High Energy Efficiency Data Aggregation Methods For Wireless Sensor Networks Based On Compressive Network Coding

As the top ten emerging technologies of the21st century, Wireless sensor networks (WSN) connect the physical world and information networks through sensors, microelectronics, and Wireless network integration, greatly expanding human’s capability of understanding and changing the world. However, on the one hand, sensors without energy harvesting functions are constrained to limited battery capacity. On the other hand, nodes’ energies in a WSN are mostly in imbalance usage states. These two facts severely restrict the network life and network expansion. Although some studies have shown that sensor nodes can capture energy through energy harvesting technologies, the conversion rate and the applied scenarios hinder the wide deployment of such nodes. In addition, as some critical nodes are in imbalance power expenditure, this becomes a main bottleneck issue affecting the network life and network expansion. Hence, it is important to improve the network-level energy efficiency for ensuring the large-scale, reliable network deployments and applications. The key issue on designing such a high energy-efficiency WSN lies in the analysis on network characteristics. Since the collected data by sensors is highly correlated in both spatial and temporal domain, it is feasible to compress and forward the data by leveraging the correlations. This can effectively enhance the network energy efficiency. As two independent new techniques in recent years, network coding and compressed sensing are changing the methodology about data transmission mechanisms, and provide us a new research horizon. However, these two technologies have their limitations in addressing the energy constraint and imbalance energy expenditure of WSNs. The theoretical and implementation combination of network coding and compression sensing still needs to be studied in dept.In this thesis, through in-depth analysis on network coding and compressed sensing, as well as their intrinsic connections, a compressed network coding (CNC) enabled WSN data fusion scheme is proposed. The key issue in the presented scheme comes from the data fusion characteristics and data correlations among sensor nodes. Based on the establishment of data fusion model for WSNs, an efficient distributed data transmission method is discussed. The network energy-efficiency is effectively improved by reducing the total energy consumption. In addition, by analyzing nodal energy consumption and enabling central scheduling in the network, an energy-balanced data fusion scheme is given. This can effectively solve the nodal failure resulted by imbalance energy expenditure, so as to improve the network life via reducing the total energy consumption. Finally, the effectiveness of the proposed scheme is evaluated through a testing platform.The main contributions of this thesis are as follows.(1) A WSN data fusion model based on CNC is proposed. Through extensive literature review on state-of-the-art in WSN data fusion models, their general advantages and disadvantages are discussed. By analyzing the coding gain in generalized butterfly based network coding, and the sparsity in collected data, the feasibility of CNC enabled data fusion model is discussed. Then, the detailed CNC enabled data fusion framework, including network topology, nodal functions, and nodal operation modes, is proposed.(2) An efficient data transmission scheme for CNC enabled data fusion is presented. Based on the CNC enabled WSN data fusion model, an Orthogonal Matching Pursuit Random Linear Compressed Network Coding (OMP-RL-CNC) based data fusion scheme is given. In order to meet the Restricted Isometry Property (RIP) in network coding, network coding vectors for WSN are designed. In addition, by combining network coding and compressed sensing together, a CNC enabled efficient data transmission scheme for WSN is further studied, which can address the all “or nothing” problem in Finite Field based Random Linear Network Coding (FFRL-NC) and the low efficiency in reconstruction. This can significantly reduce the total network energy consumption by enhancing data transmission efficiency.(3) Energy balancing schemes for CNC enabled data fusion is studied. Based on the aforementioned data fusion model, an Enhanced Orthogonal Matching Pursuit Random Linear Compressed Network Coding (E-OMP-RL-CNC) enabled transmission scheme and an tow-times Enhanced Orthogonal Matching Pursuit Random Linear Compressed Network Coding (E2-OMP-RL-CNC) transmission scheme for maximizing the network life are respectively given. Furthermore, energy consumptions in Collection Tree Protocol (CTP), network coding based data transmission, E-OMP-RL-CNC and E2-OMP-RL-CNC are analyzed. It is verified by the simulations that, the proposed E-OMP-RL-CNC and E2-OMP-RL-CNC based transmission schemes outperform others, showing that the proposed schemes consumes less energy. They thus can prolong the network life and make the network expansion possible.(4) Theoretical feasibility is verified through a testing platform. We built an outdoor WSN environment, and implemented the proposed CNC enabled data fusion in such a scenario. By transplanting hardware, difference data fusion mechanisms and the associated energy consumptions are tested. It is shown that the proposed CNC enabled data fusion is effective for WSN in terms of high energy efficiency and balanced energy consumption.