Data-Centric Node Selection for Machine-Type Communications with Lossy Links

While node selection has been popularly studied in the literature for wireless sensor networks, a majority of papers assume a simplistic wireless model without taking communication costs such as radio resource usage and link loss into consideration. In a lossy environment, since data sent back by the selected subset of sensors may suffer from random losses, it may become necessary to use more radio resource usage by either selecting more sensors than needed as backups or providing more transmission opportunities to the selected sensors. In this paper, we investigate how the limited radio resource can be effectively allocated to a selected subset of sensors using machine-type communications for minimizing the data reconstruction error in a data gathering application with lossy links. We first formulate a node selection problem and then investigate two algorithms as solutions. The first algorithm exploits meta-heuristic randomized search in the search space to find a near-optimal solution. The second one, on the other hand, incurs a much lower computation cost by greedily selecting most informative sensors one by one to represent the population. Through computer simulation, we show that providing more transmission opportunities to the selected subset of sensors can achieve a more desirable performance in terms of radio resource usage and energy conservation than selecting more sensors as backups for machine-type communications with lossy links.