【作者】 张志芬；

【导师】 柴毅；

【作者基本信息】 重庆大学 ， 控制理论与控制工程， 2008， 博士

【摘要】 数据融合技术是一门新兴综合性技术,在军事和民用上均有广阔的应用前景。航天器多信源测控数据融合受多种因素、多个环节的影响和制约。把系统理论分析和建模数值仿真相结合,逐步拓宽对数据融合的认识和实践,将对我国航天测控技术的发展产生积极意义。本文以航天器多信源测控数据融合与决策技术为研究对象,针对融合技术在测控数据融合过程中存在的问题,研究和探讨解决问题的有效途径,并通过仿真实验对系统设计方案和理论技术进行分析和验证。本文在研究数据融合技术、目标识别技术、特征矢量提取技术和证据推理决策技术的基础上,分别在数据融合的三个层次上完成了相关融合方法研究。为了认识数据融合的整体架构,论文阐述了数据融合的基本概念、基本原理、主要特点;总结分析了数据融合功能模型和结构模型的原理、结构特点及其主要融合算法。以航天器多信源测控为背景,分析了数据融合与决策系统的内容和结构,建立系统的研究框架,阐述了融合模型中各个层次的主要研究内容。数据关联是数据融合的首要环节,文章分析和改进了多信源航天器测量控制的数据关联理论和过程算法,提出了一种跟踪门的形成方法及测控数据关联算法,包括跟踪门的数论法计算、数据关联过程以及修正的联合概率数据关联算法。采用概率均匀采样的方法来提高状态分布模拟采样的代表性,并用于计算跟踪门的大小和形状。该方法适合于三维空间的跟踪门设置,计算过程简单。修正的联合概率关联算法将被跟踪目标的预测区域分成几个彼此相互独立的跟踪区域,对具有公共测量的目标和各目标相关跟踪门内的多个测量的概率分布密度分别进行计算,并进行概率加权,形成修正的联合概率关联算法。通过数值仿真验证,本算法的关联成功率较高。为了提高数据层融合精度,采用缓冲算子和灰色关联理论对测控数据进行预处理,减小测量数据的随机误差,根据测量数据序列的变化特性,采用样条滤波和强跟踪滤波技术对测量序列进行数据层融合。利用四阶样条函数建立系统状态方程,结合系统观测方程,对滤波过程的状态估计误差和观测误差进行补偿,有效地抑制了测量数据的随机观测误差,平滑了测量数据。强跟踪滤波器对模型参数失配问题具有较强的鲁棒性,对初值状态统计特性不敏感,收敛快,关于状态突变具有良好的动态性能,即使受到剧烈扰动,滤波器仍具有保持稳定的能力。以数据层处理结果为输入,提出了测控数据特征值提取的有效方法,即对数据层处理结果先进行检择,以测速定位和相互定位测距为特征,分析了数据融合特征层的计算理论和方法,推导出数据层不同处理结果条件下航天器位置速度的实时计算新方法,并将样条滤波和数论法有效结合,在相对距离最近时刻,对追逐方和逃逸方的位置分布进行伪随机采样,以较高的精度反映了最小相对距离的分布特性和统计特性。论文结合样条滤波,利用追逐逃逸过程仿真数据,对追逐方和逃逸方飞行过程的位置、速度进行了仿真,验证了最小相对距离矢量的计算方法。作为数据融合的应用之一,本文提出了一种证据加权数据融合及判决方法,用于解决证据冲突的合成问题,对航天器型号辨识、测控设备选优、空中目标交汇评估等问题进行了仿真分析和验证。本文根据D-S证据合成的一般过程,结合置信度函数和似真度函数等概念,利用证据加权融合,抑制证据冲突问题,针对型号辨识、测控设备选优和空中目标交汇评估,分别采用不同的加权计算方法,减小了个别误差较大的特征值对总体决策的影响,增强了系统的容错性。更多还原

【Abstract】 Data fusion technology is a new integrated technology, and has a wide applications perspective in both military and civilian. As to spacecraft measurement and control data, Multi-source data fusion is influenced and constrained by many factors. It is significant for the improvements of Chinese spacecraft testing and launching technology. In order to deepen the understanding and practice on data fusion, it is the principle to combine the theoretical analysis and simulation together.In the paper, multi-source data fusion to measurement and control of spacecraft and decision-making technology is explored to solve the problems on the application of data fusion technology. Through simulations and experiments the scheme, theory and technologies are analyzed and validated. This paper introduces some kernel methods on three-level data fusion, which is foundational data fusion technology, target identification technology, feature vector extraction technology and evidence reasoning decision-making technology respectively.To illuminate the main framework, the paper expounds the basic concepts, basic principles and the main features of data fusion, analyzes the principle of function model and structure model, structural features and main algorithms of data fusion. On the background of multi-resource from spacecraft measurement and control, the paper analyzes the structure of data fusion and decision-making system, establishes the research framework of system and expounds the main research contents of every level in the fusion model.With respect to the data from multi-source spacecraft measurement and control, the theory and process algorithm of data association is analyzed and improved. The formation methods of tracking gate and measurement data association algorithms is proposed, which includes the calculation of tracking gate based on number theory method, the process of data association and modified joint probability data association method and so on. The representative of sampling is improved by using the method based on uniform probability sampling, which is used to calculate the size and form of tracking gate. The method is suitable for three-dimensional tracking gate set because of a small amount of calculation and simple process. Modified joint probability data association method divide the intersection region of the tracked target into several independent tracking regions, and then carries on probability calculation of public measurement objects in common space and the probability density of multi-data measured in each correlated gate. The probability density values of multi-measurement data in tracking gate related to every target are weighted by probability. So it gets high precision.Buffer operator and gray association are used to preprocess measurement and control data. It reduces the random error of measurement data primarily. Measurement sequence is fused by using spline filtering and strong tracking filter technology according to variation characteristics of measurement sequence. Fourth-order spline function is used to establish system state equation for spline filtering. It compensates the estimation error and observation error of the system state combining the observation equation, restrains the random observation error of measurement data, smoothes the measurement data. STF is strongly robust to the parameter mismatch of model. It may restrain noise and low initial statistical characteristic, which is of fast convergence. In particular, it has strong track ability to state mutation and is stable in presence of large disturbance.The paper analyses the calculation theory and method of feature layer of data fusion aiming at the measured velocity and position, mutual position range of multi-source spacecraft measurement process, derives a new methods for real-time calculation of spacecraft velocity on the condition of different results in data layer and combines spline filter and number theory method to provide pseudo-random number for the position distribution of the escape and chase when relative distance is closest in order to reflect real-time state distribution precisely.Using the simulation data of the chase and escape, the paper combines spline filter and kalman filter to evaluate the position and velocity of the chase and escape and analyses the estimated result of relative distance vector.Evidence weighting data fusion and adjudgement method are proposed to solve the combination problem of conflict evidence, and validate it by simulation. The evidence decision-making problem in the type identification of spacecraft, the selection of measurement and the assessment of target miss distance is solved. According to the process of D-S evidence reasoning, the paper combines the conception of confidence function and likelihood function, uses evidence weighted fusion restrain evidence conflict. Different weight calculation methods are used in the type identification of spacecraft and the assessment of target miss distance in order to reduce the overall influence due to minor sensors, and to make use of known information furthest and enhance the system’s fault tolerance in bad condition.更多还原