Distributed networks consist of a set of agents which have the ability of self-processing and studying,whose agents are connected in a certain manner and they can solve distributed estimation problems by local incorporation.The objective of distributed estimation over networks is to enable the agents to estimate a vector of parameters of interest from observed date.Various distributed algorithms have different convergence rate,steady-state misalignment and robustness.To enhance the convergence performance and robustness in an impulsive interference environment,this thesis proposes two kinds of diffusion sign algorithms,i.e.,the diffusion sign subband algorithm with enlarged cooperation and robust diffusion affine projection sign algorithm.The diffusion sign subband algorithm with enlarged cooperation exchanges not only the weight information but also measurements within individual neighborhoods,and the robust diffusion affine projection sign algorithm minimize the error vector function based on the combined L1-L2 norm,so that they can obtain better convergence performance.Moreover,to improve the convergence performance in estimating sparse parameters,this thesis proposes a proportionate diffusion sign subband algorithm with enlarged cooperation and diffusion proportionate affine projection algorithm.Adapting the unknown parameters proportionately can accelerate the convergence rate or reduce the steady-state misalignment.Besides,this thesis also proposes a variable step-size diffusion affine projection algorithm to obtain fast convergence rate and low steady-state misalignment in Gaussian noise environments. |