Research On Localization Algorithms Based On TDOA And AOA

Target localization has become an indispensable technology in modern scientific and technological fields.How to achieve positioning conveniently and abtain high-accuracy location is the key issue in wireless localization.Influenced by noise,non-line of sight propagation,multipath effect and other unfavorable factors in wireless channel,the signal measurements are always noisy,which leads to inaccurate localization.Thus,this thesis mainly studies how to reduce the influence of these unfavorable factors and obtain high-precision localization algorithms.In this thesis,we focus on time difference of arrival(TDOA)and angle of arrival(AOA)location techniques and propose corresponding localization algorithms.The main works are included as follows:1.The mathematical models of TDOA and AOA location techniques are given based on the basic pricinple of localization.Several classical location algorithms are described and analyzed.In addition,the sources of positioning errors are analyzed and the evaluation criteria of positioning performance are given.2.A location method combining closed-form solution and iterative algorithm and a variance weighted location algorithm are proposed under line of sight(LOS)propagation environment.The closed-form solution is relatively robust and does not diverge.The iterative solution has a high accuracy if the geometric dilution of precision(GDOP)is good,but it is easy to diverge in the case of poor GDOP.The proposed method,which combines the closed-form solution and iterative algorithm,integrates both advantages and avoid their disadvantages.It ensures stability and acheives high-accuracy performance.The high computational burden caused by matrix inversion makes it hard for previous high-performance solutions to be applied to low-cost hardware devices.To solve this problem,a simple closed-form solution for a multistation redundancy localization system is derived by using the estimation variance as the weighting coefficient to compute an average of each group's localization result.The proposed method,with simple algebraic solution,requires no matrix inversion and can be used for low-cost hardware devices.3.For the positioning problem under non-line of sight(NLOS)propagation environment,two novel localization algorithms based on machine learning are presented in this thesis.Conventional learning location methods,which predict the value of a previously unseen data point by building up a mapping from signal feature to target position,actually complicate the problem.It is not necessary to take training data to build up the non-linear relationship between LOS measurements and target position.This non-linear relationship has already been accurately modeled.Thus,using gradient boosting regression tree(GBRT)algorithm to correct TDOA measurements and solving target location with WLS algorithm,a learning location method based on distance difference correction is proposed in NLOS environment.Previous NLOS localization algorithms based on channel impulse response(CIR)either utilize only part information of CIR or require time synchronization.By using nonparametric kernel regression(NKR),a learning location algorithm based on relative time CIR is derived for wireless location.This approach does not need time synchronization at the transmitter and utilizes the whole imformation of CIR to enhance positioning accuracy.Winprop electromagnetic simulation software is applied here to model the channel with 3D ray tracing.Numerical examples are provided to illustrate the performance of the proposed method in root-mean-square error(RMSE).The proposed method has a significantly reduced RMSE compared with the method based only on TDOA.