A Study On Key Techniques Of Aggregate Query In Wireless Sensor Networks

It is currently a hot topic to study the Wireless Sensor Networks (WSN) that is self-organized distributed those randomly exiting sensors. In the emerging area of wireless sensor system, a significant challenge is to develop scalable, energy efficiency, fault-tolerant algorithms to extract useful information from the data the sensors collect. Aggregate query is one of the key operations in query processing for wireless sensor networks (WSN). Aiming at scalability, robust and less transmission cost on high efficient aggregation query, our works present algrithems to several key queries of distinct count query, median query, average query, and aggregation query in mobile environments, which have academic and practical value for advancing the theory and practicability of high efficient aggregation queries in WSN.This dissertation firstly studied the distinct count query in wireless sensor networks. Due to power, network size and robust constraint, centralized algorithms are generally impractical, so many systems used in-network aggregation and multi-path routing methods to reduce network traffic and increased fault–tolerant in these environments. To conserve energy and to avoid double-counting, an approximate algorithm for distinct count query (AADC) based on WSN was proposed. In AADC, each node in WSN created a new Flajolet Martin (FM) subsequence summarizing its own observed values and the received FM subsequences from child nodes, and then broadcasted its subsequence to the parents. Finally, these FM subsequences were combined to a single FM sequence in the root node which its data structure was far smaller than the size of the whole data set. The approximate value for distinct count query could be introduced from the FM sequence quickly. Analytical and experimental results show that the proposed algorithm has the advantages of low power consumption, strong fault-tolerant capability, adjusting error range, and is able to significantly prolong system life.Duplicate-sensitive aggregates can't be computed exactly because of node and communication failures or double-counting. To avid this expense, two approximate algorithms for median query were proposed. Firstly, a sampling algorithm for median query (SAMQ) was proposed, which used sampling theory and in-network aggregation technology to produce approximate results with low communication. Secondly, an approximate algorithm for median query (AAMQ) was proposed. In AAMQ, each node estimated the number of times a value appeared, and created a subsequence samples which get k percent of the most commonly used values form its own observed values, and then broadcasted its subsequence to the parents. Finally, these samples were combined to a single sample across the network in the root node. The approximate value for median query could be introduced quickly from the sample across the network. Analytical and experimental results show that the proposed algorithm can greatly reduce the communication, and has strong fault-tolerant capability.At the present time,almost all the aggregate algorithm need collect data about the global status of the system, which required data fusion and long-distance communication .Therefore these algorithm have disadvantages of bad scalability,high power consumption. A distributed aggregate algorithm for average query (DAAQ) based on WSN was proposed. In DAAQ, the computation process of each node in WSN used the information gathered from just a few nearby neighbors. The algorithm offered a fundamentally distributed solution to analyze data locally without necessarily collected the information of whole nodes to a single central site, and did not require data fusion and long-distance communication. The algorithm could adaptively adjust query range according to query results as well. Experimental results show that the proposed algorithm has the advantages of good scalability, low power consumption, and is able to significantly prolong system life.While there are some advantages to using LEACH distributed cluster formation algorithm, this protocol offers no guarantee about the placement and energy of node. A distributed clustering algorithm using local threshold (DCLT) base on WSN in mobile environments was proposed. In DCLT, cluster-heads are elected based on the residual energy of node and the distance between node and variable centroid of the cluster, which can evenly distribute the energy load among all the nodes. The algorithm offered a fundamentally distributed solution to analyze data locally without necessarily collected the information of whole nodes, and did not require long-distance communication. The results of simulation indicate that the DCLT can provide better load-balancing of cluster heads and less protocol overhead. Comparing with LEACH protocol, DCLT saves energy greatly so that the network lifetime was prolonged.Finally, for better research the characters of wireless sensor networks and analysis the performance of the proposed algorithms in real system; we implemented an indoor temperature monitoring systems for wireless sensor networks.