Energy-Aware Set-Covering Approaches for Approximate Data Collection in Wireless Sensor Networks

Energy-Efficient Node Selection Algorithms with Correlation Optimization in Wireless Sensor Networks

“…Note that this paper focuses on the active node selection problem by exploiting correlations among nodes but has no concern with the aggregation operators or probabilistic models. We compare the proposed CSB and HREF algorithms with the DClocal, DCglobal [22], EEDC [24], and Snapshot [28] by running them in the same networks as well as the same parameters for the environment.…”

“…An intuitive approach for the node selection process is to use the partially ordered tuple (data coverage range, residual energy) [23]. Another approach is to use partially ordered tuple (residual energy, data coverage range) to select active nodes with higher residual energy [22]. However, the number of selected nodes is generally larger than the former approach, which means that more energy consumption is necessary when providing perception service during the given epoch.…”

“…The active node selection problem with respect to the concept of data coverage range is essentially a set covering problems [27]. We regard the problem of selecting a smallest size of active node set as the problem of selecting the minimum size of subset in set-covering issue [22]. Similar to the greedy approximation algorithm of set covering problem, CSB also takes the greedy strategy to maximize the size of data coverage range for each new added active node.…”

“…Accordingly, a subset of active nodes can be selected to provide the required sensing service within error threshold, and the rest nodes can go to sleep and preserve energy. In this way, the active node selection strategy with correlation optimization not only prolongs the network lifetime, but also helps to solve other issues in dense wireless sensor networks [22], such as lower network throughput, serious node conflict, and excessive packet transmissions.…”

“…The distance function is generally considered as an important model to formulate the data similarity between nodes because the sensitivity is sometime related to the distance between the source and sensing device. Here we adopt Manhattan distance between sensing data as error metric [22]. Based on the observation that the sensing data are similar to each other if they are close enough, Kotidis [23] proposes Snapshot query in which only selected active nodes report their sensing data, and sensing data of one-hop nodes is computed by active nodes.…”