Research On TOA/TDOA Parameters Estimation Methods For Wireless Location

With the vigorous development in mobile communications today, locationservice provided through wireless network becomes the new favorite ofmanufacturers, operators and users by its unique attraction practicality. Wirelesslocation service is to provide the mobile terminals with their real-time locationinformation in effective wireless coverage. It is widely applied in public security,wireless network management, radar and sonar location,intelligent trafficsystem and earthquake detection etc.Wireless location technologies estimate the geometry position of a targetmobile by detecting the characteristic parameters of the signals which transmitbetween the mobile terminal and several fixed transceivers. We often use somecommon signal characteristic parameters such as Angle of Arrival (AOA), Timeof Arrival (TOA), Time Difference of Arrival (TDOA) and Strength of Arrival(SOA). Although the SOA location method is easy, it is bad in precision. TheAOA location method is of certain precision, but the receiving set is verycomplex. The TOA location method is high precision but has high requires ontime synchronization. The TDOA location method can eliminate the dependenceon time datum, reduce the cost and still guarantee certain locating accuracy. Themethod to mix the methods above can satisfy certain location precision, but itneeds much modification with the network equipment of the cell system existed.Consider comprehensively with the location precision,the cost and other factors,the dissertation adopts the wireless location technology based on TOA/TDOAdetecting.This dissertation focuses on TOA/TDOA, the main technology used inwireless location system. We have given the TOA/TDOA estimation model, andhave done thorough research on the following algorithms under the condition offixed time delay and time-varying delay. We have studied Generalized CrossCorrelation(GCC) method and Cross-Bispectrum method in accordance withfixed time delay, and studied a series of Bayesian methods in accordance withtime-varying delay, such as Extended Kalman Filter(EKF),Unscented KalmanFilter(UKF) and Particle Filter(PF).The focus of this dissertation is to make a study of Bayesian methods.The essence of Bayesian estimation principle is to attempt to construct theposterior probability density function of the system states based on all availableinformation. It is that to predict the prior probability density of the state based onthe system model, then update it with the variables at the latest time step, so wecan obtain the posterior probability density.If the system is linear, and the noises are Gaussian, then the Kalman Filteralgorithm provides the optimal solution. Otherwise, if the system is nonlinear,then a local linearization of the system equations may be a sufficient descriptionof the nonlinearity. The EKF is based on this approximation, and provides asuboptimal estimation. The UKF is based on UT translation, it adopts thestructure of linear Kalman Filter, and doesn't need to calculate the partialderivative of Jacobian matrix. The UKF can achieve a precision of the secondorder at least. But the two algorithms mentioned above do not take account of thewhole statistic characteristics of the process, and they may result in low accuracy.The Particle Filter algorithm is a Bayesian approach based on the ideal ofMonte Carlo, it has been brought up in recent years. The main feature of the PF isto use a large number of particles to describe the posterior probability densityfunction of the related parameter approximately, and to obtain efficient estimationof the parameters we need. The PF can estimate nonlinear and non-Gaussiansystems efficiently.There are six chapters in this dissertation.Chapter 1 has recounted the basic location methods of wireless locationtechnologies, and brought up the signal model of time estimation based onTOA/TDOA and the development of time estimation. At the end of this chapterthe main content of this dissertation is introduced.Chapter 2 has introduced some basic theories of the work in the dissertation,such as some characteristics of high order cumulants.Chapter 3 has studied the time estimation methods of GCC algorithm basedon the second order statistics and Cross-Bispectrum algorithm based on the thirdorder statistics.Chapter 4 has introduced the concepts of Bayesian estimation, and analyzedthe KF and the algorithms extended in detail, such as EKF,UKF. Furthermore,the dissertation has given the demonstration of the recursive formulas andsummaries of the algorithms. In this chapter, the author has brought up the idealof using the EKF and UKF algorithms to estimate time-varying delay, laid out thedetail deduction steps and the simulation results.Chapter 5 has introduced the concept of Particle Filter, and explained thekey ideal of the PF algorithm. Besides the dissertation has analyzed the SISalgorithm in detail, and further introduced the improved algorithm—SIR. Also,the author has given the simulation comparison of these two methods undernonlinear and non-Gaussian models. This chapter has reveald the ideal of usingthe PF algorithm to estimate time-varying delay, and laid out the detail deductionsteps and the simulation results.Chapter 6 has summarized the work content of this dissertation, and madefurther prospects according to the status of the research work now.The main works and contributions of this dissertation are as following:The dissertation has studied the characteristics of high order cumulants,analyzed GCC and Cross-Bispectrum algorithms applied in the signal model offixed time delay concretely, and made simulations under the condition of singlepath propagation and multi-path propagation.The dissertation has deducted the theory of classic Kalman Filter, analyzedthe EKF and the UKF which is not maturity, and summarized the steps ofalgorithms, moreover. The author has brought up the ideal of time-varying delayestimation based on the EKF, UKF, and provided the simulation results.The dissertation has analyzed the basic ideal of Particle Filter which is ahotspot with nonlinear and non-Gaussian system in recent years, introduced theSIS which is the basic algorithm in PF in detail, and summarized the algorithmsteps according to the analysis, then brought out the SIR algorithm in accordancewith degeneracy problem. The author has brought up the ideal of time-varyingdelay estimation based on the PF, and deducted the steps that the SIR applied totime-varying delay estimation model.This dissertation has also given some opinions on the combination of theUKF and the PF, and made some prospects of this ideal. It is the work in the nextstage with the author.