Objectives With the popularization of smartphones and the evolution of global navigation satellite systems (GNSS), navigation contributes as an indispensable function for smartphones, and the demand of high-precision positioning for mass-market smartphones is becoming higher. From 2016, developers can obtain GNSS raw observations through Google Android application programming interface (API). Android 9.0 and above versions, provide the option to turn off the duty cycle, which makes it possible for smartphones to conduct precise point positioning (PPP).
Methods This paper conducts quality analysis of the GNSS observations for several common-used dual-frequency smartphones, Huawei Mate 20/30 and Xiaomi 8, and propose a set of quality control schemes that is suitable for smartphone PPP. Gross errors can be removed by data detection according to satellite elevation, signal-to-noise ratio, pre-residuals, etc., while cycle slips can be detected by the accumulated delta range state (ADRS) identifiers as well as the time difference carrier phase (TDCP) observations. A real-time PPP application based on Android platform was developed. It supports for singe- and dual-frequency, real-time and post-processing PPP. To verify the performance of PPP with smartphones, several field experiments were performed with different smartphones in different scenarios.
Results Field results show that: (1) The pseudorange quality of Xiaomi 8 and Huawei Mate 20 are similar, while the carrier phase quality of Huawei Mate 20/30 is significantly worse than Xiaomi 8. (2) The ADRS identifier of Xiaomi 8 can better identify the cycle slip, while it is difficult for Huawei Mate 20/30 to directly determine the cycle slips through the ADRS identifier. By combining the ADRS and TDCP can detect most of cycle slips. (3) The positioning accuracy of single-frequency PPP for Xiaomi 8 and Huawei Mate 20/30 reach 0.5-0.6 m in horizontal, and 1.0-2.0 m in vertical. The positioning accuracy of dual-frequency PPP is better than that of single-frequency PPP after convergence, especially in the vertical direction. (4) For the real-time and post-processing PPP, they both have a similar positioning accuracy in the horizontal components, while the positioning accuracy of post-processing PPP in the vertical direction is higher than that of real-time PPP by 20%-40%.
Conclusions An accuracy of a few decimeters can be achieved for real-time PPP with Android smartphones. Different from geodetic GNSS receivers, the widely-used smartphones use low-cost linearly-polarized antennas. In addition, smartphones are often used in urban complex scenarios, thus the GNSS data they collect have more gross errors, cycle slips and multipath effects. Further study should focus on the multipath mitigation, multi-frequency and multi-GNSS combination, and fusion with other sensors such as gyroscopes, magnetometers, barometers, bluetooth, Wi-Fi, etc. to realize continuous and seamless positioning both indoor and outdoor.