Scalable and Robust Aggregation Techniques for Extracting Statistical Information in Sensor Networks

Wireless sensor networks have stringent constraints on system resources and data aggregation techniques are critically important. However, accurate data aggregation is difficult due to the variation of sensor readings and due to the frequent communication failures. To address these difficulties, we propose a scalable and robust data aggregation algorithm. The novelty of our work includes two aspects. First, our algorithm exploits the mixture model and the Expectation Maximization (EM) algorithm for parameter estimation. Hence, it captures the effects of aggregation over different scales while keeping the communication cost low. Second, our algorithm exploits loss-tolerant multi-path routing schemes. Hence, it obtains accurate statistical information even in the presence of high link and node failure rates. We demonstrate that our techniques reduce communication cost while retaining the precious statistical information otherwise neglected by other aggregation techniques. Our evaluation shows the proposed techniques are robust against link and node failures, and perform consistently well.