conference

DGNSS Corrected Pseudorange and Time-Differenced Carrier Phase (TDCP) Measurements using Differentiable Factor Graph Optimization (DFGO)

Hoi-Fung Ng, Penghui Xu, Yihan Zhong, Guohao Zhang, Weisong Wen, Li-Ta Hsu

ION GNSS+, The International Technical Meeting of the Satellite Division of The Institute of Navigation · Oct 2024

time-differenced-carrier-phase differentiable-factor-graph-optimization

Abstract

Google Smartphone Decimeter Challenge (SDC) provides a dataset for the communities to verify their leading-edge algorithm for precise smartphone GNSS positioning under the same baseline. This year’s dataset provides more variety on the vendors of smartphones’ GNSS chips, where the new data from affordable smartphones is believed to be noisier, increasing the position uncertainty. This study integrates the pseudorange, time-differenced carrier phase (TDCP) measurements, and estimated velocity using factor graph optimization (FGO) to provide an optimal and robust position estimation. In the optimization process, properly excluding and weighting the measurement is the key for only using healthy measurements for state estimation. This study also uses the differentiable FGO (DFGO) to determine the covariance for the GNSS positioning adaptively. The positioning results show that the network can provide a more robust and reliable performance than the traditional model-based approach.

Problem Addressed

Precise smartphone GNSS positioning is challenged by the noisier GNSS chips of affordable smartphones included in the Google Smartphone Decimeter Challenge, which increase position uncertainty. Robust estimation requires properly excluding and weighting unhealthy measurements.

Key Contributions

Factor graph optimization integrating DGNSS-corrected pseudorange, time-differenced carrier phase, and estimated velocity
Differentiable factor graph optimization adaptively learning measurement weighting for GNSS positioning
Outlier detection and exclusion schemes using CORS-based observation-space representation corrections
Least-squares initialization with consistency checks for position, velocity, clock offset, and clock drift

Conclusions

Evaluation on the challenge training data shows the FGO framework provides better positioning performance than the traditional model-based approach, with outlier exclusion and Sigma-epsilon weighting giving the best results. Selecting and weighting healthy measurements is shown to be critical for reliable FGO-based GNSS positioning.