Study On Distributed Algorithm For Non-negative Parameter Estimation In Wireless Sensor Networks

With the advantages of low cost,low power consumption and small size,wireless sensor network(WSN)is widely used in areas such as environmental monitoring,industrial and agricultural control,city management etc.The application of distributed algorithm for parameter estimation in WSN can effectively solve the problems of limited energy,limited communication capacity and limited storage capacity.Therefore,the study on distributed algorithm for parameter estimation is significantly important.In many practical systems,some parameters are subject to nonnegative constraints,such as the parameter in the combined forecasting model and the abundance in the problem of hyperspectral data unmixing.Based on the analysis for previous studies on distributed parameter estimation,this paper studies the problem of distributed estimation for nonnegative parameter in WSN.The main work is summarized in the following:(1)A cost function with nonnegative constraint for the parameter is established.Then the distributed algorithm for nonnegative parameter estimation is deduced by using gradient descent method.(2)The analysis for mean and mean-square stability of estimation error is performed.(3)Through numerical simulations,the proposed algorithm is compared with several other algorithms.On one hand,the theoretical analysis for the proposed algorithm is verified.On the other hand,the higher performance of the proposed algorithm in estimation for nonnegative parameter is shown.(4)The proposed algorithm is applied in intelligent warning system for gas in mines,and satisfying performances are achieved.In a word,this paper proposes a distributed algorithm for nonnegative parameter estimation in wireless network.In the field of distributed algorithm for nonnegative parameter estimation,there is still much room for future study when the topology structure,environment noise and node anomaly are also considered in the wireless network.