%0 Journal Article
%T A Unified Framework for GPS Code and Carrier-Phase Multipath Mitigation Using Support Vector Regression
%A Quoc-Huy Phan
%A Su-Lim Tan
%A Ian McLoughlin
%A Duc-Lung Vu
%J Advances in Artificial Neural Systems
%D 2013
%I Hindawi Publishing Corporation
%R 10.1155/2013/240564
%X Multipath mitigation is a long-standing problem in global positioning system (GPS) research and is essential for improving the accuracy and precision of positioning solutions. In this work, we consider multipath error estimation as a regression problem and propose a unified framework for both code and carrier-phase multipath mitigation for ground fixed GPS stations. We use the kernel support vector machine to predict multipath errors, since it is known to potentially offer better-performance traditional models, such as neural networks. The predicted multipath error is then used to correct GPS measurements. We empirically show that the proposed method can reduce the code multipath error standard deviation up to 79% on average, which significantly outperforms other approaches in the literature. A comparative analysis of reduction of double-differential carrier-phase multipath error reveals that a 57% reduction is also achieved. Furthermore, by simulation, we also show that this method is robust to coexisting signals of phenomena (e.g., seismic signals) we wish to preserve. 1. Introduction Multipath is defined as one or more indirect replicas of the line-of-sight (LOS) signal from satellites arriving at a receiver¡¯s antenna from a satellite. It normally occurs due to reflection from objects in the vicinity of the receiver and constitutes a major error source that contaminates receivers¡¯ measurements, resulting in performance degradation of GPS positioning solutions. The errors induced by multipath are typically up to 15 meters for C/A code [1] and a few centimeters for carrier-phase measurements [2]. Multipath mitigation is hence important for a variety of applications which utilize this data, such as ionospheric monitoring [3], geodesy [4, 5], and navigation [6]. On the one hand, multipath mitigation is a very challenging task. As multipath is site-dependent, differencing measurements among multiple short-baseline receivers (DGPS) are unlikely to help. Furthermore, aggressively removing multipath error may harm wanted coexisting information and perturbations such as seismic signals induced by an earthquake, as their frequency spectra likely overlap with that of the contaminating multipath error. Various mitigation approaches have been proposed in the literature, classified into either frequency-domain or time-domain processing. The former is based on spectral analysis of multipath error in the frequency domain using fast fourier transform (FFT) [7], or wavelet decomposition [8, 9]. However, they unintentionally rule out other coexisting signals. To
%U http://www.hindawi.com/journals/aans/2013/240564/