Research On Moving Source Localization Based On Enhanced Semidefinite Programming

Moving source localization is an important research topic in signal processing owing to its wide use in many applications like radar,wireless communications and sensor networks.The main thought of moving source localization is constructing localization equations to determine the location and the velocity of the target based on the parameters which are collecting from the received signals.However,solving localization equations is a nontrivial problem owing to its nonlinearity and the noise exists in the localization parameters.Up to now,Source localization problem has been solved for low noise case,but still lacks solution in high noise or complex electromagnetic environment.In mathematics,source localization is a quadratically constrainted quadratic programming problem,and semidefinite programming is an effective method for such problems.However,using semidefinite programming directly in source localization problem leads to poor accuracy when the noise level is high.In order to solve this problem,this thesis proposed four enhanced semidefinite programming methods for source localization problems in passive location systems and distributed MIMO radar systems based on the recent researches in convex optimization problem.The proposed approaches can obtain high localization accuracy under high noise levels.The main contributions and innovation points of this thesis are listed as follows:(1)A source localization algorithm based on semidefinite programming and reformulation linearization technique is proposed.To overcome the shortcoming of the insufficient accuracy of the semidefinite programming method,this algorithm first utilizes reformulation linearization technique to convert the localization problem to convex optimization problem,then introduces semidefinite relaxation constraints to tighten the feasible region of the problem.This algorithm can get an accurate estimate of the location and the velocity of the source under high noise levels.(2)An enhanced semidefinite programming solution for source localization based on the geometric analysis to the feasible region is proposed.This algorithm introduces bound constraints according to the localization model first,then tightens the semidefinite relaxation by driving new valid constraints.This algorithm has a better performance than other existing methods when the noise level is high.(3)A source localization algorithm under unknown noise power knowledge is proposed.To overcome the shortcoming that most pervious approaches rely on the exact knowledge of the measurement noise power,this algorithm formulates the time difference of arrival and the frequency difference of arrival equations respectively,and by doing so,the variances of the TDOA and FDOA measurement noises can be separated and neglected.Then this algorithm obtain the estimation of the source position and velocity by solving two alone SDP problems.This algorithm can get an accurate estimate of the source position and velocity when the noise power knowledge is absent.(4)A maximum-likelihood based semidefinite programming algorithm for distributed MIMO radar systems is proposed.This algorithm transforms directly the maximumlikelihood localization problem into a semidefinite program which can be solved efficiently using interior-point methods according to the characteristics of the localization model,then introduce an additional procedure to improve the estimate based on the measurement model.This algorithm significant outperforms the existing techniques in distributed MIMO radar systems.