【作者】 贺琛；

【导师】 陈肇雄； 黄河燕；

【作者基本信息】 南京理工大学 ， 模式识别与智能系统， 2003， 博士

【摘要】 随着信息技术、芯片技术的迅猛发展以及互联网的个性化、移动化趋势，使得嵌入式Internet系统逐渐成为研究的热点。可以想象，如果嵌入式系统能够连接到Internet上面，则可以方便、低廉地将信息传送到世界上几乎任何一个地方，真正做到随时随地、在任何设备上存取信息，其应用前景十分广阔。所以，对嵌入式系统网络信息获取研究不仅在学术上十分必要，而且在经济上也能取得显著的效益，最终能加速社会信息化的变革。 然而嵌入式系统所处的环境相对恶劣，再加上嵌入式硬件的功耗、体积、成本、可靠性、带宽、速度、处理能力等方面均受到各种制约，这就决定了嵌入式系统无法像通用系统那样方便的获取Internet上的各种信息。同时，现有的浏览器／服务器模式、压缩技术和互联网技术由于不是专门面向嵌入式系统设计，所以目前仍然无法很好的满足嵌入式应用。如何克服上述的弱点，减少网络数据传输、节约有限的带宽、提高响应时间，也就成为了本课题的研究重点和难点。 针对以上问题，本文进行了下列研究工作： 1)提出了基于智能代理架构的嵌入式系统网络信息获取技术，并应用该技术实现了一个智能代理模型系统。该系统分为浏览器端代理(BSB)和服务器端智能代理(SSA)两个物理部分，从而无须改动Web服务器或者浏览器的设置，便于系统部署。从逻辑层次上该系统通过增加表示过渡层和逻辑过渡层将大量的计算和网络流量集中于智能代理-服务器一边，从而减轻了嵌入式设备的计算负担，减少了浏览器到嵌入式设备网关之间的数据传输、节约有限的带宽、提高了响应时间。 2)提出了嵌入式Internet动态页面缓存管理概念，从数学上归纳了动态页面缓存的定义和定理，提出了动态页面的后台生命周期的概念、动态页面的缓存一致性策略和替换算法。同时，设计实现了基于知识库系统的请求分析器、以块为单位的差异提取算法的差分引擎。动态页面缓存管理能大幅减少BSB和SSA之间的传输量，有效的提高了浏览器的响应时间。 3)定义了资源的多态性和保真度以及E值函数模板和约束条件的概念，给出了单一约束条件最优化和多约束条件最优化的解决方案，并设计实现了一种可以人机交互的自适应Web资源优化策略。该策略能较好的表达出制作者的意图，同时又提高了嵌入式浏览器对页面的表现力。 4)提出并实现了一种改进的面向半结构化文本信息的压缩算法：LZWX算法。通过对其中的结构化信息的提取，在压缩时LZWX算法可以充分利用这些信息，通过引入元串和滑动窗口的概念，使得编码／解码不再是以单个字符而是以元串为单位摘要博」:论文进行，扩展了传统的LZW算法。LZWX算法的应用能有效的减少客户端到SSA的传输量，从而减少了带宽的占用和能量消耗。 实验结果表明该系统有效的克服了1沃入式Internet系统的弱点及其在应用浏览器/服务器模式时的局限性。更多还原

【Abstract】 Along with the development of chips and the trends of Internet mobilization and personalization, Embedded Internet System (EIS) is concerned by more and more researchers. We can image that if Embedded Systems (ES) link to Internet, information will be transmitted to any place in the world conveniently and cheaply, which will enable us retrieve information at any time, at any place, and by any equipment with wide prospect. Researching Embedded Internet Information Retrieval (EIIR) is not only very necessary academically but also profitable economically, and will accelerate social renovation of information ultimately.However, ES’ environment is execrable relatively; its hardware energy loss, size, cost, reliability, bandwidth, speed, processing ability, etc are all in the collar. Meanwhile, the Browser/Server mode, the compression technology and today’s Internet technology are not designed for the ES, therefore at present, they can’t satisfy the embedded applications, and neither can ES retrieve information conveniently from the Internet. How to overcome the disadvantages mentioned above are the emphases and the difficulty of our research.This dissertation addresses the above issues and presents the following solutions:1) Proposes a Broker-Agent (B-A)-architecture-based Embedded Internet information retrieval technology, and implements a model system. The system is composed of two physical parts: one is Browser Side Broker (BSB), another is Server Side Agent (SSA), so the system need not change the settings of Web server or browser, and is easy to be deployed. By adding the Presentation Interlayer and the Logic Interlayer from the point of view of logical level, the system can pool large numbers of computing and net flow onto agent-server side, so it can not only minimize transmission data between the browser and embedded facilities’ gateway, and save the limited bandwidth but also improve the response time.2) Proposes the dynamic pages caching management strategy, which induces its definitions, theorems, concept of background Time-to-Live of dynamic pages, cachecoherency and replacement algorithm of dynamic pages. Furthermore, designs and implements a knowledge-based request analyzer, a differencing engine based on the block differences isolating algorithm. Applying this strategy can largely reduce transmission data between BSB and SSA and can improve the response time efficiently.3) Defines the concepts of resources multi- modality, fidelity, E-value function template and constraints, and proposes the solution of single-constraint optimization and multi-constraint optimization. Designs and implements an adaptive Web resources optimal strategy, which can better express the producers’ notion and improve the expressive force of the embedded browser.4) Proposes and implements an improved semi-structured-oriented data compression algorithm: LZWX. LZWX expands some concepts of the traditional LZW algorithm, so that text is not encoded/decoded char by char but meta-string by meta-string. The algorithm can reduce transmission data between browser and SSA, so as to cut down the bandwidth occupation and the energy consumption.Finally, the experiment results indicate that our system can overcome the weakness of EIS effectively and the limitation when applied the Browser/Server mode.更多还原