【作者】 丁春；

【导师】 冯志勇；

【作者基本信息】 天津大学 ， 计算机应用技术， 2010， 博士

【摘要】 普适计算强调人、计算机以及环境的相互融合,这就对传统的上下文预测提出了挑战。在普适计算环境中,不仅仅是对一个用户空间位置上下文的变化进行预测计算,更要考虑用户交互、上下文变化波动等特性,从而使向用户提供服务的普适计算环境更加智能。本文在经典上下文预测技术的基础上,针对普适计算环境下上下文的有限性、交互性和波动性等特征,对交互和波动情景的有限上下文预测计算这一课题进行了深入研究,取得了下述主要成果:1、提出了有限上下文计算方法,给出了基于深层关系发现的上下文计算策略及具体架构,通过发现上下文相关信息中隐藏的高层抽象关系,突破有限性对上下文计算的局限。分析了上下文预测的规律性与主动性特点,给出了基于情景关系的上下文预测计算策略,从而为实现交互和波动情况的上下文预测提供了前提条件。2、提出了交互上下文预测计算理念,给出了交互上下文的定义、模型及基于本体的建模方法,以及基于交互上下文的基本预测方法。从上下文感知的基础出发,对普适计算中的上下文概念进行了完善与扩展,为普适计算中存在用户交互情况的上下文预测计算的深入研究建立了基础。3、给出了基于规则的交互上下文预测方法,通过分析用户交互情况中上下文变化的影响因素,给出了交互规则的定义与分类,在基于规则的交互上下文预测架构基础上,给出了交互规则的发掘算法和交互预测算法,实现了利用交互过程中的抽象规则对未来上下文的有效预测。4、提出面向波动复杂性的上下文预测方法,提出相似上下文序列的概念,给出了上下文序列相对距离和相似上下文序列的定义,以及相关的上下文预测推理算法和架构。通过计算上下文序列之间的相对距离,确定相似的上下文序列,根据相似的上下文序列进行上下文预测。该预测方法不仅提供了一种度量上下文序列的新方法,而且解决了传统精确匹配预测方法的局限性问题,有效地提高了普适计算中上下文预测的能力,使其具有更强的适应性和实用性。更多还原

【Abstract】 Pervasive computing emphasizes the mutual fusion of human, computer, and environment, that’s a challenge traditional context prediction method. In pervasive computing environment, context prediction not only focus on spatial location of single user, but also need to consider features such as user’s interaction and fluctuation during context changing, so that the provision of services to users in ubiquitous computing environment more intelligent. Based on the traditional context prediction technology, this thesis performs deep research in the subject of limited context prediction computing for interactive and fluctuant scene, which focuses on context characteristics: limited nature, interactivity and fluctuant nature, and gets some results below:1. The limited context computing method is proposed, the context computing approach based on deep relationship finding and its architecture are presented in this thesis. The limitation of context computing brought by limited nature is broke through the discovery of hide high level abstraction relationship in context information. The context prediction characteristics of regularity and proactivity are analyzed. The scene relationship based context prediction strategy is given, so that the condition of context prediction for interactive and fluctuant circumstance is laid.2. The interactive-context predictioncomputing notion is proposed, the definition of interactive-context and its modeling method are introduced, and basic prediction approach based on interactive-context is given. On the base of context awareness, the concept of context in pervasive computing is improved and expanded; the foundation for further study on context prediction of users’interaction is established.3. An approach for interactive-context prediction based on rule is presented. By analyzing the factors affecting user interaction in the context changes, the definition and classification of interactive rule are given. According to the architecture of interactive-context prediction based on rule, the interactive rule mining algorithm and interaction prediction algorithm are proposed, and the future context was predicted effectively by using abstract rules in interaction process.4. An approach for fluctuation complexity oriented context prediction is proposed, the concept of similarity context sequence is introduced. The definitions of context sequence relative distance and the similarity context sequence are given, and related context prediction reasoning algorithm and framework were presented. The similarity context sequence was obtained according to the calculated relative distance between context sequences. Then the future context was predicted based on the similarity context sequence. This prediction approach provides a new measurement method for context sequences and overcomes the limitations of traditional precise matching prediction method. The context prediction capacity in the pervasive computing was effectively improved with more adaptability and practicality.更多还原