuators, 2018, 31(5):705-709. (in Chinese)

隐含层节点数	网络学习MSE
2	0.02438
4	0.01924
6	0.01015
8	0.00198
10	0.00178
20	0.00175
40	0.00175
60	0.00174

隐含层节点数	网络学习MSE
2	0.02438
4	0.01924
6	0.01015
8	0.00198
10	0.00178
20	0.00175
40	0.00175
60	0.00174

转速/((°)·s^-1)	Q	R	MSE
1.5	0.0015	6.3651	4.9634×10^-7
-1.5	0.0024	5.7802	4.859×10^-7
3	0.001	6.7147	2.8848×10^-8
-3	0.00201	5.3647	1.1947×10^-7
6	0.0018	5.3162	2.8958×10^-6
-6	0.0012	6.1819	3.6192×10^-6

转速/((°)·s^-1)	Q	R	MSE
1.5	0.0015	6.3651	4.9634×10^-7
-1.5	0.0024	5.7802	4.859×10^-7
3	0.001	6.7147	2.8848×10^-8
-3	0.00201	5.3647	1.1947×10^-7
6	0.0018	5.3162	2.8958×10^-6
-6	0.0012	6.1819	3.6192×10^-6

角速度/((°)·s^-1)	评价指标	原始数据	本文方法	BP-KF	QPSO-KF	VMD-WTD
1.5	MAE	0.1336	6.98×10^-4	4.64×10^-2	3.98×10^-2	4.34×10^-2
	MSE	2.80×10^-2	4.96×10^-7	3.20×10^-3	1.90×10^-3	2.70×10^-3
-1.5	MAE	0.1263	6.87×10^-4	4.01×10^-2	3.35×10^-2	3.69×10^-2
	MSE	2.48×10^-2	4.86×10^-7	2.50×10^-3	1.60×10^-3	2.00×10^-3
3	MAE	0.1327	1.20×10^-4	4.59×10^-2	4.00×10^-2	4.32×10^-2
	MSE	2.75×10^-2	2.89×10^-8	3.10×10^-3	1.90×10^-3	2.60×10^-3
-3	MAE	0.1292	3.30×10^-4	3.52×10^-2	2.57×10^-2	3.12×10^-2
	MSE	2.59×10^-2	1.19×10^-7	1.90×10^-3	9.55×10^-4	1.50×10^-3
6	MAE	0.1265	1.70×10^-3	3.90×10^-2	3.10×10^-2	3.57×10^-2
	MSE	2.51×10^-2	2.90×10^-6	2.30×10^-3	1.30×10^-3	1.90×10^-3
-6	MAE	0.1283	1.90×10^-3	3.91×10^-2	2.98×10^-2	3.57×10^-2
	MSE	2.59×10^-2	3.62×10^-6	2.30×10^-3	1.30×10^-3	1.90×10^-3