Research On Key Techniques Of Context-Aware Computing Based On Ontology

With the wide use of software and hardware technologies in the fields such as sensor technology, sensor networks and embedded system, context-aware computing, being as an important branch of pervasive computing, has gotten rapid development. Context-aware computing aims to take the initiative to perceive the context information around users, to identify the users’ situation and intention through interpreting and reasoning, and to provide users with needed information and appropriate services proactively. This emerging paradigm can relieve users from excessive interactions with various kinds of devices and services in pervasive computing environments, in order to create a user-centric favorable interactive atmosphere and to enhance user experience. In context-aware computing research area, some key problems such as the formal representation of context, knowledge sharing and reuse, interoperability and semantic reasoning, can be effectively solved using ontology technology due to its standardization, formalization, rich semantic expression and reasoning capabilities. Ontology is widely regarded as the most promising technology in context-aware computing area.Although the researches on ontology-based context-aware computing have gotten greater progress, there are still many urgent issues should be addressed. In this dissertation, ontology-based context modeling, context situation reasoning and context-aware services selection, provision and adaptation are discusse d. T he main contents and contributions are given as follows:1. Most of the existing context models were designed for specific domain. They have deficiency in supporting the generality and description of quality of context(QoC). We utilize ontology technology to establish a general, scalable and layered context model which is represented using OWL language. This context model has a strong capability of semantics express, and supports the description of QoC information. It can be easily extended and used for general purpose. A general and computable QoC model is further built using UML language and EMF framework, and QoC information is labeled into the ontology-based context model through OWL built-in annotation proper- ties without extra burden on ontology-based context reasoning. Being as an important basis of this dissertation, the ontology-based context model lays the foundation for the succeeding research work such as context situation reasoning, semantic similarity calculating and context-ware Web services technologies.2. The existing context situation reasoning approaches have deficiency in supporting uncertainty and users’personalization, we propose a hybrid uncertainty context situation reasoning approach. Fuzzy logic theory is introduced into context ontologies to solve the problem that ontology technology has little ability to express and process uncertainty. The basic context situational model which describes the situations in which the average user can be involved, is expressed in SWRL semantic rules, and is adapted to the specific user by exploiting genetic algorithm. Users’ situations can be inferred and identified using a reasoning method that is based on both rule and ontology, and can deal with the uncertainty of context. Through handling certain and uncertain context variables respectively, the complexity of the semantic reasoning can be effectively reduced. The approximate rule matching algorithm based on semantic similarity is designed for the situation that the rule does not match current context. The experiment results show that the proposed approach can improve the effectiveness and timeliness of context situation reasoning.3. In the context-aware systems, context-aware services are often selected for users in terms of manual decision rules. In this dissertation, an automatic decision rule generation approach based on granular computing is proposed for context-aware service selection. In this approach, context history is employed and regarded as a decision information system, equivalence classes are described by granular matrix, relationship matrix is defined for testing and maintaining compatibility. Using logic AND operation, the logic-based knowledge granulation process is transformed into matrix computation, so that granulation can be realized quickly in different levels, and the redundant context attributes and attribute values can be reduced by granularity computation. Based on the reduction results, decision rules could be generated automatically. When context changes, rule-based reasoning technology is used to select suitable services for users in terms of generated rules. Considering the QoC information involved in the service sele- ction process, QoC-driven service provision approach is put forward. Instead of immediately calling and executing the selected services, the service provision mode would be adaptively adjusted according to the QoC to overcome such problems as error automatic operations.4. In order to improve the adaptability of services, an efficient context-aware service adaptation approach is proposed in this dissertation. In this approach, for the service composed of components, service adaptation is realized by dynamically adding, replacing and deleting its components, and it mainl y refers to the following three aspects:1) extend the traditional service description mechanism for supporting the context conditions description; 2) improve the semantic matching relationships, the ontology-based semantic matching can be transformed into numerical calculation, supporting semantic matching of function, context conditions and quality, and components organization can be achieved based on the improved semantic matching relationships, so as to reduce the number of semantic matching and increase efficiency of the component selection; 3) separate the functionality and adaptation concerns at both the application and component levels which can simplify verification of service adaptation. The experiment results show that the proposed approach can flexibly combine with traditional syntactic matching, improve efficiency of the component selection and context-aware adaptation.In this dissertation, some key issues of context-aware computing, including context modeling, context reasoning and context-aware Web service technologies are studied, and some methods and technologies are put forward, which can provide effective guidelines for further research, as well as strong support for development and deployment of context-aware infrastructure and applications, to promote transparent interactions in context-aware computing.