Data aggregation schemes for Machine-to-Machine gateways: Interplay with MAC protocols

This paper focuses on the use capillary M2M (Machine-to-Machine) networks for the estimation of spatial random fields. The capillary network is coordinated by an M2M gateway which is in charge of (i) collecting the set of spatially correlated sensor observations; and (ii) aggregating (compressing) such observations prior to their re-transmission to a remote application server. We consider a realistic scenario where a contention-based Medium Access Control scheme is used for sensor-to-GW communications. In this context, we attempt to determine the optimal duration of the two transmission phases, namely, sensor-to-gateway and gateway-to-application server, in such a way that the best trade-off in terms of packet collisions and compression level is attained. By doing so, the distortion in the reconstructed random field can be minimized. We also investigate how the optimal operating point is affected by the specific data aggregation scheme, both of them being based on the Karhunen-Loève transform, or the spatial correlation properties of the underlying field.