In convergecast or data collection applications of wireless sensor networks, the sensor nodes generate a large amount of redundant data. To save energy and communication costs when collecting meaningful data from convergecast traffic, efficient data aggregation is required. A network clustering method has been proposed in the literature for efficient data aggregation. However, further improvement is required in order to make such a clustering algorithm generally effective in any environment or situation. With this observation, this work proposes a technique that adaptively chooses from hybrid “clustering-based data aggregation techniques [based] on the network condition.” It is meant for data aggregation in target-tracking applications.

The proposed hybrid mechanism takes advantage of static (lower energy consumption and lower delay) and dynamic (higher aggregation rate) clustering. Two types of hybrid algorithms are proposed to improve the performance of data aggregation. The first algorithm is combined cluster-based data aggregation, where multiple aggregations are used together in different segments of the same network. The choice of the aggregation method by a node is based on the current location of the sensor node with respect to the sink node. The second algorithm is adaptive cluster-based data aggregation, where the sensor node switches its aggregation method based on the network’s current status. The status of the network is a function of the attributes of moving targets: number of targets, velocity, and so on. The combined aggregation algorithm is claimed to be more stable in terms of network performance. It tries to solve the problems in static and dynamic clustering. However, it can’t effectively adapt to scenarios of multiple targets and high-velocity mobile targets. The adaptive algorithm provides a solution for these scenarios.

The authors propose two methods for dynamic and adaptive data aggregation for data collection in target-tracking applications. The related works section contains citations about dynamic and static clustering works. Theoretical analyses (on a grid network) are done for sensor node distribution. The performance is evaluated on a network of 100 sensor nodes on a Qual-Net simulator. The data aggregation performance is evaluated in relation to target speed and sensing range.

Overall, the idea of adaptive data aggregation and using advantages of both static and dynamic clustering is interesting, although the proposed schemes are heuristics, with no theoretical analysis applied to general networks. The protocol involves the flooding of the network with control messages, so it can include delay and congestion. Therefore, it will be key to evaluate the performance of this protocol in real networks. Readers interested in static, dynamic, and hybrid clustering for data aggregation in sensor networks should read this paper.