Autoregressive model based data gathering algorithm for wireless sensor networks with compressive sensing

The recently emerged compressive sensing (CS) theory provides a whole new avenue for data gathering (DG) in wireless sensor networks (WSNs) with the benefits of energy efficient. The exiting CS based DG approaches design the routing processes based on the assumption that there is a transform that can sparsely represent the sensor readings. However, most of the real sensor readings are well approximated by the sparse signals, which increases the amount of the data traffic to obtain the exact reconstruction, specially in the noisy situation. In this paper, we present an adaptable CS based DG scheme in WSNs. The autoregressive (AR) model based optimization problem is considered for the successive reconstruction and the reduction of the data traffic, in which AR model and the routing processes are jointly introduced into the reconstruction of the sensor readings. The AR parameters are evaluated with the historical sensor readings at the sink, by taking advantage of the similar spatial correlation. Then, with the known AR parameters, the sensor readings are reconstructed by extending the BPDN algorithm to the AR model based optimization problem. Our experimental results indicate that the proposed DG scheme outperforms the exist random walk (RW) based DG scheme in terms of reconstruction error and energy efficient, which results in better reconstruction accuracy with the RWs with the significantly smaller length.