Research On Key Technologies Of Context-Aware Based On WSN In Pervasive Computing Environments

With continuous developments and progress of the sensor technology, wireless communication technology and computing technology, intelligent human-computer interaction has become the dominant development trend of computing model. As the initial attempt to get rid of the shackles of computing devices towards human activities, the"Human-Oriented"concept was gradually approbated by more and more people, therefore lots of pervasive computing applications have come out. Their main goal is to widely deploy all kinds of pervasive devices which have certain computing power and small size in various everyday life environments, and combine them with the existing Internet technology, in order to achieve mobile, seamless, transparent and ubiquitous computing support and services.Owing to pervasive computing systems usually running in extremely dynamic and heterogeneous computing environments, the intelligent entities need to be able to sense changes in the environment , and adaptively adjust the service status based on those dynamic scenes. Wireless sensor networks (WSN), as a practical extension of pervasive computing technology, join a large number of small computing devices (such as sensors) to meet the requirements of the people to explore, utilize and management of the physical world. With WSN, people can obtain fine-detail information using a more microscopic point of view of the surrounding environment or the individual object phenomenon. According to these obtained information about phenomena, sates and behaviors, people can achieve the reverse control or monitoring. Whereupon, the logical information world and the objective physical world can be closely fused together, to achieve the"pervasive computing"concept.In pervasive computing environments, collecting, transmitting and processing a variety of contexts is the key step to achieve pervasive computing support and services. Thus, based on the physical characteristics and application characteristics of WSN in pervasive computing environments, we focus on a number of key technical issues of context-awareness to carry out our in-depth study. Aiming at those key technical issues, such as the edge sensor node detection of the events region in large-scale WSN monitoring areas, the unified context information representing, the high-level context reasoning and predicting, and the secure context fusion and transmitting, we deal with the solutions in depth in the special pervasive computing environments. Finally, we integrate our research results into the UbiCAM middleware to build a context-aware computing prototype system for the elderly healthcare. This can deal with the inter-operation peoblems caused by the heterogeneous nature of hardware and software system in the pervasive environments and improve the flexibility of context-aware applications.The main works and contributions of this dissertation are as follows:(1) In order to effectively monitor those emergencies, an important issue in WSN is the edge detection of the event area. To optimize the system performace, most of existing methods need information of their non-direct neighbors and lack of adaptivity in the process of edge detection. WSN is essentially distributed networks, so we modelled it as a Multi-Agent System (MAS) to study the energy-efficient edge node detection algorithm. Each autonomous agent inhabited in a sensor node independently tests and values the consistency criterion of those measurements which obtained by sensor nodes within the two-dimensional monitoring region grid. According to the feedback from interaction with the environments, as the response to input stimulus of the local environments, autonomous agents correspondingly select and execute appropriate behavioral responses, and ultimately, can adaptively, effectively locate and label all the different homogeneous region. Comparing with the congeneric edge node detection (PR-Classifier algorithm and T-Fit algorithm), experimental results show that the edge node detection algorithm based on autonomous agent have better performance on the average distance of the boundary nodes and the total energy consumption.(2) Timely and accurate perception of a variety of context information in pervasive environments is the basis for intelligent interaction beween the users and the environments. In the context information collection applications which utilizing WSN, due to the influence of the sensor physical characteristics and the environment noise, the obtained context information has certain degree of ambiguity and uncertainty. Those traditional time series predicting methods usually can not effectively deal with such unavoidable problems. Using fuzzy prediction theory, a context prediction method was proposed based on fuzzy time series analysis, and was applied to predict the physiological signs of users in the context-aware middleware UbiCAM. Experimental results show that the method can obtain accurate predictions of the pulse, the blood oxygen (arterial) and the body temperature, and can meet the needs of practical applications. (3) Currently, there are few studies on the establishment of comprehensive evaluation system based on the interaction among sensor nodes in WSN, and thus to obtain secure data aggregation by quantifying the degree of trustiness. Based on the thought of establish node reputation according to the relevant evaluation from other nodes in P2P network studies, we study the secure data aggregation using Set Pair Analysis (SPA) theory (an uncertainty system theory). During the phase of data collecting and transmitting, taking the SPA results of node reputation into account, a new secure data aggregation method based on sensor node reputation SPA was proposed. In the data collecting phase, according to the actual needs of different applications, the predefined reputation threshold was presented to filter the raw data. In the data transmitting phase, through the evaluation of energy, distance, prefernce and other attribute information, the sensor node with best comprehensive evaluation was selected to forward the raw data. Experimental results show that the method has satisfying performace improvements comparing with the congeneric BTSR algorithm and LEACH algorithm on the energy consumption of cluster head node, the probability of safety, the accuracy of aggregation and other performance indicators.(4) Using the unique advantages of sensor networks on real-time, multi-attribute information acquisition and transmission, based on WSN, we have designed and implemented a context-aware middleware UbiCAM prototype system for elderly medical healthcare application in pervasive computing environments. It effective separates the upper context-aware application development from the underlying context acquisition. On the hardware part, we detailedly described the design and implementation of the WSN platform Ubicell and the gateway nodes. On the software part, we proposed the hybrid conext modelling method, evidence-based uncertainty reasoning, and the PCA-based in-cluster context aggregation method. Through the design and development of the UbiCAM middleware prototype system, we integrated our related research results into the prototype system, and demonstrated the innovation and practicability.