Research On Compressed Network Coding In Wireless Sensor Networks

Many types of wireless sensor networks are playing a significant role in themodern society. The communication scheme of sensing data, of which thereliability and efficiency determine the value of system’s application prospect,affects the performance of the system very directly. There are some defects of thetwo existing communication schemes, especially the all-or-nothing effect ofnetwork decoding, which seriously impacts the application of sensing system.This paper discusses an efficient communication scheme called compressednetwork coding in which we combine network coding(NC) with the concept ofcompressive sensing(CS) to solve the all-or-nothing problem.The structure and data transmitting form are introduced first in this paper.Considering the reliability of the wireless sensor network, we use a distributedcontrol architecture and flooding-based routing technology. Considering thesimilarity between the operation in random linear network coding(RLNC)scheme and the random projection operation in Compressed Sensing, howrandom linear network coding works, such as the data processing at intermediatenodes and the formation of the local encoding vectors and global encodingmatrix, is detailedly analyzed. CS theory is briefly introduced in this paper, andrecovery problem of the compressed signals is discussed by using OrthogonalMatching Pursuit(OMP) method as an example.On the base mentioned above, this paper presents a Compressed NetworkCoding(CNC) scheme to improve the network communication efficiency.Utilizing the correlation of the readings from nodes of sensor networks, weintroduce a method combined compressive sensing and network coding to solvethe all-or-nothing problem of network decoding by designing the packet form andlocal coding vectors. CNC scheme guarantees that the sensing data can beaccurately recovered with a high probability even if the number of receivedpackets is less than that of source nodes in the sensor network. Simulation showsthat we just need much less than the number of packets which conventionalnetwork coding scheme needs to reconstruct measurements with reasonablequality. Compared with the traditional NC scheme, CNC increases the efficiencyof data gathering over40%.