Study Of Multi-station Short-wave Time-difference-of-arrival Localization Technique Based On Ionosphere-layer Reflected Signal

The focus of this paper is a short-wave time-difference-of-arrival(TDOA)localization system which deals with ionosphere-layer reflected signals,and the target of interest is on the earth surface(on the ground or at the sea).Besides,the system is ground-based,which locates a signal source by passively measuring the TDOAs between its signals which arrive at different base-stations via the ionosphere-layer reflected paths.It brings benefits including wide detection range,novel concealment and lost cost,which make the system potentially complement the active short-wave detection system.For short-wave signal sources,this paper focuses on multi-station TDOA localization,deals with TDOA measurements.The problems to be solved belong to three categories,including basestation deployment,TDOA data association and TDOA localization.The main research contents are summarized as follows:Firstly,the ionosphere-layer reflection transmitting path of signal and the corresponding TDOA data model are described.The traditional base-station deployment methods,TDOA data association methods and TDOA localization methods are also formulated.These form the foundations for the chapters to follow.Secondly,the accurate priori information about ionosphere-layer virtual height(IVH)is rarely known in the practice while the information mainly determines the localization performance of TDOA localization.Thus,this paper will design base-station geometry based on the identity of IVHs to alleviate the impact of IVH errors,where the IVH identity means the IVHs from a target to different sensors are approximately identical under some conditions.Additionally,with the influences of land distribution,electric supply and electronic interference,the available places for base-station deployment are usually limited,which is the so-called geographical constraint in this paper.This motivates us to study base-station geometry design considering the geographical constraint.On this basis,we construct a fractional quadratic integer programming problem based on the approximate cost function of CRLB(Cramer-Rao Lower-Bound)and propose an improved SD(Sphere Decoding)algorithm for the problem,which makes it possible to quickly design sensor geometry.Besides,the ionosphere-layer identity between closely spaced sensors is the basis for base-station design and TDOA localization,for which the identity-test problem of IVHs and an improved solver of this problem will begiven.Theory analyzes and simulation results show that: the proposed improved SD method is faster than EXM method,with the computational complexity acceptable for off-line design process under some conditions;the proposed improved identity-test method is superior to the current state-of-the-art in terms of detection efficiency.Thirdly,when multi-source signals are not separable by frequency and angle,the TDOA measurements for an individual signal are not obtainable and thus the TDOA localization cannot be implemented,which leads to the TDOA data association problem.To take advantage of the additional priori information in multi-source TDOA data association problem,sparse Bayesian recovery will be used.For this reason,a new complex nonlinear operator and subspace-based robust sparse Bayesian inference(SRSBI)will be proposed.Theory analyzes and simulation results show that,the TDOA measurements dealt with by the proposed nonlinear operator are beneficial for recovery accuracy;and,compared with traditional algorithms,SRSBI can provide better TDOA data association results.Finally,the multi-source TDOA localization can be converted into multiple singlesource TDOA localization problem after the right data association procedure,for which the TDOA localization algorithm based on single signal will be studied in this paper.The limitations of traditional parametric algorithms in our background are analyzed;and a robust grid-search(RGS)algorithm based on weight matrix training is proposed for a drawback of traditional grid search algorithms,i.e.,they do not promise to find the nearest grid-point(NGP)of target.Simulation results show that,the matrix training process of RGS is usually feasible by adjusting grid density and the computational cost is acceptable for an off-line progress.Theory analyzes and simulation results also show that,RGS is superior to the current state-of-the-art in terms of localization accuracy and computational complexity.Of the aforementioned research contents and achievements,improved SD algorithm and RGS algorithm can be extended to other signal process applications.Hence,this paper is somewhat meaningful for the development and progression of passive localization technique.