Research On Theory And Methodology Of GNSS Multipath Estimation

Multipath is one of the main error sources in global navigation satellite systems (GNSS). Inapplications of GNSS with high precision, multipath is the most difficult error source to dealwith. Multipath estimation is one of the receiver-internal multipath mitigation approaches. In thisthesis, the multipath estimation technology is researched in details, and several multipathestimation methods are proposed on the basis of in-depth analysis of existing multipathmitigation and estimation methods. The main works and creations are as follows.1. The effect of multipath on GNSS receiver tracking loops and the performances of somerepresentative multipath mitigation technologies are analyzed, and lots of conclusions are drawnfrom the analysis, which provide a basis for the establishment of multipath estimationtechnology.2. For multicorrelator multipath estimation, multipath grid maximum likelihood estimation(GMLE) is proposed based on the idea of-deploying method. A proper configuration for thebank of correlators is advised to avoid the interference from ineffective multipath signals. Thesteps of multipath detection and estimation method are established, through which the delay,amplitude, and carrier phase estimations of multipath signals can be estimated.3. The GMLE is analyzed from two aspects: statistical accuracy and numerical stability,revealing that the ill-conditioning in the normal matrix is the causation of poor performances ofthe GMLE. Several types of biased multipath estimations including ridge estimation andtruncated singular value decomposition (TSVD) estimation are proposed, and the property of theestimations are discussed. The selection methods for the bias parameters of the estimations areadvised based on the characteristics of the ill-conditioning existing in the model. Biasedestimation mitigates the strong effects of noise to a certain extent.4. Multipath maximum likelihood estimation and the optimization methods to solve theestimation are researched based on the highly compressed and analytic measurement model.Several nonlinear optimization technologies are utilized to solve the multipath maximumlikelihood estimation, and the performances of the different optimization technologies arediscussed. For the problem of the unknown number of the multipath signal components, basedon the theory of model order selection, the rule and method to fix the number of the multipathsignal components are proposed.5. For the characteristics that the noise in multipath measurement is additive Gaussian whitenoise, square-root unscented Kalman filter using nonaugmented state vectors is proposed andapplied in the multipath estimation problem. Compared with traditional unscented Kalman filter, the new method has lower computational burden, better accuracy of estimation and betterresistance performance to filter dispersion. The new method can mitigate multipath effect moreefficiently than traditional multipath mitigation technology, and can estimate multipathparameters in almost the same precision as marginalized particle filter with much lowercomputational burden.6. Cubature Kalman filter is applied in the multipath estimation problem. The computationalburden of square-root cubature Kalman filter is lower than square-root unscented Kalman filter,and the estimation precisions of the two filtering methods are at the same level.7. For the problems that the transformation of multipath signal components is complicatedand it is difficult to characterize how multipath signal components appear and disappear and howthe values of multipath parameters transform, the kinematic model of only the direct signalcomponents is proposed, and multipath estimation based on particle filtering is implementedusing the kinematic information of the direct signal components. In the update steps of theparticle filtering, the parameters of the direct signal component are estimated combiningobservation and kinematic information while the parameters of multipath components areestimated using only observation information. The building of proposals and the implementationof the filtering are given. The method can use the kinematic information of the direct signals toimprove the accuracy and meanwhile avoid misusing of the incorrect kinematic information ofmultipath signals.