Precise relative positioning using smartphone GNSS carrier-phase-only observations

Abstract

Today, GNSS (Global Navigation Satellite System) functionality is ubiquitous in smartphones, with billions of devices in use globally. However, the positioning accuracy offered is limited to several meters. In this paper, we aim to provide a precise, centimeter-level solution using recent smartphones and relative positioning mode with the MAFA (Modified Ambiguity Function Approach) algorithm fed with carrier-phase-only observations. The impact of baseline length and employed constellations was analyzed. Data was collected by three smartphones: Xiaomi Mi 8, Google Pixel 7, and Samsung S24+. The results show that positioning performance varies across devices even under similar observation conditions. Nonetheless, the number of satellites observed by each smartphone had the greatest influence on the results. The results demonstrated the high applicability of the MAFA method fed with smartphone GNSS data, not only in a single-constellation case but also in a multi-constellation case. Moreover, thanks to the method’s unique property, it is immune to cycle slips and discontinuities in ambiguity, which frequently occur in smartphone data.

The results show the expected degradation in positioning accuracy with increasing baseline length. The largest decrease occurs in the vertical component, while the horizontal components are less affected. For the Google Pixel, the greatest accuracy degradation is observed for BDS-2. For the ultra-short baseline, the STDs are 2.5 cm and 1.4 cm for the north and east components, increasing to 10.6 cm and 2.9 cm, respectively, for the 5 km baseline. Interestingly, the vertical STD changes only slightly. For the Xiaomi Mi 8 and Samsung S24+, baseline-induced STD changes do not exceed 3 cm in any component. The multi-constellation solution shows the smallest accuracy degradation, with STD differences below 1 cm between baselines. While coordinate biases are negligible for the ultra-short baseline, noticeable biases appear for all three smartphone models with the longer baseline.