| Summary: | 碩士 === 國立交通大學 === 資訊科學與工程研究所 === 94 === In this paper, we focus on the issue of approximate query processing in wireless sensor networks. Since there are usually spatial correlations between the readings of sensors in vicinity, it is energy-efficient to group these sensors into clusters and select one r-node, i.e. cluster-head, for each cluster to answer queries. We propose an innovative concept called data-coverage to address the problem. We prove that the data-covering problem which selects minimal number of r-nodes to fully data-cover the whole network is an NP-complete problem by reducing the set-covering problem to our data-covering problem. In order to solve the data-covering problem, we devise two heuristic algorithms DCglobal and DClocal. The first algorithm, DCglobal, is a centralized algorithm executed at the sink. The ratio between the number of r-nodes selected by DCglobal and that of the optimal solution is bounded. Since the energy consumption
of DCglobal is extremely high, we devise another algorithm called DClocal. DClocal is a distributed algorithm executed by each sensor to locally select r-nodes. We prove that the ratio between the number of r-nodes selected by DClocal and that of the optimal solution is bounded. We then discuss the optimal value of the parameter used in DClocal. Through the experimental study, it can be seen that the energy consumption of DClocal is less than that of DCglobal. In addition, through comparing with other algorithms, the performance of DClocal is much better in terms of network lifetime and energy consumption. Thus, we conclude that our algorithms are energy-efficient and can provide the users the answers that satisfying their requirements for precision.
|
|---|
