One-bit Recursive Least-Square Algorithm Over Sensor Networks

With the continuous advancement of modern wireless communication technology,embedded technology and sensor technology,the theory and technology of wireless sensor networks have become one of the research hotspots at home and abroad,and it has been widely used in the fields of environmental monitoring,military defense,industry,agriculture,medical treatment and so on.Parameter estimation is one of the key problems in wireless sensor networks.It establishes a certain statistical model through the measured values collected by the nodes,and uses the corresponding algorithm to obtain the estimated values of the parameters of interest.For the problem of parameter estimation over sensor networks,how to reduce the cost of using the network as much as possible,and at the same time get better parameter estimation performance has attracted more and more attention from scholars at home and abroad in recent years.In order to reduce the energy consumption,save bandwidth and storage resources of the network,this thesis considers to use adaptive parameter estimation algorithm to estimate the unknown parameters with one-bit measurement values of each node in the sensor network.First,each node compresses the measurement information into one-bit data.Then,according to whether there is a fusion center or not in the topology of the network,this thesis studies the parameter estimation problem into centralized and distributed.In the centralized parameter estimation,each node transmits the perceived data to the fusion center with powerful computing function for centralized processing.In the distributed parameter estimation,each node transfers the data to the neighbor node according to a certain cooperation mode to obtain the local optimal estimation value.Since the measured value after quantization is nonlinear with the unknown parameter,the classical adaptive algorithm which directly uses the minimum mean square error criterion cannot produce satisfactory estimation results.This thesis combines the expectation maximization algorithm and the recursive least squares method to propose one-bit centralized RLS algorithm and one-bit distributed RLS algorithm for the two different working modes of the network,and further extends the two algorithms to the case where the noise variance of each sensor node in the network is unknown.The thesis compares the proposed algorithms with other corresponding algorithms and conducts a series of simulation experiments.The results show that the proposed algorithms have better convergence and stability,and can obtain similar estimation accuracy with the classical RLS algorithm using non quantitative measurements.At the same time,experimental results show that the centralized parameter estimation accuracy is higher than that of distributed,but in this way,once the fusion center fails,the entire network will stop working and the robustness is poor.