Session E4b: Smartphone Decimeter Challenge (Co-sponsored by Google)

Third Place Winner of the Smartphone Decimeter Challenge: Improving Smartphone Accuracy with a Two-step Accurate Velocity Estimation
Jeonghyeon Yun, Suyeol Kim, Taejin Youn, Gyeongmin Kim, Wonwoo Park, Byungwoon Park, Sejong University
Date/Time: Thursday, Sep. 19, 4:23 p.m.

Peer Reviewed

This paper describes the work that won third place in the Google Smartphone Decimeter Challenge 2023-2024 (GSDC 2023-2024), held from September 2023 to May 2024. Google provided dynamic smartphone GNSS and IMU data for 196 different drive routes, offering a wider variety of smartphone chipsets compared to the previous GSDC 2021 and GSDC 2022 competitions. The goal of the competition was to obtain accurate position information using smartphone measurements, which are characterized by high noise and low quality compared to commercial GNSS receivers. The competition allowed the use of various post-processing strategies, such as non-commercial open-source programs or machine learning (ML).
In this paper, we present a real-time GNSS positioning strategy for smartphones that can be applied in real-life situations. First, the raw GNSS measurements of pseudo-range, Doppler, and carrier-phase are processed through two steps: predict and update. Using time-synchronized Doppler measurements, the current position and velocity are predicted from the previous state. The current position and velocity are then updated by the corrected pseudo-range for DGNSS position and the time-differenced carrier-phase (TDCP) velocity. The Doppler-based velocity and TDCP velocity, calculated at different times, are used in the position time update and to validate and exclude invalid pseudo-range and carrier-phase measurements.
Through the proposed algorithm, we achieved mean (50%, 95%) scores of 0.928m (public) and 1.342m (private) on the final leaderboard, securing third place. The final scores for the late submission after the competition were 0.890m (public) and 1.191m (private).