Research On Data Collection For Wireless Sensor Networks

Wireless sensor network is a very important technology for internet of things. Data collection, which is a fundamental function for wireless sensor networks, is the foundation for various smart services of internet of things, thus it is worthy to be studied. Existing data collection methods can effectively reduce the traffic burden of wireless sensor networks. But, they can not simultaneously satisfy the requirement of large-scale long-term persistent date collection on scalability and self-adaption. Existing mobility based data collection methods can effectively balance the traffic burden of wireless sensor networks. But, they ignore the impacts of both the communication capability and the travel path of mobile node, when multiple mobile nodes are used to collect data. As a result, the data collection efficiency of mobile node is reduced. Aiming at the problems mentioned above, this thesis studies data collection in wireless sensor networks from the following four aspects:data spatial-temporal correlation model, distributed data processing and collection, mobile nodes travel paths optimization and multiple mobile nodes cooperated data collection. The main contributions of this thesis can be summarized as follows:(1)We propose a layered data correlation model based on directed graph. According to the spatial and temporal correlation in sensor data, our data correlation model is divided into two layers:the local estimation model and the data approximation model. The local estimation model uses a linear model to model the temporal correlation of sensor data. Based on the local estimation model, the data approximation model uses a directed graph to describe the spatial correlation among the data of sensor nodes. Our data correlation model can estimate the real sensing data of all sensor nodes by the data of a small portion of sensor nodes, and can support distributed data processing.(2)Based on our layered data correlation model, we propose a distributed self-adaptive approximate data collection scheme, called ADC, which can be used in large-scale long-term persistent data collection. ADC reduces the traffic burden of network by only transmitting the data of a small portion of sensor nodes which are called representative nodes. ADC converts the selection of the representative nodes into finding a minimum dominating set of a directed graph and then adjusts the representative nodes according to the change of sensor data to guarantee the accuracy of the collected data and satisfy the self-adaption requirement of long-term data collection. ADC archives distributed data processing and excellent scalability by distributing the correlated computation to each cluster header.(3)We propose a multiple mobile nodes path planning algorithm based on loop-like travel path. Minimizing the number of mobile nodes is NP-hard, when multiple mobile nodes are used to collect data via one-hop wireless communication. By comparing the data collection efficiency of mobile nodes with different travel path pattern, we find that the data collection efficiency of the loop-like travel path is higher than that of non-loop-like travel path. Hence, we propose a heuristic path planning algorithm based on loop-like travel path to optimize the travel paths of mobile nodes.(4)We propose a data collection pattern based on multiple mobile nodes cooperation. In this pattern, data are transferred to the sink node by using multi-hop transmission among mobile nodes. In this way, the communication capability of mobile node required by data collection can be reduced. We formally give the constraints that the travel paths and the movements of mobile nodes must satisfy for the cooperation of mobile nodes, and prove that minimizing the number of mobile nodes under such constrains is NP-hard. Hence, we propose a heuristic path planning algorithm and a movement planning algorithm to stisfy the requirements of the cooperation of mobile nodes and improve the data collection efficiency of mobile nodes.Extensive experiments demonstrate the correctness and effectivity of our data collection methods. Our distributed self-adaptive approximate data collection scheme not only can guarantee the accuracy of the collected data and has excellent self-adaption, but also can effectively reduce the traffic burden by comparing with existing methods. Our path planning algorithm and data collection pattern can improve the data collection efficiency of mobile nodes and reduce the number of mobile nodes simultaneously.