Closely spaced multipath mitigation in GNSS receiver based on maximum likelihood estimation

Multipath is the dominant source of positioning error in modern GNSS receiver. Maximum likelihood (ML) parameter estimation is an optimal method to mitigate the multipath effects while ML involves nonlinear optimization and requires iterative algorithms. Iterative methods usually lack of global convergence when the paths are closely spaced, if the initial value is arbitrarily assigned. In this paper, however, we first employ a grid search method to choose the initial value before iteration. Most computation of the grid search can be done offline. After that, an iterative method with simple forms is used to improve the parameter accuracy and global convergence can be achieved with just a few iterations. The simulations results show the estimator of time delay is almost unbiased when the time relative delay of two paths is larger than 0.20 chips.

I. Introduction

The presence of multipath greatly affects the localization performance of GNSS system. In GNSS receivers, due to multipath, the correlation between the local PN sequence and received signal no longer has an ideal triangular shape but is superimposed by many delayed versions of auto correlation function (ACF). Conventional methods such as Early-Late (E-L) delay locked loop (DLL) which are easily trapped into the wrong equilibrium point, cannot tackle this problem, thus resulting great ranging error, especially when the LOS path is not dominant. In the past years, various methods have been proposed to reduce the multipath errors, which can be categorized into two groups: modified DLL method and multipath parameters (complex amplitude, time delay) estimation methods. The first group of methods try to simply modify the traditional DLL and make it capable of mitigating multipath, typical methods including narrow spacing correlator, strobe correlator, double delta correlator [1]–[3]. The latter group of methods employ parameter estimation methods to resolve the multipath [4]–[11].