Research On Time Delay Estimation Algorithm In Radiolocation

Radiolocation is very ancient and young technology. With the development of communication technology and the advent of information age, radiolocation technology has been applied into all area of people's life which include security services, information services, navigation services, tracking services and many other aspects, it's changing the way of people's life. Radiolocation technologies estimate the geometry position of a target mobile by detecting the characteristic parameters of the signals which transmit between the mobile terminal and several fixed transceivers. Common characteristic parameters include Strength of Arrival (SOA), Angle of Arrival (AOA), Time of Arrival (TOA) and Time Difference of Arrival (TDOA). The location method based on TOA/TDOA is widely used at present, and time delay estimation is the main contents of this thesis.The estimation algorithms for time delay parameter in two cases of fixed time delay and time-varying delay are studied respectively and the Bayesian estimation methods for time-varying delay is mainly studied in this paper.Generalized Cross Correlation (GCC) method, third-order cumulant method, bispectrum method and joint method have been studied in accordance with fixed time delay. GCC method is classical and simple, but it requires that signal and noise, noise and noise are independent, while high order cumulant method can solve this problem. A series of Bayesian estimation methods including Extended Kalman Filtering (EKF) method, Unscented Kalman Filtering (UKF), Particle Filtering (PF) method, Extended Kalman Particle Filtering (EPF) method and Unscented Particle Filtering (UPF) method have been studied in accordance with time-varying delay. Kalman Filtering is the optimal estimation for linear system under Gaussian noise. When the system is nonlinear, EKF does linear approximation for nonlinear system using first order Taylor series expansion and it can provide a suboptimal estimation. UKF is based on UT translation, it adopts the structure of linear Kalman Filtering, doesn't neglect higher order terms and does not need to compute Jacobian matrix, so it has a higher accuracy than EKF. PF is a Bayesian method based on Monte Carlo which is widely used in recent years. Its basic idea is to use a series of sample particles to approximate the posterior probability distribution of the state, and further to estimate the state. Its characteristics and advantage is to deal with state estimation under the nonlinear and non-Gaussian environment. Choice of importance function is the key of Particle Filtering, which directly determines the filtering performance of particle filter. The importance function of standard PF does not consider the current observations, the importance function of EPF is generated based on EKF, the importance function of UPF is generated based on UKF, these two important functions have fully considered the current observations and they are much closer to the realistic posterior probability distribution, so the filtering performance of EPF and UPF is better PF. The said algorithms are discussed in detail and used to estimate the time delay in the paper, whose experimental results have certain theoretical significance and practical value.Finally, a summary of this thesis is advanced, and the future works on the subject are viewed.