Sparse Random Projection Compressive Data Gathering in Lossy Wireless Sensor Networks
Xuan-Gou WU,Zhao-Bin CHU,Xiao ZHENG,Xiu-Jun WANG,Pan-Long YANG
DOI: https://doi.org/10.11897/SP.J.1016.2019.00388
2019-01-01
Abstract:Compressed sensing is considered as a promising data sampling and compression technology compared with the traditional transformed based data compression technology.Existing research results show that compressive sensing based data gathering approaches can reduce network communication costs and balance network energy consumption in wireless sensor networks.Although many existing approaches have been proposed to reduce data transmission cost and balance network energy consumption in wireless sensor networks, these approaches mainly focus on reliable wireless sensor networks.That is, no network packet loss occurs on the network links between all sensor nodes.However, the experimental results of real sensor network system show that the network packet loss is a common phenomenon in wireless sensor networks under network packet loss retransmission mechanism.Unfortunately, our experimental results show that therecovery data performance of these approaches would be disturbed severely by the network packet loss during wireless sensor network data gathering.Even if there is a small amount of network packet loss on wireless sensor networks, the recovery sensory data of the traditional compressive sensing based data gathering approaches will be still arbitrarily bad.To make compressed sensing meet the data gathering requirements of lossy wireless sensor network, we propose a sparse random data gathering approach with packet loss consideration.Our proposed approach can effectively cope with the impact of network packet loss on the recovery quality of sensory data in wireless sensor network data gathering.In our approach, we mainly exploit two aspects of techniques to achieve the robustness of sensory data in lossy wireless sensor networks.First, unlike the existing random sparse projection approaches, we design our sparse projection matrix according to the network routing in wireless sensor networks.Our designed sparse projection matrix can effectively reduce the damage of gathered measurements caused by packet loss during wireless sensor network data gathering.Second, in order to satisfy the conditional requirements of compressed sensing theory for sparse projection matrix and sparse representation basis, we design a low-coherence sparse representation basis with sparse projection matrix constraint.In other words, our designed sparse projection matrix and sparse representation basis guarantee that the sensory data can be reconstructed successfully even if some packets are lost during wireless sensor network data gathering.Meanwhile, we present a sparse representation learning algorithm by the historical sensory data, which satisfies the low-coherence between the sparse representation matrix and the sparse projection routing matrix.Furthermore, we also give the corresponding distributed data gathering algorithms of our approach in wireless sensor network.The experimental results of real network data show that the sparse representation basis obtained by our proposed representation base learning algorithm can compress the network sensory data effectively, the proposed joint sparse projection data gathering approach can recover and reconstruct the sensory data accurately, even when the network packet loss rate is up to 20％in wireless sensor networks.At the same time, the simulation results also show that the proposed scheme can effectively reduce network energy consumption compared with the existing compressed sensing data gathering approaches in wireless sensor network data gathering.