Evaluation, Refinement and Fusion of Software-Based Pseudorange Multipath Mitigation Techniques

Multipath remains a significant error source to the GPS user even when employing a receiver and antenna designed for multipath rejection. This study evaluated data processing methods to further remove pseudorange multipath from observations collected by a GPS reference station using multipath-resistant hardware. Both templateand surface-based empirical models were implemented as batch processes. Both analytic and algorithmic enhancements to the processes are presented. Previous studies have evaluated the effectiveness of multipath mitigation algorithms by looking at higher order statistics of the multipath in the observations after filtering. In contrast, this effort evaluated the effectiveness of multipath removal techniques by forming position estimates with the filtered observations. The quality of the position estimate is summarized using two metrics, the volume of the 3σ error ellipsoid and RMS of range residuals. After investigating several error models and processing techniques for pseudorange multipath and related errors, the most effective models were applied to a control set of observations. The application resulted in consistent improvement in positioning results. The RMS of range residuals improved by an average of 18%, and the volume of the 3σ error ellipsoid reduced by 60%. INTRODUCTION Multipath occurs when reflections of an electromagnetic wave interfere with the line-of-sight wave. In GPS, multipath interference corrupts the signals on both the L1 and L2 frequencies. Figure 1 depicts how multipath can propagate similarly to optical reflection. Both the direct and reflected signals carry time codes; the receiver must extract these codes to estimate the range to the GPS satellite. However, multipath signals corrupt the codes carried by the direct path signal, skewing the code correlation processes within the receiver. The skewed correlation results in errors in the measurement of all the range measurements, notably the pseudorange and accumulated Doppler range (ADR) observations. Figure 1: Multipath on the GPS Signal There are a number of ways to reduce the impact of multipath on the GPS pseudorange. Selecting a site with a First presented at ION-GPS 2002, Portland, Oregon, September 25, 2002 minimal number of nearby reflectors is the first step. Choke rings greatly diminish the amplitude of multipath received from low elevation reflectors or via the ground plane. Multipath can be mitigated within the receiver as well. Most geodetic quality receivers employ correlation processes that can detect and remove the effect of a single reflector. Finally, some receivers smooth the pseudorange using the carrier-phase. Despite these measures, multipath errors still corrupt the pseudorange. This study evaluated post-processing methods to further eliminate pseudorange multipath for reference stations that employ multipath-resistant hardware. Two classes of models have been developed to empirically model pseudorange. These models fall into two general categories: templates [1, 4] and surfaces [3,6,9]. Mapping multipath to bins by azimuth and elevation was proven an effective modeling strategy for calibrating carrier-phase measurements collected on spacecraft [11]. This technique is equally promising in mapping the pseudorange multipath for a static GPS reference station. A primary goal in this study was to evaluate the strengths and weaknesses of the various empirical modeling techniques by applying the models. Pseudoranges, corrected for multipath, were used to generate position estimates with precise ephemerides. The multipath models were validated using two years of observations from a reference station participating in the IGS. By using a large dataset, model consistency in space and model deterioration over time was explored.