MEMS惯性传感器阵列系统搭建及数据融合技术研究

发布时间：2018-03-13 07:39

本文选题：传感器阵列　切入点：数据融合　出处：《昆明理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：本论文主要围绕MEMS惯性传感器阵列及数据融合技术介绍了一系列相关基础知识、解决了一系列问题。在MEMS技术应用越来越广泛的今天,MEMS惯性传感器成本低、耐用、体积小、易集成、误差基本满足要求的特点使得其占据了非常大的市场。MEMS惯性传感器主要包括两种传感器,即MEMS的加速度计和陀螺仪,有时还包括磁力计。由于技术逐渐成熟,现如今几乎所有的手机都配有加速度计,有的还配有陀螺仪,或者直接配有集成了多个传感器的惯性传感器,而这些惯性传感器多数是MEMS惯性传感器。如苹果公司的iPhone6手机内部集成了 InvenSense的六轴惯性传感器MPU-6700(集成了 MEMS陀螺仪和MEMS加速度计)、Bosch公司的三轴加速度计BMA280和AKM公司的AK8963C磁力计,即使用了两个加速度计、一个陀螺仪和一个磁力计。除了手机还有很多其他电子设备配有MEMS惯性传感器,如VR、游戏手柄、汽车、飞行器、航天器等等。在需求越来越多的情况下,MEMS惯性传感器的误差问题变得越来越重要。本论文正是针对这一问题提出了使用传感器阵列的数据融合技术,该技术可以有效的降低低成本MEMS惯性传感器的误差,为系统提供更为可靠的数据。围绕这一技术,作者搭建了多传感器阵列平台,并使用MATLAB软件作为分析、建模工具(第二章);使用Allan方差分析法对传感器的噪声进行识别,并作为判定数据融合模型优略的一个指标(第四章);通过构建状态方程和更新方差,作者建立了三个用于数据融合的Kalman滤波器(第六章),并通过实验对三个模型进行了验证比较(第七章)。其中,Allan方差分析法可用于辨识导致数据噪声的底层随机过程的特征,即通过对数据的整个长度执行某些操作来辨识惯性传感器数据中的噪声种类及其强度。除了上述工作,作者还分析、总结了 MEMS惯性传感器的常见误差来源(第三章),介绍了 ARMA模型的理论基础并给出了 MATLAB软件内ARMA模型的应用举例(第五章)。在文章最后的第八章,作者对所做工作做出了总结和展望。
[Abstract]:This paper mainly introduces a series of basic knowledge about MEMS inertial sensor array and data fusion technology, and solves a series of problems. Easy to integrate, error basically meet the requirements of the characteristics that make it occupy a very large market. MEMS inertial sensors mainly include two kinds of sensors, that is, accelerometers, gyroscopes, and sometimes magnetometers. Today, almost all mobile phones are equipped with accelerometers, some with gyroscopes, or with inertial sensors directly integrated with multiple sensors. Most of these inertial sensors are MEMS inertial sensors. For example, Apple's iPhone6 phone has InvenSense's six-axis inertial sensor MPU-6700 (integrated MEMS gyroscopes and MEMS accelerometers, BMA280 and AKM). Company's AK8963C magnetometer, Even with two accelerometers, a gyroscope and a magnetometer, there are many other electronic devices besides mobile phones that have MEMS inertial sensors, such as MEMS, game handles, cars, aircraft. The error problem of MEMS inertial sensors becomes more and more important in the case of increasing demand. In this paper, the data fusion technology using sensor array is proposed to solve this problem. This technique can effectively reduce the error of the low cost MEMS inertial sensor and provide more reliable data for the system. Around this technology, the author built a multi-sensor array platform, and used MATLAB software as the analysis. Modeling tool (Chapter 2); using Allan variance analysis method to identify the sensor noise as an index to determine the data fusion model excellence (Chapter 4th); by constructing the equation of state and updating the variance, The author establishes three Kalman filters for data fusion (Chapter 6th), and verifies and compares the three models by experiments (Chapter 7th), in which the method of variance analysis can be used to identify the characteristics of the underlying stochastic processes that lead to data noise. That is, by performing some operations on the entire length of the data to identify the noise types and their intensity in the inertial sensor data. This paper summarizes the common error sources of MEMS inertial sensors (Chapter 3), introduces the theoretical basis of ARMA model and gives an example of the application of ARMA model in MATLAB software (Chapter 5th, Chapter 8th). The author makes a summary and prospect of the work done.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP212

【参考文献】