QQ客服 官方微博 反馈留言

题名	挖掘信息提高GPS/INS导航精度的理论与方法研究
作者	柴艳菊
学位类别	博士
答辩日期	2008-06
授予单位	中国科学院测量与地球物理研究所
授予地点	武汉
导师	欧吉坤
关键词	GPS/INS组合 doppler测速 模糊度动态解算 坐标约束 模糊度检验 不适定问题 kalman滤波 参数可测性 挖掘隐含信息 附加隐含信息约束的滤波模式 自适应选权滤波 姿态检验 MATLAB语言
学位专业	大地测量学与测量工程
中文摘要	INS和GPS作为两个主要的导航系统，其优缺点具有明显的互补特性，二者组合不仅可以克服INS误差快速积累、GPS导航不完整等问题，而且为用户提供高采样率、高可靠性、参数丰富的导航服务，成为目前许多领域研究的热点问题。尽管GPS/INS组合系统是近二十年来才发展起来的一项新技术，但是目前关于组合系统的文献非常丰富。但是以往的研究多集中在GPS动态数据处理、滤波算法的改进等方面，而对常用组合模型存在的固有缺陷分析比较少。鉴于目前我国INS技术水平和GPS的发展状况，本文主要研究中低精度INS与GPS的组合。研究内容主要涉及GPS模糊度动态解算方法及其模糊度正确性检验的理论与方法研究；根据系统的特点和具体的运动环境等，如何挖掘系统存在的隐含信息，进一步改进数据融合模型的研究；针对GPS/INS滤波的特殊情况，选权自适应滤波方法如何实现等方面。概括起来，具体研究工作主要集中在以下几个方面：一、 研究背景及相关的基础研究工作的总结，主要包括GPS/INS组合系统研究的意义、组合系统数据融合处理的基本理论、模型及方法、提高组合系统导航精度的现有的主要方法及其优缺点，进而论述了本文在解决提高组合系统导航精度方面的切入点。二、 具体研究工作的系统介绍，主要包括： 1、 在GPS动态模糊度解算方面： ① 如何获取足够精确的动态近似坐标并确定合理的约束权阵，提高模糊度动态解算的成功率； ② 模糊度动态解算结果如何检验； 2、 在GPS/INS数据融合的理论与方法方面 ① 研究了根据组合滤波模型的增益矩阵讨论参数的可测性； ② 针对组合系统的固有缺陷，深入研究了挖掘系统的隐含信息并对其建立函数模型和随机模型的理论与方法； ③ 系统研究了如何将隐含信息纳入滤波观测模型，以提高部分参数的可测性的新的组合模式； ④ 针对组合系统的直接可测参数和间接可测参数，详细研究了如何将自适应选权滤波方法拓展应用到GPS/INS数据融合处理中； ⑤ 根据系统的现有信息，研究了姿态参数的合理性检验方法； ⑥ 编制了组合数据处理的初步软件； ⑦ 利用实测地面GPS/INS组合数据对本文提出的新算法进行检验。 1. 提出了附加doppler测速预报坐标约束的模糊度动态解算方法组合系统中，为了对INS的误差进行准确估计，GPS模糊度必须实现实时解算，而GPS模糊度实时解算的主要困难是相位观测方程秩亏，必须附加足够准确的坐标约束条件。本文根据doppler高精度测速特点，在模糊度初始化后，提出采用对doppler测速预报的高精度近似坐标附加约束，一方面有效解决了相位观测方程秩亏问题，另一方面相位观测方程在比较准确的近似坐标处线性化，大大降低了方程的线性化误差。通过与经典的伪距和相位联合单历元模糊度解算及组合中INS辅助GPS模糊度解算方法进行比较，新方法在提高模糊度固定的成功率方面具有明显优势。 2. 提出了模糊度动态解算结果的检验新方法模糊度固定后，如何检验其正确性是模糊度动态解算中的又一关键问题。通过对目前的检验标准ratio值进行分析可知，ratio值实质上是一个统计指标，其成立的前提是样本趋于无穷大，因此利用这一指标检验静态模糊度解算的正确性是合理的、可靠的。但是在模糊度动态解算中，由于观测信息比较少，不满足统计检验的前提条件，ratio值检验不一定完全可靠。基于此，本文提出利用模糊度固定后的定位结果反求速度和doppler测速结果进行比较，并结合相邻历元模糊度结果的综合判断方案。 3. 提出了参数可测性判断的新方法通过分析以往判断参数可测性指标存在的不足，提出从滤波的关键矩阵——增益矩阵入手对各参数的可测性进行研究，并对各参数解的特性进行了深入分析。从滤波解的特性分析可知，对间接可测的参数的估计依赖可测参数的滤波解和状态参数预报方差阵，而状态参数预报方差阵难以准确确定是引起间接可测参数估值不准的主要原因。 4. 根据系统的特点，提出了附加系统隐含信息约束的新的组合模式分析了常用组合滤波模型的主要缺陷是待估参数多于观测个数的秩亏问题，如何增加观测信息提高参数的可测性是本文的主要研究思想。通过对组合系统的特点分析可知，输入组合滤波器的外部观测信息除了GPS直接观测信息（包括原始观测值或定位、测速结果）外，系统中存在一些有用的隐含信息，如载体系中速度满足X、Z方向为0、运动情况下观测卫星多于3颗时，GPS可以计算比较准确的方位角信息、车辆在平坦的路面上运动时，高程变化为0等。这些隐含信息与间接可测参数有关系，本文详细介绍了这些隐含信息的函数模型及其精度估计方法。针对本文所用的GPS/INS数据中GPS的观测情况，提出了分两种情况在观测方程中附加约束的思想。一种是当卫星多于3颗时，附加GPS测方位约束的松散组合模型；另一种是卫星少于4颗时，附加速度和高程约束的紧密组合模型，以增加部分状态参数的可测性。不同的观测情况和运动状态，系统中蕴含的隐含条件也不同，关键是如何根据系统具体情况挖掘出隐含信息并对其建模和精度估计。 5. 自适应选权滤波算法的拓展及实现针对组合系统中待估参数多于观测个数的秩亏问题，本文深入研究了自适应选权滤波方法在GPS/INS组合中的应用，主要包括观测信息权阵（或方差阵）的确定和状态预报信息等价权因子阵的确定。提出了对观测信息和附加约束信息定权根据GPS观测精度、观测结构、高程变化等情况综合定权；对于状态预报信息，首先将状态参数进行分类，对于完全可测参数（如位置误差参数和速度误差参数等）等价权因子阵依据当前历元观测量和状态预报方差阵来确定，对间接可测的参数（元件误差参数）等价权因子阵依据运动的机动性、采样间隔等因素来确定。当采样间隔比较短时，可以认为元件误差的变化比较小，对此不进行自适应估计。 6. 姿态合理性检验研究姿态矩阵校正后的合理性检验一直是组合系统中比较困难和棘手的问题，尤其对于中低精度的INS。通过对以往姿态检验方法存在的不足以及载体系速度的特点进行了分析，指出当INS导航误差经过校正后，速度精度接近GPS测速精度，约为cm/s级，如果姿态矩阵正确，将其转换至载体系后X方向的速度应为cm/s量级，否则，如果载体系X方向的速度比较大，其误差主要是由姿态矩阵引起的。本文以此提出根据载体系X方向速度为0的特点来检验姿态矩阵是否合理的方法。 7. 组合数据处理软件的研制本文基于MATLAB语言和VC++语言编写了GPS/INS组合数据处理软件，该软件具有GPS单点定位测速和差分定位测速功能、INS单独导航功能、GPS/INS组合导航功能、结果浏览和绘图功能等，界面友好，操作简便，但是还需要进一步优化和改进。
英文摘要	GPS/INS integrated system is very popular in many fields because this system not only overcomes the shortcomings that the INS navigation errors are accumulated quickly and the GPS positions are discontinuous because the signals are disturbed or sheltered, but also provides the users with navigation service of high frequency, high reliability and full navigation information. Though the system is developed only for about twenty years, its productions are very fruitful. The previous researches mainly focus on the GPS data processing, kalman filtering improvement and so on, but the analysis on the inherent limitation of the common models including the loosely coupling and the tightly coupling is very lack. Based on the INS technical level and GPS development in our country, the dissertation mainly researches on the medium precision or low-cost INS/GPS integration. The main contents involve the theory and method about the GPS On-The-Fly ambiguity resolution and its test. According to the character and actual dynamics of the integration system, how to dig the hidden conditions in the system and improve the data fusion models using these hidden conditions. For the especial filtering model of the GPS/INS, how to implement the adaptive kalman filtering by selecting the parameter weights. Generally, the detailed works are as follows: First part: the systematic introduction and summarization of the background and the previous research results of this work. This part mainly includes the research significance of the paper, the base theory, model and algorithm of the GPS/INS and the existing methods for improving the GPS/INS precision and the virtues and drawbacks of these methods. Second part: the systematic introduction of the main contributions and research methods of this paper. One is the GPS On-The-Fly ambiguity resolution. ① How to obtain the enough accurate dynamics coordinates and determine the reasonable weight matrix of them, advance the success rate of On-The-Fly ambiguity resolution. ② How to test the validities of fixed ambiguities. The other is the data fusion theory and method of the GPS/INS ① The parameter observability is discussed according to the gain matrix of the filtering. ② How to dig the hidden conditions in the system and build their function and stochastic models. ③ How to bring these hidden conditions into the observation models and construct the new fusion modes for improving the observability of part parameter is deeply researched. ④ How to extend the adaptive kalman filtering by selecting the parameter weights to the GPS/INS integration is researched. ⑤ The method how to verify the correctness of the attitude is researched without the additional information. ⑥ The software for the GPS/INS data processing is written with the MATLAB and VC++ language. ⑦ The new methods proposed in the paper are tested using the ground GPS/INS data. 1. Present On-The-Fly ambiguity resolution by restricting coordinates predicted by Doppler. In the GPS/INS integration, the GPS On-The-Fly ambiguity resolution is essential for accurately estimating the errors of INS. The main difficulty of kinematic positioning using carrier phase observations is that the norm equation is rank deficiency, so the enough precision coordinate constraints must be added. According to the character of high precise velocity calculated from doppler observations, a new method by restricting the coordinates predicted by the velocity calculated from doppler observations after the ambiguity initialization is put forward. This method has two values. One is that the rank deficiency problem is overcome effectively. The other is that the linearization errors of the phase observation equations are reduced greatly because the approximate coordinates are accurate. By comparing the ambiguity results of the new method with that of the method combining the carrier phase observations with the code observations and that of the method by aiding the INS, the success rate of ambiguity resolution is advanced with the new method. 2. Propose a new method for testing the validities of the fixed ambiguities The other key problem is how to test that the fixed ambiguity is accurate. In generally, ratio value is taken as the index. It is materially a statistic index based on the sample size tending to infinite, so it is reasonable and reliable for verifying the fixed ambiguity of the static positioning but not always reliable for the kinematic positioning because the magnitude of the observations is very small. The author put forward a synthetical test index that the velocities calculated from GPS positions after the ambiguities being fixed are compared with that calculated from the doppler observations, at the same time, combining with the ambiguity results of the previous and next epochs. 3. Propose a new idea for judging the parameter observability Through analyzing the shortcoming of the previous method for judging the parameter observability, a new idea is proposed that parameter observability is judged from the gain matrix of the filtering, and the characteristic of the parameter estimator is deeply analyzed. The filtering estimators of the indirect measurable parameters are determined by the estimators of the direct measurable parameters and the predicted covariance matrix. it is well-known that the estimators of the direct measurable parameters can be estimated accurately by GPS observations but it is very difficult that the covariance matrix is predicted precisely, so the incorrect estimators of the indirect measurable parameters are mainly influence by the predicted covariance matrix 4. Propose the new data fusion models by adding the restrictions of the hidden conditions dug from the system The drawback of the classical integrated models is rank deficiency because the number of the parameters is more than that of the observations, so the idea for resolving this problem is how to add observation equations in the filtering model. By analyzing the characteristic of the GPS/INS integration system, some usable conditions are hidden among the system besides the GPS observations which are GPS original observations or the positions and velocities calculated from GPS original observations. For examples, the component velocity in X or Y direction should be zero in body frame. The yaw angle can be calculated using the velocities from GPS when the number of the satellites is more than three and the velocity is not zero. The height change is zero when the ground is comparatively flat. These hidden conditions are related to the indirect measurable parameters, so the method for constructing the functional models and estimating the precision of the hidden conditions is described in detail. According to the GPS observation status, the models added the hidden condition restrictions are divided two cases. One is loose coupling model by adding the yaw angle calculated from the GPS velocity measurement when the number of the satellites is more than three. The other is tight coupling model by adding the velocity in the body frame and height change restrictions when the number of the satellites is less than four. The hidden conditions vary with the GPS observation status and dynamics, so the key is how to dig them according to the actual cases. 5. Extend the adaptive filtering by selecting the parameter weights to GPS/INS system Aiming at the coefficient matrix of the observation equations is rank-deficient in GPS/INS system, the author deeply researches how the adaptive kalman filtering by selecting the parameter weights is implemented, which mainly includes that the weights or variances of the observations and predicted states are determined. The paper puts forward that the weights or variances of the observations are determined according to the observation precision, GPS structure, height change and so on. But the weights or variances of the predicted states are more difficultly determined. At first, the states are divided into two classes, which are full measurable parameters and indirectly measurable parameters. The equivalent weight factor of the full measurable parameter, such as position error and velocity error, is determined according to the magnitude of observation and predicted covariance matrix of the current epoch in kalman filtering. But the equivalent weight factor of the indirectly measurable parameter, such as sensor error and attitude error, is determined according to the dynamics, trajectory and sampling space and so on. When the sampling space is comparatively short, the sensor error changes very little, so the equivalent weight factor of the parameter equals one. 6. Propose a new idea for verifying the attitude rationality How to verify the attitude rationality is always comparatively difficult after the attitude matrix is corrected, especially for the low-cost INS. In the dissertation, a new verifying idea is put forward that the X component of the velocity in body frame is zero. We all know, the velocity precision of INS should be close to that of GPS after velocity corrected, namely cm/s level, so X component of the velocity in body frame should be close to cm/s level if the attitude matrix is correct. Otherwise, X component error of the velocity in body frame is brought by attitude matrix error. 7. Write the data processing software for GPS/INS The data processing software for GPS/INS is written using MATLAB and VC++ languages. This software has the following functions, which are GPS point and double-differenced positioning and velocity, only INS navigation, GPS/INS loose coupling and tight coupling, result display and plot. Its virtues are the friendly interface and the convenient operation. But it needs to be further optimized and perfected
语种	中文
公开日期	2013-01-17
内容类型	学位论文
源URL	[http://ir.whigg.ac.cn//handle/342008/3689]
专题	测量与地球物理研究所_学生论文_学位论文
推荐引用方式 GB/T 7714	柴艳菊. 挖掘信息提高GPS/INS导航精度的理论与方法研究[D]. 武汉. 中国科学院测量与地球物理研究所. 2008.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。