SCCS: Spatiotemporal clustering and compressing schemes for efficient data collection applications in WSNs

The development of large-scale wireless sensor networks engenders many challenging problems. Examples of such problems include how to dynamically organize the sensor nodes into clusters and how to compress and route the sensing information to a remote base station. Sensed data in sensor systems reflect the spatial and temporal correlations of physical attributes existing intrinsically in the environment. Noteworthy efficient clustering schemes and data compressing techniques proposed recently leverage the spatiotemporal correlation. These include the framework of Liu et al. and schemes introduced by Gedik et al. However, the previous clustering schemes are based on an impractical assumption of a single-hop network architecture and their cluster construction communication cost is relatively expensive. On the other hand, the computational overhead of recent compressing techniques (e.g. the work of Liu et al. and Douglas et al.) is quite significant; therefore, it is hard for sensor nodes with limited processing capability to perform these techniques. With such motivation, we propose a novel and one-round distributed clustering scheme based on spatial correlation between sensor nodes, and propose a novel light-weight compressing algorithm to effectively save the energy at each transmission from sensors to the base station based on temporal correlation of the sensed data. Besides, the aim of the proposed clustering scheme is not only to group the nodes with the highest similarity in observations into the same cluster, but also to construct and maintain a dynamic backbone for efficient data collection in the networks (with the consideration of sink mobility). Computer simulation shows that the proposed schemes significantly reduce the overall number of communications in the cluster construction phase and the energy consumed in each transmission, while maintaining a low variance between the readings of sensor nodes in the same clusters and high reliability of the compressed data.