【作者】 李鹏；

【导师】 黄心汉；

【作者基本信息】 华中科技大学 ， 控制科学与工程， 2010， 博士

【摘要】 随着传感器、计算机以及信息等技术的迅猛发展,各种具有复杂应用背景的多传感器系统不断涌现,使得信息的获取、处理和融合变得多样化。多源信息融合技术如今已在军事和民用领域都得到了广泛的重视和应用,其理论和方法已经成为智能信息处理和控制的重要研究领域。迄今为止,学者们针对实际应用提出了许多信息融合算法,其中的证据理论允许人们对不完善信息进行建模及推理,符合人类思维方式,得到了广泛的推崇。最近提出来的Dezert-Smarandache证据理论(DSmT)可以处理几乎所有类型的信息,其优势更是体现在对不确定、不精确以及高冲突信息的处理上,拥有完备的描述方法,是一种优秀的证据推理理论。本文在简要介绍了Dempster-Shafer Theory (DST)、DSmT和中智理论后,根据信息的特征,对不完善信息进行了归类,并从定量融合和定性融合两个角度对基于证据理论的不完善信息融合技术进行了深入研究,针对不完善信息中的不同种类信息分别提出了相应的融合算法,并对这些算法进行了深入的分析和比较,构建了完整的多源不完善信息融合理论体系,为多源不完善信息融合技术的推广奠定了基础。本文还对证据理论领域现有的融合框架进行了分析,提出了新的兼容其他融合结构的广义融合框架,并在此基础上构建出了广义证据推理机,定义了广义推理机的工作原理,使其可以推广到其他证据理论中。随后本文将DSmT信息融合技术应用于移动机器人中,对声纳数据进行了分析,构建出了二维声纳数学模型,将本文提出的融合算法和广义证据推理机应用于机器人的信息融合系统中,同时提出了单体机器人的Multi-Agent系统结构,实现了移动机器人在复杂动态未知环境下的精确地图构建、定位以及自主导航,验证了信息融合技术在机器人技术应用中的有效性,同时也验证了单体Multi-Agent结构的优越性。同时还尝试构建了多机器人系统,初步实现了多机器人联合地图构建,为多移动机器人探测未知动态环境的深入研究奠定了基础。本文还采用Visual Studio 2008为软件开发平台,结合OpenGL,开发了一套机器人实验及仿真平台,由于其通用性强,操作简便,人机交互界面友好,已成为移动机器人研究的良好工具。更多还原

【Abstract】 With the rapid development of the sensor, computer and information technology, a variety of applications with complex background of multi-sensor system continue to emerge, making the obtaining, processing and integration of the information more diversified. The multi-source information fusion technology has been given more and more attentions and applied in both military and civilian fields. The theories and methods of multi-source information fusion technology has become the important research area of intelligent information processing and control.So far, many researchers have brought forward a lot of information fusion algorithms. Among these algorithms, the evidence reasoning theory allows people to accord their thinking mode to model and reason with imperfect information. It has been widely spread. The recently proposed Dezert-Smarandache Theory is an excellent evidence reasoning theory with the capabilities of dealing with nearly all types of information, especially with uncertain, imprecise and high conflicting information.After briefly introducing the DST (Dempster-Shafer Theory), DSmT and the neutrosophic theory, based on the information characteristics, the imperfect information is classified. The imperfect information technology based on evidence reasoning is studied further with two aspects of quantitative fusion and qualitative fusion. The responsible fusion algorithms for different types of information of the imperfect information are proposed with deeply a analyzing and comparing. The multi-source information fusion technology can be spread well with the complete multi-source information fusion theory system constructed in this article.After analyzing the existent fusion frame in evidence reasoning field, a new general fusion frame compatible with other fusion structures. And the general evidence reasoning machine is constructed, and it can be extended into other evidence reasoning theories with the definitions of its working principles.And then the DSmT information fusion technology is applied to mobile robot. The sonar data are analyzed and the planar sonar mathematic model is constructed. The algorithms and general evidence reasoning machine proposed in this paper is applied to the information fusion system of the robot. The multi-agent system architecture of single robot is proposed, too. With these achievements, the mobile robot can build accurate maps, positioning precisely, and autonomously navigating in complex dynamic unknown environment. The results of the experiments verify the validity of information fusion technology in robotics applications and the superiority of the monomer multi-agent contracture. And the multi-robot system is attempted to be construct. The joint multi-robot map building is achieved basically. A solid foundation for multi-robot detecting unknown dynamic environment has been set up.A mobile robot experiment and emulation platform is developed with Visual Studio 2008 and OpenGL. It has become an excellent tool for mobile robot research with the advantages of versatility, easy operation and friendly Friendly interface.更多还原