Research On Energy-Efficient And Self-Organizing Approaches For Wireless Sensor Networks

Through integrating computer techniques, sensor techniques, wireless communicationtechniques, and network techniques together, wireless sensor networks have emerged as akey technique to merge the logical information world and the physical world. By offering arevolutionary way for gathering data from the physical world, wireless sensor networks cansolve the problem of the data source of the logical information world. At the same time,wireless sensor networks will greatly advance the interaction between human and the phys-ical world, and thus accelerate the notion of pervasive computing. Wireless sensor networkwill realize the interconnection between the physical world, the information world and thehuman society. As a consequence, wireless sensor networks have been considered as one ofthe most important techniques of the twenty-first century, and have been one hotspot of theinformation science and technology research.Wireless sensor networks are highly energy-constrained networks, and this makes howto effectively make use of the limited energy to maximize the network lifetime the all-important challenge in wireless sensor networks. Furthermore, wireless sensor networksgenerally have no infrastructure and sensor nodes are generally deployed randomly. Thuswireless sensor networks are self-organizing networks which means the sensor nodes ina wireless sensor network should configure and manage themselves to form a multi-hopwireless data gathering and relaying network system. Many challenging research issuesare brought by these two characteristics. Aiming at wireless sensor networks for monitor-ing applications and centering on energy-saving and self-organizing, this dissertation studiesfollowing issues: (1) how to collect the mass data generated by the sensor nodes in an energy-efficient way? (2) how to organize the sensor nodes into an effective structure so that the sinkcan gather data energy-efficiently? (3) how to overcome the packet loss, network conges-tions and energy-consumption hotspots in collecting mass sensor data? (4) how to monitorthe state of wireless sensor networks energy-efficiently to backup the data collecting, self-organizing and network management? (5) how to manage the metadata energy-efficientlyso as to backup the data management and query processing? The research results of thisdissertation are as follows.1. Utilizing the temporal and spatial correlations existing in sensor data, and from theangle of saving energy by reducing the size of data transmitted, this dissertation proposesa model-aided data gathering approach. In this approach, the sensor nodes figure out the temporally changing models and spatial distribution models using the temporal and spatialcorrelations of sensor data, and then the replica of these models are sent to the base station.In such a way, the base station can use the replica models to estimate the actual measurementdata of sensor nodes. Meanwhile, the original models are adopted by sensor nodes to judgehow the estimations of the models agree with the actual measurement data. A measurementdatum will be reported to the base station only when the error of corresponding estimativefigure exceeds the allowable error bound.2. Considering the spatial correlations existing in sensor data, this dissertation proposesan energy-efficient clustering approach. By uniforming the data gathering energy cost in aclustered wireless sensor network as the energy cost of collecting the data generated by asensor node in a data collecting cycle, the clustering problem is modeled as a node-weighteddominating set problem, and corresponding centralized and distributed algorithms are de-signed to organize the sensor nodes into effective network structures that can support thedata gathering and in-network processing effectively.3. To solve the funneling effect problems caused by centralized data gathering, multi-hop communication and many-to-one traffic pattern, based on the idea of supply and demandnetwork, a distributed algorithm is proposed by this dissertation to organize the sensor nodesinto a balanced supply and demand network. The network generated by this approach canbalance the overload of the sensor nodes so as to overcome the problems of packet loss, net-work congestions and energy-consumption hotspots.4. Monitoring the state of wireless sensor network can help the effective network man-agement. The characteristics of large scale and being highly energy-constrained bring greatchallenges to network state monitoring. State monitoring framework and approaches arestudied from the angles of network applications, network types, monitoring demands andthe monitored states, and a prediction-based energy-efficient state monitoring approach isproposed for long-running and continuous state monitoring.5. To assist the data management and query processing, this dissertation analyzes thedemands and challenges imposed on metadata management by the characteristics of wire-less sensor networks. Based on that, this dissertation proposes an energy-efficient metadatamanagement approach, which includes a metadata management framework and metadatacollecting approach.