Research On Distributed Algorithm Against Noise Interference In Wireless Sensor Networks

The wireless sensor network is composed of sensor nodes,which can collect and process information.Distributed estimation can estimate target parameters through cooperative communication between nodes within the network,it is widely used in wireless sensor networks due to its unique characteristics such as robustness,stability,and effectiveness.When sensor nodes collect and process data,the interference of background noise can not be avoided.Therefore,solving noise interference in wireless sensor networks is an important task in distributed estimation.In previous studies,it is found that distributed least mean squares algorithms(DLMS)and distributed sign algorithms(DSA)are robust against Gaussian noise in wireless sensor network.However,besides Gaussian noise,impulse noise also exists widely in signal processing applications,and it also destroys the estimation performance of these algorithms.To solve the problem of output noise interference,we have proposed the distributed least logarithmic absolute difference(LLAD)algorithm over wireless sensor network.The proposed algorithm based on least logarithmic absolute difference cost function is resistant to impulsive interference.By using such a relative cost measure,the proposed algorithm can effectively adjust the traditional cost function on the basis of error optimization.The distributed least logarithmic absolute difference algorithm achieves a better performance than distributed least mean squares algorithms and distributed sign algorithms in impulsive noise environment.The distributed LLAD algorithm has similar performance with the distributed LMS algorithm with Gaussian noise.In the theoretical analysis,we study the stability and convergence of our algorithm.Our simulation results demonstrate that the distributed LLAD algorithm achieves better robustness against impulse interference than distributed LMS and distributed SA.The noise in the system not only exists in the output,but also exists widely in the input.Aiming at the problem of noise interference in input and output,a distributed minimum total error entropy algorithm is proposed in this thesis.The proposed algorithm is also based on diffusion strategy over wireless sensor networks in which the node only interacts with defined communication range neighbors.By using the minimum error entropy function,the optimization strategy is adjusted effectively in the face of large error changes.In the theoretical analysis,we study the stability and convergence of the algorithm.Experimental results show that the algorithm can resist input Gaussian noise and output impulse noise.Summarized the above work,we have studied the problem of noise interference in wireless sensor networks.It is found that different types of noise have great differences on the performance of distributed estimation algorithm.Aiming at the noise interference problem in input and output,we propose a series of distributed estimation algorithms,which can resist both Gaussian and impulse noise.In the next work,we will study more noise types and apply the idea of distributed estimation to more scenarios.