Towards GNSS Ambiguity Resolution for Smartphones in Realistic Environments: Characterization of Smartphone Ambiguities with RTK, PPP, and PPP-RTK

With a lower noise level of GNSS phase measurements compared to pseudoranges, there will be a significant improvement in GNSS positioning performance if the integer property of ambiguity terms can be recovered for smartphone positioning. In a realistic smartphone user environment, GNSS measurements suffer from severe noise and multipath effects, which can cause problems in estimating reliable and stable ambiguity terms and further decrease the possibility of conducting ambiguity resolution with raw smartphone GNSS observations. Therefore, this paper focuses on evaluating the characteristics of ambiguities in smartphones with different precise positioning modes: PPP, RTK, and PPP-RTK. Datasets collected in realistic driving environments which contain open sky, moderately suburban, overpass, and vegetation are used in this study. RTK, PPP, and PPP-RTK processing are applied to obtain double-differenced and un-differenced ambiguities, respectively. Analysis on ambiguity availability shows that there is a heavy signal loss in observations, with a maximum percentage of 60% for E33 in test dataset, and only a few satellites have continuously tracked dual-frequency data. RTK has the best positioning performance of 1.1 m for 95th percentile horizontal positioning error, while with regional atmospheric corrections, PPP-RTK performs slightly better than PPP. The ambiguity time series shows similar trends for all three modes, and GLONASS has the poorest ambiguity stability with an STD of 0.35 m, which is larger than one cycle of the wavelength. The Melbourne-Wübbena-derived wide-lane ambiguity shows a good consistency with raw ambiguity-derived wide-lane ambiguity, but some of the differences can reach 5-10 cycles. However, the fractional parts of both wide-lane ambiguities look like noise, which would cause difficulty in correctly fixing WL ambiguity in realistic environments.