Research On The Key Technologies Of Environmental Change Awareness

With the rapid development of ubiquitous networks and the Internet, context-aware technology has been widely used in the Internet of Things (IoT), composite service self-adaptive systems and various areas of mobile communications, becoming one of important supporting technologies of self-adaptive runtime system and a variety of auto run systems. In recent years, many researchers at home and abroad have designed a number of typical context-aware system frameworks, and conducted in-depth research on related technologies such as identification of data sources, data acquisition and data preprocessing and context-aware information inference, and achieved substantial results. Among them, context change awareness techniques characterized by awareness of changes in context have been the topic of research because they directly trigger and impact the decision-making behaviors of application systems and self-adaptive actions of self-adaptive systems.Currently, research on context change awareness mainly deals with three types of changes:changes in applications in the application layer, changes in monitoring tools in the context monitoring layer, and changes in state of contexts being monitored in the context layer. These three types of changes can all affect the runtime quality of systems. Specifically, in the application layer, addition and reduction of applications will disrupt the original equilibrium of systems, resulting in problems like decreased performance of related supporting subsystems. Changes in monitoring tools in the context monitoring layer will affect the quality of monitoring data, leading to problems such as inaccuracy, deficiency or redundancy of monitoring data, and thereby increase the burden of context change awareness. Changes in state of contexts being monitored will make the original system no longer suitable for current context, and will thus require the system to make corresponding decisions according to context changes. However, existing researches have focused on the awareness of context state, while lacking attention to context change awareness. To this end, this thesis carries out research on key techniques related to context change awareness. Targeting the issue of data source selection caused by changes in application layers in IoT, a sensor selection method for IoT applications is studied. Targeting the issue of data missing caused by changes in wireless sensors in the context monitoring layer, a collaborative filtering-based data complementation method is studied. Targeting the issue of lowered QoS of self-adaptive composite service systems caused by changes in physical context state in context layer, context change event identification methods for self-adaptive composite services are studied. More specifically:(1) Targeting the issue of selection of data sources during deployment of new applications on existing wireless sensor networks in IoT context, an IoT application-oriented sensor selection method is proposed. By analyzing the status of sensor multiplexing and the state of each sensor in sensor networks, and utilizing coverage-oriented optimal selection algorithm, the method selects a subset of sensor nodes which can maximize the using time of entire sensor network, deploys redundant sensors for potentially problematic sensor nodes in this subset, and thereby solves the problem of data source selection during deployment of new applications. Experimental results show that the sensor subset selected using this method improves the using efficiency of sensors, and prolongs the overall using time and maintenance interval of sensor networks while allowing the wireless sensor networks to meet IoT application needs.(2) Targeting the issue of missing data complementation after missing of monitoring data caused by unreliability of ireless sensors and self-organizing property of sensor networks in context change awareness systems with wireless sensor networks as data sources, a collaborative filtering-based sensor data completion method is proposed. By classifying sensors with missing data, and exploiting spatio-temporal correlation among sensor data in sensor networks, this method selects similar sensors using different similarity assessment approaches to find similar sensor data to complement the missing data, so as to ensure the accuracy of estimated values, and thereby improve the accuracy and reliability of context change information sensing capability of context change awareness systems. Experimental results indicate that the method has a higher value estimation accuracy than other data complementation methods when monitoring data of sensors are changing drastically.(3) Targeting the issue of missing data complementation after missing of monitoring data caused by unreliability of ireless sensors and self-organizing property of sensor networks in context change awareness systems with wireless sensor networks as data sources, a collaborative filtering-based sensor data completion method is proposed. By classifying sensors with missing data, and exploiting spatio-temporal correlation among sensor data in sensor networks, this method selects similar sensors using different similarity assessment approaches to find similar sensor data to complement the missing data, so as to ensure the accuracy of estimated values, and thereby improve the accuracy and reliability of context change information sensing capability of context change awareness systems. Experimental results indicate that the method has a higher value estimation accuracy than other data complementation methods when monitoring data of sensors are changing drastically.(4) Targeting the issue of the impact of changes in network contexts on runtime QoS of composite services, the impacts of context changes likely to occur in network context on application service systems are analyzed systematically. An approach which computes changes in QoS of affected services by identification of network context changes is proposed. On this basis, a context change identification method for dynamic composite service QoS computation is given, a context aware function from context monitoring to change event output is implemented, context monitoring and context change identification strategies are presented, as well as context change identification algorithm and method for computation of change identification threshold, in order to provide strong support for improving runtime QoS of Web services, and context self-adaptability of composite services. Experimental results demonstrate that this method can effectively identify changes in network contexts in the network context the composite services are in and compute the degree of impact of network context changes on QoS of services, and thus provides reliable information on context and service changes for self-adaptive decision-making of composite services.