The Research On Distributed LMS Algorithm In Wireless Sensor Network

With the increasingly development of low-power wireless communication technology and the miniaturization of micro-sensors and micro-processors, Wireless Sensor Networks (WSN) has become an extremely hot spot of present researches. However, in the large scale of WSN, each node should transmit its own sensing data, which will lead to a lot of data redundancy inevitably, and the power consumed by data transmission is far greater than the power used by data processing. Therefore, it is very necessary to perform data aggregation before data transmission. Relying on the efficient algorithms, data aggregation technology can not only reduce the power consumption in network and lengthen the life circle of sensor node, but also improve data collection efficiency to obtain more accurate information. In this thesis, the key research point lies in the distributed data aggregation technology based on LMS algorithm.Firstly, the relative knowledge of WSN is introduced in this paper, including its problems and challenges, the application of the data aggregation technology as well as the development status at home and abroad. Then the derivation process of the adaptive LMS algorithm is given and the evaluating parameters of algorithm performance are discussed, those laid the theoretical foundation for the subsequent chapters.Secondly, the collaborative model of the WSN is introduced, and then two kinds of data aggregation algorithms-centralized and distributed are analyzed. Based on distributed algorithm,aiming at the limited data rate in WSN, an improved distributed diffusion sign LMS algorithm are proposed.It can effectively reduce the computational complexity and communication among nodes when applying to quantify the error signal and the estimated signal.Simulation results prove the correctness of the proposed algorithm,and show good performance by comparing with other algorithms.Especially for the BPSK signal, the proposed scheme outperforms the traditional diffusion LMS algorithm.Finally, to deal with the problem for WSN in a non-secure environment which malicious nodes tamper signal data, the based on reputation mechanism distributed diffusion LMS (R-dLMS) algorithm and diffusion normalized LMS (R-dNLMS) algorithm is proposed. The propose algorithms can allocate the appropriate reputation value according to the contribution of each node in the network. The reputation value of malicious node is minimized and non-malicious nodes are opposite, so that the impact of malicious nodes on the network is minimized. Simulation results verify the effectiveness and superiority of the proposed algorithm, and show that the proposed can greatly improve the performance compared to the algorithm without reputation value, as the same time, the performance of R-dNLMS algorithm has been further improved based on R-dLMS algorithm.