Research On Beamforming And Time Synchronization In Wireless Sensor Networks

Wireless sensor network is a new type wireless network. In recent years, with the rapid development of wireless sensor networks, wireless sensor networks has been widely applied in various fields, such as, military, smart city, medical etc. At present, researchers are researching wireless sensor networks on network mode, topological structure, communicating protocol etc. The Key technology includes time synchronization, data management, information fusion and beamforming. This paper discusses the study of beamforming and time synchronization in wireless sensor networks.This paper proposes a beamforming method base on sparse Bayesian learning to solve the problem of beamforming with manageable pattern in wireless sensor networks. According to the characteristics of wireless sensor networks, the method introduces sparse linear array optimization to sparse planar array optimization to complete beamforming in wireless sensor networks reducing the difficulty of beamforming. In this paper, to ensure good beamforming pattern, the wireless sensor networks beamforming pattern is required to approach the reference pattern and we get a sparse linear model about sparse antenna array excitation vector. Then, the sparse Bayesian learning method is used to analysis the prior, to fit the posteriori and determine the maximum a posteriori of the excitation vector. Finally, the method completes beamforming with satisfactory pattern using the maximum a posteriori of the excitation vector. The simulation results verify the feasibility and effective of the method.Due to the uncertain time delay caused by two-way time information exchange mechanism effecting the precision of time synchronization, this paper proposes the RB particle filter (Rao-Blackwellised particle filter) time synchronization method base on DPM (Dirichlet process mixture) model. In this paper, the two-way time information exchange process is equivalent to Markov dynamic process and the uncertain time delay is equivalent to the observation noise. Thus, the clock offset estimation of time synchronization is transformed into the Markov dynamic model state estimation. The method uses DPM model to represent the observation noise distribution. Then it use RB particle filter to estimate DPM model parameters and state variables getting noise distribution estimation described by Gauss mixture model and getting clock offset estimation. Finally, the method eliminates clock offset and realizes time synchronization. Simulation results show the effectiveness of the method. Compared with other time synchronization method, it has better time synchronization precision.