农机多传感器组合导航方法研究

CORC > 沈阳自动化研究所 > 中国科学院沈阳自动化研究所 > 信息服务与智能控制技术研究室

题名	农机多传感器组合导航方法研究
作者	刘晓光
学位类别	博士
答辩日期	2014-05-16
授予单位	中国科学院沈阳自动化研究所
导师	胡静涛
关键词	组合导航 Kalman滤波 H∞滤波混合滤波数据融合
其他题名	Research on Multi-sensor Integrated Navigation Method for Agricultural Machinery
学位专业	机械电子工程
中文摘要	随着应用需求的不断提高，对导航系统定位精度和稳定性的要求越来越高。理想的导航系统能够提供全天候的高精度定位定向信息，具有较好的自主性和隐蔽性而且不受周围环境的干扰。但至今没有任何一种单一的导航系统能够具备所有的优点。为了满足这种高条件的需求，目前组合导航成为当今导航技术研究的热点之一。本文以农业机械作为应用背景，针对农机多传感器组合导航系统及其关键技术展开研究。在惯性传感器误差处理方面，针对微机械惯性传感器的随机误差进行了建模和去噪方法研究。在农机多传感器组合导航方法研究方面，重点研究了农机组合导航系统的建模问题、H∞滤波的自适应问题和H2/H∞滤波的权值增益取值问题。本文的主要研究内容和主要工作包括以下几个部分：第一，首先对组合导航系统的发展概况和现状进行了回顾和介绍，总结了农机导航系统的发展方向。重点针对组合导航系统数据融合方法和农机组合导航系统的国内外研究现状进行了详细的介绍和深入分析。最后给出了本文的主要研究工作和章节安排。第二，提出了基于时间序列分析的惯性传感器随机误差建模方法。惯性传感器的随机误差是影响农机组合导航系统定位精度的一个重要因素。为了提高农机多传感器组合导航系统的定位精度，对惯性传感器的随机误差进行了分析。通过分析惯性传感器的工作原理和误差来源，发现惯性传感器的随机误差源是多样的、变化的。而基于时间序列分析的建模方法具有较大的灵活性，能够较好地描述惯性传感器的随机误差。第三，提出基于改进小波阈值的惯性传感器去噪方法。惯性传感器的测量数据中含有大量的测量噪声，直接影响建立的惯性传感器随机误差模型的精度，而小波分析具有多分辨率的特性，而且能够较好地保留信号的边缘信息，非常适用于处理惯性传感器的数据。在分析常规小波阈值去噪的基础上，本文构造了一种新的阈值函数，提出了一种改进的小波阈值去噪方法。该方法能够有效的去除惯性传感器的测量噪声。第四，提出了基于农机运动学分析的多传感器组合导航方法。根据农机运动学分析和建立的惯性传感器误差模型，建立了考虑惯性传感器随机误差的农机组合导航系统的模型。为了验证所提出的方法，搭建了农机组合导航系统仿真试验平台。并利用提出的方法在农机多传感器组合导航试验平台上进行了验证。第五，提出了基于自适应H∞滤波的多传感器组合导航方法。分析了Kalman滤波和常规H∞滤波的不足，针对常规H∞滤波参数固定不变的问题，基于滤波新息与H∞滤波参数之间的关系，推导了H∞滤波器参数的自适应方法。该方法克服了常规H∞滤波具有较大保守性的问题，从而具有更好的滤波精度。针对Kalman滤波方法与H∞滤波方法的特点，提出了基于H2/H∞滤波的组合导航方法。该方法采用矩阵不等式的有关理论，根据kalman滤波与H∞滤波的估计方差矩阵推导了H2/H∞滤波的增益权值自动调节的迭代方法。该方法能够结合Kalman滤波和H∞滤波的优势，从而兼顾系统的精度和鲁棒性。第六，在上述研究内容的基础上，完成了农机组合导航系统和车载试验平台的设计。基于ISO11783协议完成了农机组合导航系统通信协议和农机组合导航系统的设计。针对GPS导航系统的野值，提出了一种基于农机运动学模型的野值剔除方法。最后，利用农机组合导航系统车载试验平台的实验数据，分别对本文提出的农机多传感器组合导航方法进行了实验验证。实验结果表明本文提出的农机多传感器组合导航方法具有一定的实用性和可行性。
索取号	TP212/L75/2014
英文摘要	With the development of application requirements, the demand of positioning accuracy and stability of are increasing in navigation systems. Perfect navigation system can provide all-weather, autonomous and concealed high precision navigation information, and not be influenced by environmental distraction. However, so far, none of single navigation systems has all the advantages. In order to meet the high demand of navigation system, integrated navigation system becomes a research hotspot. Based on the application of agricultural machinery, the dissertation focuses on researching multi-sensor integrated navigation system and its core technology. In the inertial sensor error processing aspect, the dissertation proposes a stochastic error modeling method of the micromechanical inertial sensor and a micromechanical inertial sensor de-noising method. In the multi-sensor integrated navigation method aspect, The dissertation focuses on the multi-sensor integrated navigation method based on kinematics analysis of agricultural machinery, adaptive H∞filter and adaptive H2/H∞filter. The specific contents and research work of this dissertation are as follows: First of all, the development history and current situation of integrated navigation system are reviewed and introduced, and then the future development orientation of the integrated navigation system is concluded. Research situations of the data fusion method in integrated navigation system and the integrated navigation system of agricultural machinery are detail reviewed and thoroughly analyzed. Then, the main research work and chapters arrangement of this dissertation are described. Secondly, a stochastic error modeling method of inertial sensor based on time series analysis is proposed. Inertial sensor stochastic error is one of the important factors that affect the positioning precision of integrated navigation system. In order to improve the positioning accuracy of agricultural machinery multi-sensor navigation system, the stochastic error of inertial sensor is analyzed. It found that the sources of inertial sensor stochastic error are diverse and changing by analyzing the working principle and error sources of inertial sensor. As the modeling method that has greater flexibility based on time series analysis can have a better description to the inertial sensor stochastic error. Thirdly，a de-noising method of inertial sensor based on improved wavelet threshold is proposed. The measurement value of inertial sensor contains a large number of measurement noise, which directly affect the accuracy of the inertial sensor error stochastic model, while wavelet analysis has the characteristics of multi-resolution and can get better edge information of signal. So the wavelet analysis is very suitable for processing the inertial sensor data. The dissertation establishes a new threshold function based on the analysis of conventional wavelet threshold and proposes an improved wavelet threshold de-noising method. The improved method can remove the measurement noise of inertial sensor effectively. Fourthly, a multi-sensor integrated navigation method based on kinematics analysis of agricultural machinery is proposed. The integrated navigation system is designed by combining the kinematic model of agricultural machinery and the stochastic error model of inertial sensor. In order to validate the proposed method, a simulation platform for integrated navigation system of agricultural machinery is built. The proposed multi-sensor integrated navigation method for agricultural machinery is verified based on the simulation platform. Fifthly, a multi-sensor integrated navigation method based on adaptive H∞ filter is proposed. The dissertation analyzes the shortcomings of Kalman filter and conventional H∞filter. Aiming at solving the problem of conventional H∞filter is more conservative due to the parameter is set manually, the adaptive method for parameter of H∞ filter is derived based on the relationship between filtering innovations and parameter of H∞ filter. The adaptive method which has better filtering accuracy, overcomes the conservative issues of conventional H∞filter. According to the characteristics of Kalman filter and H∞ filter, H2/H∞ filter is proposed. The method derives the gain weight coefficient using the theory of matrix inequalities and estimation variance matrix of Kalman filter and H∞filter based on the least trace criterion. The method can inprove the accuracy and robustness of system by combining the advantages of Kalman filter and H∞filter, Finally, the dissertation has completed the design of integrated navigation system and vehicle experiment platform based on the above research. A communication protocol for integrated navigation system is designed based on ISO11783 protocol. And the integrated navigation system is designed based on the communication protocol. In order to remove outliers in GPS navigation system, an outliers removing method is proposed based on kinematic model of agricultural machinery. Finally, the proposed multi-sensor integrated navigation methods for agricultural machinery are tested using the experiment data of vehicle test platform for multi-sensor integrated navigation system. Experiment results show that the proposed multi-sensor integrated navigation methods for agricultural machinery has good practicality and feasibility.
语种	中文
产权排序	1
页码	108页
分类号	TP212
内容类型	学位论文
源URL	[http://ir.sia.ac.cn/handle/173321/14810]
专题	沈阳自动化研究所_信息服务与智能控制技术研究室
推荐引用方式 GB/T 7714	刘晓光. 农机多传感器组合导航方法研究[D]. 中国科学院沈阳自动化研究所. 2014.