Correlation Based Data Aggregation Techniques In Wireless Sensor Networks

Wireless sensor network is limited by the node’s hardware, especially the energysource is restricted and irreplaceable, and this makes the energy problem still the biggestobstacle of WSN’s development. Then data fusion technology emerges from simplestpackets merger originally to different technologies including routing protocol, MACprotocol and data prediction, etc, while the research purpose becomes to extend thenetwork’s lifetime and reduce the data communication latency and enhance the dataquality. In WSN, the data collected by nodes have several different correlations:temporal correlation, spatial correlation and attribute correlation, the last one emerges inrecent years. The research of data correlation can promote data aggregation technologiesin WSN, now the research of data correlation mainly focus on excavation andestimation of data correlation, data prediction and data aggregation. This paper can bedivided into three parts as follow:1. Data aggregation technology based on data correlation. Data correlation is themost important feature of WSN. These data correlation can be used for the developmentand enhancement of the key technologies of WSN. In chapter two of this paper, thetemporal and spatial and attribute correlation are analyzed, including the cause of threedifferent data correlation and compare the data aggregation technology they pushforward (data prediction technology derived by temporal correlation, cluster based datacollection technology derived by spatial correlation, attribute correlation derived byWSN’s multi-attributes and its excavation and estimation).2. Periodic data prediction is WSN. Data prediction has been emerged as animportant way to reduce the number of transmissions in wireless sensor networks(WSN). This paper proposes a Periodic Data Prediction Algorithm called P-DPA inWSN. The P-DPA takes the potential law hidden in periodicity as a reference to adjustthe data prediction, which helps to improve the accuracy of prediction algorithm. Theexperiments of temperature, humidity and light intensity based on the dataset whichcomes from the actual data collected from54sensors deployed in the Intel BerkeleyResearch lab proved that the P-DPA has an obvious enhancement to the existing dataprediction algorithms.3. Spatial correlation based data collection protocol. In the fourth chapter, this paper proposes a new way of building cluster through the analysis of spatial correlation.The known cluster structure of WSN data collection protocol is divided into clustersstructure and tree structure, ordinarily the cluster structure is commonly used on nodelevel of the data collection while use the tree structure on cluster and sink level. In thispaper the nodes choose which cluster to join according to the similarity between datacollected by the nodes and the cluster heads, this makes the nodes in one cluster havehigh similarity, which is helpful in improving the quality of collected data whencombined with data prediction algorithm. After that, the periodic data predictionalgorithm in chapter three is applied to wireless sensor nodes which equipped multisensors on one node, here exist a problem that different sensors have differentprediction results while they have only one radio, then a solution with tradeoff betweendata accuracy and energy consumption is proposed to solve this problem. After thisproblem solved, this paper combines the data prediction algorithm into data collectionprotocol which further improved the energy saving effect of data collection protocol.This paper starting from the research of data correlation, then extended up to theresearch of data aggregation technologies, then elaborate and analysis the researchstatus and technical characteristics, and proposed a periodic data prediction algorithmand applies it to the multi sensors node condition. At last, this paper combines the dataprediction into data collection protocol to further prolong the WSN’s lifetime.