Abstract
Data fusion in sensor networks involves the processing of data from multiple sources. For distributed settings, the fusion operation has to be executed at individual nodes, which makes data synchronization among nodes a crucial task to ensure, consistency. If the sensor nodes are mobile, the communication channel between them can sometimes be broken. When communication is re-established, the snapshots of the nodes may no longer be consistent. The challenge is further made more difficult given limited communication bandwidth.
In this paper, we propose two algorithms for data synchronization in distributed and constrained mobile sensor networks. The first one uses timestamps to locate data records to be transferred between nodes. The second one employs hash digest and index tree to locate the discrepancy set between two sensor nodes. For the hash-based algorithm, we further investigate the cases for complete and rounded data. The evaluation results show that the timestamp-based algorithm is better if complete data are required for fusion, while the hash-based algorithm can outperform the timestamp-based one if the fusion algorithm can tolerate errors within some given threshold.