Research On Sampling-based Data Aggregation Algorithms In Wireless Sensor Networks

Wireless Sensor Network is made up of large number of sensor nodes. Thenodes can communicate with each other wirelessly, and they can spontaneouslyform a network. Besides, the sensor node can sense and collect the information ofthe environment around, it can also do some computation. Nowadays, wirelesssensor networks can be used for environment monitoring、wild animals monitoring、traffic control、military surveillance and so on. The Wireless Sensor Network alsoplays an important role in the Cyber Physical System and the Internet of Things.Sensor nodes are supported by micro battery, so the energy constraint is thebottle-neck that restrict the ability of the wireless sensor network. Data aggregationin wireless sensor networks can show the overview of the monitored environment,and it can also help save the energy consumption. The weighted data aggregationcan describe the environment more objectively, by assigning different weights todifferent nodes. Early research on data aggregation in WSNs always return users theaccurate result. In fact, it is often impossible to get the true accurate results, becauseof the instability of sensor nodes and the networks. Recently, a lot of research arefocused on the approximate data aggregation, and some algorithms can fulfillarbitrary accuracy requirement given by the users. However, no previous works payattention on weighted data aggregation. At the same time, no work concentrates onthe path selection problem during the computation of in-network approximateaggregation, which can help save the energy used for communication.Based on the analysis above, this paper proposed a new algorithm forapproximate weighted aggregation, this algorithm can ensure arbitrary accuracyrequirement. Based on the uniform sampling method, this paper put forwards thegroup-based sampling method, and uses it for the algorithm. Theoretical analysisproves our method can reduce the number of samples. Then this paper shows how todecide the number of samples and gives the exhaustive working flow of thealgorithm. Experiments prove that our algorithm works well. This paper also putsforwards routing strategy for in-network sampling processing, we outline thedifference between this one and previous works. For the snapshot query, our algorithm can ensure the nodes make locally best choice for routing, and prove itscorrectness by theoretical analysis. For the continuous query, a heuristic algorithmis proposed. This algorithm can significantly reduce the number of inner nodes,especially in dense distributed WSNs. As experiment results show that ouralgorithms can save lots of energy when compared with previous works.