Research On Robust Distributed Estimation Algorithms

A distributed network consists of a number of nodes which are connected according to a certain cooperation strategy and has the ability of information interaction.An adaptive network is a special distributed network whose nodes have the capacity of adaptive signal processing and exchange information between nodes within their neighborhoods so as to estimate parameter or vector of interest.The factors,such as the connection mode of networks,the type of task to deal with,the constrain on the communication conditions,environment noise,and so on,will result in effect on the performance of the distributed estimation algorithms.Impulsive noise is an interference which is often encountered.This type of interference will degrade the performance of the distributed estimation algorithms that are derived based on the l2-norm optimization criteria.Therefore,to study how to restrain the effect the of impulsive noise on distributed estimation is of great significance.This thesis firstly introduces the normalized least mean absolute third cost function into distributed estimation and then proposes a distributed normalized least mean absolute third algorithm,which can reduce the disturbance of impulsive noise on distributed estimation.In order to improve the performance of estimating sparse systems,the proportionate matrix is used in distributed normalized least mean absolute third algorithm and then presents a distributed proportionate normalized least mean absolute third algorithm.Secondly,this thesis introduces the maximized correntropy criterion into multitask adaptive networks and then proposes a multitask distributed algorithm based on maximized correntropy criterion.Considering that in some multitask network applications,the difference between unknown parameter vectors belong to connected clusters,an l2,o-norm similarity-constrained multitask distributed algorithm based on maximized correntropy criterion is also proposed.Lastly,in order to improve the convergence performance of distributed estimation used in multitask networks and to reduce computational complexity,this thesis also presents a maximum Versoria criterion-based multitask distributed algorithm,and uses a variable parameter method to further improve the convergence performance of the algorithm.Simulation results show that the proposed algorithms are robust against impulsive noise.