【作者】 吴捷；

【导师】 盛焕烨； 王东；

【作者基本信息】 上海交通大学 ， 计算机应用技术， 2010， 博士

【摘要】 计算技术、通信技术、传感技术等新科技的飞速发展,使物联网(The Internet of Things)正在成为现实。射频识别技术(Radio Frequency Identification,RFID)作为物联网的重要支撑技术,受到了广泛关注。以高速发展的RFID技术及其产业化为核心,推广其在物流与供应链、交通、医疗卫生乃至更广阔领域的应用,是打造无所不在的物联网的必然途径。当前RFID应用存在的问题是自动化、智能化程度较低,并没有将RFID技术的潜能完全发挥出来。这一问题存在的深层次原因为:(1)缺乏对RFID应用中各种信息进行有效表示的模型;(2)现有RFID信息不确定性推理模型的表达能力有限,推理准确性有待进一步提高;(3)缺乏对RFID数据进行高层处理的框架。构建情境感知的RFID应用,对解决现有RFID应用在自动化和智能化等方面的不足将起着一定积极作用。RFID技术作为感知情境的主要手段之一,实时感知环境中出现的对象并对其进行跟踪;情境感知计算则将环境中各个对象(计算设备、用户、机器、物等对象)以及物理环境自身智能地连接起来,实现信息的实时扩展,通过信息融合、信息推理使得环境中的各个对象具有了认知能力,并可以根据这些信息进行交流、活动,最终实现机器与机器之间的“对话”,物与物之间的“对话”,最大限度地减少业务操作中的人工干预,提升自动化、智能化水平。为了实现面向RFID应用的情境感知计算,本文从以下几个方面进行了研究讨论:(1)定义了RFID应用中的情境信息,针对其特点,对面向RFID应用的情境模型的表达能力、可验证能力、可扩展能力、存储与查询效率等方面提出了要求。根据这几方面的要求,对六类情境模型的适用性进行比较,通过实验进行分析验证。最终得出面向对象的情境模型是适用于RFID应用的情境建模机制,并以物流与供应链领域中的RFID应用为例进行了建模。所建立的模型能够清晰的表达实体关系、时序信息,支持信息的验证与扩展,通过Hibernate技术可以实现情境信息的快速存储与查询。(2)针对已有对象追踪追溯不确定性推理方法的不足,提出了面向RFID应用的基于隐半马尔可夫模型的追踪追溯不确定性情境推理方法。提供的对象追踪追溯推理包括对象当前位置解释、对象将来位置预测以及确定最佳位置序列等。构建的隐半马尔可夫模型具有更强的表达力,能够表达对象在各位置的驻留时间概率,能够表达对象在某一位置驻留时间为固定长度的情况等。通过实验证明了具有更强表达力的基于隐半马尔可夫模型的情境推理方法比基于隐马尔可夫模型的情境推理方法具有更高的推理准确性,更适合于对RFID应用中的不确定性情境进行推理。(3)阐述了面向RFID应用的情境感知服务框架的设计理念。根据设计理念的要求,提出了用于建立面向RFID应用的情境感知服务框架的可配置、可重构组件模型,总结了基于该组件模型进行系统开发的流程。提出了面向RFID应用的情境感知服务框架的体系结构,并进行了系统实现。通过实验证明了采用所提出的组件模型并不会对系统的性能有较大影响。这一情境感知服务框架与其它已有的情境感知服务框架的不同之处在于:它是面向RFID应用的;它是与现有的RFID体系框架兼容的;它具有灵活性、扩展性、布署时可配置、运行时可重构的特点。(4)在以上工作的基础上,设计和开发了一个情境感知的应用—智能销售环境。描述了该智能销售环境中的区域划分、设备部署以及所实现的主动情境感知与被动情境感知;并通过该应用检验了面向RFID应用的情境感知服务框架的各项特征,充分体现了使用情境感知计算后给RFID应用所带来的自动性、智能性、实时性方面的好处。更多还原

【Abstract】 With the rapid development of computing, communication and sensing technologies, the internet of things is on the way to become a reality. As a key technology of the internet of things, RFID attracts a lot of attentions. Focusing on RFID technology, its industrialization and promoting RFID’s applications in logistics and supply chain, transportation, health care and other industries is the only way to construct a pervasive internet of things. However, the automation and intelligence of applications is limited and the potential of RFID technology has not been fully tapped. The underlying reasons of the problem being: (1) Lack of an effective model to describe different kinds of information in RFID application; (2) Existing probabilistic inference models for RFID data processing is limited in expression and inference accuracy; (3) Lack of a framework to transform low level RFID data to high level.Construction of context-aware RFID applications will play an active role in solving the above mentioned problems. RFID technology is used to sense context, track and trace objects. Context-aware computing combines objects with physical environment intelligently. It provides information collection, fusion and inference so that each object in the environment is capable of cognition. They can communicate with each other and take proper actions, which will minimize human interference and promote automation and intelligence in RFID applications.To realize RFID oriented context aware computing, we undertook the following research:(1) Define context information in RFID application, and analyse its characteristics. The model should be expressive, verifiable and extensible. After comparison of existing models, we conclude that object-oriented context model is more suitable for modeling context information in RFID application. We use the object-oriented approach to model context information in RFID applications for logistics and supply chain management.(2) Present the necessity of probabilistic inference in RFID application. After the analysis of the shortcomings of the existing object tracking and tracing probabilistic inference methods, we put forward a hidden semi-Markov model based probabilistic inference model for RFID applications. Experiments done have proved that the proposed model performs best compared to the others.(3) Describe the design principles of the context-aware framework for RFID applications. Then we propose a service-oriented component-based configurable and reconfigurable model, the development lifecycle for the model and a workload balance mechanism to fulfill the design principles. On the basis of the component model, we designed and developed a context-aware framework for RFID applications. Different from existing frameworks, it is for RFID applications, compatible with existing RFID frameworks, flexible, extensible, deploytime configurable and run-time reconfigurable.(4) We also designed and developed a context-aware application– intelligent retailing environment. Through the application, we discuss the characteristics of the framework, proving that introducing context-aware computing into RFID application will result in considerable benefits.更多还原