【作者】 梁冰；

【导师】 王慧强；

【作者基本信息】 哈尔滨工程大学 ， 计算机应用技术， 2008， 博士

【摘要】 多传感器融合是指为了完成需要的识别、估计和决策等任务,对从时间序列上获得的各种传感器数据以及从其他途径获得的信息按照一定的准则进行综合分析与判断的处理过程。目前,对多传感器融合的研究主要集中在融合功能与融合方法方面,强调数据融合的功能模拟与融合处理方法,却忽略了融合的整体过程与计算问题,至今仍然缺乏较为有效的多传感器融合过程的计算模型。这一理论上的不足阻碍了对MSF的深入认识,无法对融合过程作出综合分析与评估,无法消除融合系统设计的盲目性。由于多传感器融合处理信息的多样性和处理过程的复杂性,目前还缺乏有效的数学工具对多传感器融合的处理过程进行统一描述和分析,开发深层次的数学工具对多传感器融合进行描述和分析成为急于解决的问题。本文从多传感器融合的计算问题出发,针对融合过程与处理功能的数学特征,采用自动机理论和方法描述多传感器融合的计算结构和处理过程,建立多传感器融合的计算模型,并对其进行分析与验证。所作的工作主要体现在以下几个方面:第一,通过引入时间序列扩展了有限自动机理论,提出了时序I/O自动机作为多传感器融合的描述、建模和分析的数学工具,时序I/O自动机比传统的自动机能够更准确地描述数据融合过程,并提出了时序I/O自动机的建模、分析及仿真方法。第二,建立多传感器融合的时序I/O自动机模型,采用时序I/O自动机的字符集描述融合处理的时序数据,提炼数据融合处理过程中关键处理状态和信息事件;利用时序I/O自动机模型输入语言和输出语言的并、连接运算表示对各传感器节点数据的时空域融合处理。通过检测多传感器融合的时序I/O自动机模型状态序列的可达性,验证并分析了多传感器融合处理过程的正确性和实时性。第三,数据关联是多传感器融合的主要处理模块,又是一种数据处理方法。利用时序I/O自动机对观测数据的获取、处理过程进行了形式化描述,建立了数据关联的时序I/O自动机模型,并提出一种基于时序I/O自动机的数据关联性能评价算法,通过宽度优先搜索时序I/O自动机模型的状态转移图验证数据关联模块生成航迹的正确性,并讨论了算法实现的关键技术。第四,传感器管理是多传感器融合过程的控制模块。当某些传感器出现故障时,本文对传感器控制策略进行分析,并利用自动机产生的语言测度实现传感器控制策略性能的评价。建立了描述传感器管理模块对多传感器逻辑控制行为的自动机模型;通过对可控制事件的非使能设定,描述不同的控制策略;并提出语言测度参数的递归估计方法与停止规则,根据语言测度获取不同传感器控制策略的性能指标。第五,采用时序I/O自动机模型产生语言的信息熵评价数据融合过程性能。多传感器融合的时序I/O自动机模型中输入信息和输出信息作为两种不同的信息源,利用信息源概率空间上的信息熵推导时序I/O自动机模型上的融合熵,定量描述了多传感器融合对降低决策结果不确定性与增加信息量的有效性。最后验证了多传感器融合的时序I/O自动机模型融合熵分析方法的正确性。第六,将基于自动机理论的多传感器融合建模方法应用于C~3I系统的建模和实时性分析。提出基于计算树逻辑CTL的C~3I系统的实时性表达方法,详细描述使用模型检测工具Uppaal对C~3I系统建模及其实时性验证的方法。基于自动机理论的多传感器融合模型采用了客观、准确的数学语言描述多传感器融合的定义、处理过程与计算结构,并能够分析融合结果的正确性、实时性和过程控制等方面的性能,为多传感器融合系统的设计以及融合算法的研究提供了有效的依据。更多还原

【Abstract】 Multi-sensor fusion is a comprehensive processs of measurement, identification and decesion dealing with data information from multiple different sensors and other sources to achieve above tasks. At present, research on multi-sensor fusion mainly concertrates on the fusion function and methods, however, to a large extent it neglects the whole course and computation and there is not any effective method for mult-sensor fusion compuating models, results of which makes the less theorical foundation baffling the development of deep research on it. Besides, it is impossible to describe and evaluate the fusion course and elimate the blindness of fusion design. As for the diversity of informantion disposed and the complexsity of disposing course, there is no effective mathematica tool for us to describe and analyze the multi-sensor fusion course in union. Therefore, it becomes an urgent problem to find out deep and underlying mathematical tool to describe and analyze the problem. From the computation of multi-sensor fusion, this dissertation adopts the automata theory and methods to describe the couse of computation and disposition of multi-sensor fusion by considering the mathematical charactertics of it, and builds the computation model of multi-sensor fusion with its analysis and validation. Aiming above analysis, this dissertation has done relative work as follows.(1) This dissertation developed the automata theory and put forward the temporal I/O automata as the mathematioca tool for describing, analyzing and modeling multi-sensor fusion. Compared with tradional automata, temporal I/O automata is more exact for describe the course of multi-sensor fusion. At last, this dissertation put forward the methods of modeling, analysis and simulation.(2) This dissertation built the temporal I/O automata model of multi-sensor fusion and adopted the character set of temporal I/O automata to describe the time data of fusion course, which was used to abstract the key treatment state and information events of its course. On the other hand, this dissertation made use of the "and" and "join" computation of output languae and input language of temporal I/O automata model. By examining the reachable activity of temporal I/O automata model of multi-sensor fusion, this dissertation validated and analyzed the correctness of the multi-sensor fusion course.(3) This dissertation used the temporal I/O automata to gain and dispose the observational data and built the data correlation temporal I/O automata model, and put forward capability evaluating arithmetic of it, with breadth priority search method to find state transferring figure of temporal I/O automata to validate the correctness of flight path of data correlation modes, and at last the key techniques of computation realized was discussed.(4) Sensor management of the mode of control course of multi-sensor fusion. This dissertation analyzed the sensor control strategies when some senor breaks down, and used the language measurement produced by automata to complete the evaluation of different strategies. The method of quantitative evaluation and comparison of different control strategies for sensor management by a real language measure based on the principles of automata theory is proposed. The experiment validates identification of the language measure parameters and got the performance measure of different control strategies.(5)The measure entropy of language generated by temporal input/output automata is defined, and the entropy theory for the temporal I/O automata model of multi-sensor fusion is built. Furthermore, several important theorems about conditional fusion entropy are proposed. Their directive significance is pointed in multi-sensor fusion theoretical study and practice.(6)Since there was not a systemic theory for analyzing and validating the dynamic behavior of C~3I information system, a timed automata model of C~3I system using formal method was presented. The Computation Tree Logic presents the properties of C~3I information system, then the system’ s key properties of real-time performance was verified by the real-time model checker Uppaal.A precise computation model of multi-sensor fusion based on temporal input/output automata can be demonstrated by practice that the modeling method provides novel approaches to analyze and validate the dynamic behavior of multi-sensor fusion. It is expected that it should lead to the development of tools that could be used by software engineers to formally derive designs of fusion systems.更多还原