MEMS陀螺仪随机漂移误差补偿技术的研究

发布时间：2019-02-13 06:40

【摘要】：MEMS (Micro Electromechanical System)陀螺仪相比于其他类型的陀螺仪,具有体积小、功耗低、成本低、质量轻、易集成以及生产批量大等优点。由于MEMS陀螺仪具有上述优势,已经在众多民用领域得到广泛应用,如汽车导航、汽车安全气囊脱扣装置、拍照摄像设备的防抖动平台、机器人姿态测量系统、电子玩具、虚拟体感游戏等。在军事领域,可以预见未来的武器系统、无人机等侦查设备必然向着数字化、智能化、小型化和高机动化发展,因此MEMS陀螺仪具有巨大的发展潜力和价值。但是MEMS陀螺仪的测量精度相对较低,这成为微小型导航系统、制导与控制系统等关键技术的发展瓶颈。解决这一问题的途径有两种：第一,从硬件内部结构上针对性地进行分析；第二,从软件层面入手,从算法角度处理。MEMS陀螺仪的随机漂移误差问题是提升测量精度这一课题的重要研究内容。本文从软件层面入手,研究了MEMS陀螺仪的静态输出噪声特性,就MEMS陀螺仪的随机漂移误差补偿技术展开研究。本文介绍了MEMS陀螺仪的若干性能指标,搭建了MEMS陀螺仪漂移数据采集系统,用Allan方差方法对MEMS陀螺仪噪声特性进行了分析。对MEMS陀螺仪输出的非平稳数据进行了预处理,应用拉伊达准则去除奇异点,通过逐步回归方法拟合漂移趋势,将其转换为平稳、零均值的随机漂移数据。根据自相关系数ACF、偏相关系数PCF、AIC等准则选取了时间序列模型和参数,采用基于自回归AR模型的线性离散系统的卡尔曼滤波方法对MEMS陀螺随机漂移进行补偿,MATLAB仿真及Allan方差计算结果表明,该方法能够有效地抑制漂移误差达50%。针对基于自回归AR模型的线性离散系统的卡尔曼滤波方法的不足,应用了AR与SVM(Support Vector Machine)支持向量机混合模型方法。用AR模型描述随机漂移数据中的线性部分,用SVM支持向量机方法对非线性数据进行处理,经过归一化处理、相空间重构、数据训练及数据预测验证等一系列步骤,计算Allan方差及其它性能指标,结果表明：应用AR与SVM支持向量机混合模型方法能够有效抑制漂移误差80%以上,相比于基于时间序列的卡尔曼滤波方法,具有更为良好的去噪效果。
[Abstract]:Compared with other kinds of gyroscopes, MEMS (Micro Electromechanical System) gyroscopes have the advantages of small size, low power consumption, low cost, light weight, easy integration and large batch production. Because of the above advantages, MEMS gyroscope has been widely used in many civil fields, such as automobile navigation, automobile airbag tripping device, anti-jitter platform of camera equipment, robot attitude measurement system, electronic toy, etc. Virtual sense of the game and so on. In the military field, the future of weapon system, UAV and other reconnaissance equipment will inevitably develop towards digitalization, intelligence, miniaturization and high motorization. Therefore, MEMS gyroscope has great development potential and value. However, the measurement accuracy of MEMS gyroscope is relatively low, which has become the bottleneck of the development of micro navigation system, guidance and control system and other key technologies. There are two ways to solve this problem: first, the internal structure of the hardware is analyzed pertinently; Secondly, the problem of random drift error of MEMS gyroscopes is an important part of improving the measurement accuracy from the point of view of algorithm and software. In this paper, the static output noise characteristics of MEMS gyroscopes are studied from the software level, and the compensation technique of random drift error of MEMS gyroscopes is studied. In this paper, some performance indexes of MEMS gyroscope are introduced, the drift data acquisition system of MEMS gyroscope is built, and the noise characteristics of MEMS gyroscope are analyzed by Allan variance method. The non-stationary data from MEMS gyroscope were pretreated, and the singularity was removed by using Ray criterion, and the drift trend was fitted by stepwise regression method, which was transformed into stationary and zero mean random drift data. According to the criterion of autocorrelation coefficient ACF, partial correlation coefficient PCF,AIC and so on, the time series model and parameters are selected, and the Kalman filter method of linear discrete system based on autoregressive AR model is used to compensate the random drift of MEMS gyroscope. The results of MATLAB simulation and Allan variance calculation show that the proposed method can effectively suppress the drift error up to 50. Aiming at the deficiency of Kalman filtering method for linear discrete systems based on autoregressive AR model, the hybrid model method of AR and SVM (Support Vector Machine) support vector machine is applied. AR model is used to describe the linear part of the random drift data, and SVM support vector machine is used to process the nonlinear data. After normalization, phase space reconstruction, data training and data prediction verification, a series of steps, such as normalization, phase space reconstruction, data training and data prediction verification, are presented. The Allan variance and other performance indexes are calculated. The results show that the mixed model method of AR and SVM support vector machine can effectively suppress the drift error by more than 80%, compared with the Kalman filter method based on time series. It has better denoising effect.
【学位授予单位】：东南大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN96

【相似文献】