Researching On Topology Identification In Sensor Network Based On Network Tomography

With the wide application of wireless sensor network, network management and maintenance become more and more important. Lots of research is focused on the measurement of sensor network performance parameters. As one of the important performance parameters, sensor network topology plays a special role in the measurement of the other parameters. The traditional popular measuring method of fixed network obtains the network topology by analyzing and processing the information collected from network internal nodes. This method requires the cooperation of network internal nodes. At the same time it not only increases the communication burden but also accelerates network node energy consumption in the process of information collecting. Obviously, this method is not suitable for resource-limited sensor network. It makes sensor network performance parameter measurement face new problems compared with the traditional one.Combining with statistical methods network tomography can infer the network topology by collecting internal information actively or passively without the cooperation of network internal node. Up to now, the research of wireless sensor network measurement based on network tomography is still in the early stage. This dissertation concentrated on studying the sensor network topology by network tomography. The network tomography was discussed in detail from its system modeling, measurement, applications and so on. This dissertation studied the network topological model and the link packet loss-rate model, compared several topology inference algorithms and analyzed their features and drawbacks. Based on the lost/received information received in sink, this dissertation proposes a level-topology identifying algorithm in sensor network based on data aggregation, the level information of the sensor network could be obtained in the same time. When the link loss-rate is small, network topology identifying algorithm proposed right now need lots of observation data in order to infer the topology, this increased the computation quantity and decreased the identifying speed. This dissertation proposed a node sleeping level-topology identifying algorithm, this method increased the virtual link loss-rate by making some nodes turn into sleeping randomly. The simulation results show that the proposed algorithm can infer the logical topology of wireless sensor network with little observation data.