A new data aggregation scheme via adaptive compression for wireless sensor networks

Abstract

Data aggregation is necessary for extending the network lifetime of wireless sensor nodes with limited processing and power capabilities, since energy expended in transmitting a single data bit would be at least several orders of magnitude higher when compared to that needed for a 32-bit computation. Therefore, in this article, a novel nonlinear adaptive pulse coded modulation-based compression (NADPCMC) scheme is proposed for data aggregation in a wireless sensor network (WSN). The NADPCMC comprises of two estimators—one at the source or transmitter and the second one at the destination node. The estimator at the source node approximates the data value for each sample. The difference between the data sample and its estimate is quantized and transmitted to the next hop node instead of the actual data sample, thus reducing the amount of data transmission and rending energy savings. A similar estimator at the next hop node or base station reconstructs the original data. It is demonstrated that repeated application of the NADPCMC scheme along the route in a WSN results in data aggregation. Satisfactory performance of the proposed scheme in terms of distortion, compression ratio, and energy efficiency and in the presence of estimation and quantization errors for data aggregation is demonstrated using the Lyapunov approach. Then the performance of the proposed scheme is contrasted with the available compression schemes in an NS-2 environment through several benchmarking datasets. Simulation and hardware results demonstrate that almost 50% energy savings with low distortion levels below 5% and low overhead are observed when compared to no compression. Iteratively applying the proposed compression scheme at the cluster head nodes along the routes over the network yields an additional improvement of 20% in energy savings per aggregation with an overall distortion below 8%.