Research On Some Key Techniques Of Context-Aware Computing

The development of computation technology makes the computation paradigm adaptable to it. With the occurrence of equipments with computation and network ability, and the trend to become smaller and embedded, the traditional computation paradigm considering one computer as the core of a system no longer meet the needs, because it cannot overcome the difficulty coming from"one person, many computer". Pervasive computation paradigm proposed by Mark Weiser which takes"transparent"and"everywhere availability"as the main characteristic has received widespread attention in recent years since it meets the demand of computation technology."Transparent"in pervasive computing does not only refer to the physical invisibility, but also refers to the interaction invisibility between the human and the computer. Like the human can avoid obvious interaction using contexts, a computer also may use sensed contexts for interaction via some implicit methods. Obviously, this implicit interaction way is a very important way to realize"transparent"in pervasive computing and now it become a hot research topic– context-aware computing.Aimming at the demand of context-aware applications under pervasive environment, this dissertation first gives a literature review on the state-of-the-art of context-aware computing; then we proposes the principle of division of context-aware computing in its narrow sense and broad sense; after that we make a discussion on the context-aware computing modeling, context-aware system infrastructure and its applications. After reviewing the existing context-aware systems, we propose a conceptual model of context-aware computing which can set up a solid base for future research in this field. Conceptually, the existing context-aware models have the drawbacks of high complexity, high cost and high modeling difficulty. To solve these problems, this dissertation proposes a standard modeling methodology which supports fast prototype development. Under the platform of Ptolemy II, based on standardized modeling of contexts, Combined the hierarchization modeling method and finite state machine(FSM) method, the methodology propose a standardized process which include steps like system hierarchization, sub-model refinement and state refinement to support context-aware systems modeling. The simulation results indicate that the modeling methodology can facilitate and speed up the development of context-aware systems, and can provide a system-level simulation environment for studying context-aware system partially or completely.One of the key issues in context-aware computing is the context inference. Obviously, complete and precise inference is impractical. The bayesian network is an appropriate inference tool in context-aware system. However, searching the optimal bayesian network in many cases is a NP-complete problem. On the other hand, a sub-optimal Bayesian Network can be obtained if the toplogy of the network can be predefined with some prior knowledge on the applications. This dissertation proposes a fast algorithm with low computation complexity. Theoretical and experimental results all show that the algorithm can find a bayesian network similar with the optimal one with affordable search time. Therefore, we say that the algorithm is a practical algorithm for context-aware computing.System infrastructure is a key factor in fast development and wide deployment of context-aware system. Based on multi-agent technology, this dissertation proposes a novel system infrastructure for context-aware system. The infrastructure takes environment agent as the center, thus decoupling the relations of sensors with applications. It uses the sensors and evolution agents to do context sense and evolution. The management agents are used to realize security and privacy strategy, while the user and application agent are used to interactive with user and other applications. In addtion, the context-aware can be used in routing algorithm and system optimization. This dissertation proposed a mobility-aware zoned MANET routing protocol, which can sense the link state context between mobile nodes and then isolate unusual node adaptively which can avoid invalid routing computing and requests and then reduce routing costs.Our contribution lies in the following aspects: 1) through the study on context-aware computing modeling, context evolution, system infrastructure and context-aware applications, this dissertation establishes a global view of context-aware computing; 2) it proposes many key technologies and algorithms for context-aware computing. All of these can provide effective direction and guidelines on researching and deployment of context-aware system.