In this paper, we address the problem of traffic aggregation in capillary M2M networks. In particular, we focus on monitoring application where sensor nodes gather information from the environment (e.g. temperature, humidity, concentration of pollutants, etc) and send it to the gateway (GW). The GW then processes and retransmits the data through a cellular network to a distant application server. In an attempt to reduce the traffic congestion in the cellular network, we propose a data compression strategy at the GW that effectively exploits the temporal and spatial correlation in the sensors observations. The data compression strategy is composed of two building blocks: an estimation block and a compression block. In the estimation block, the data streams are jointly processed by means of a multidimensional linear filter. Next, the output streams of the estimation block are compressed by resorting to a truncated version its Karhunen-Loeve expansion.
The different versions of the original document can be found in: