PositionNet: CNN-based GNSS positioning in urban areas with residual maps

Xu, P., Zhang, G., Yang, B., Hsu, L. T.

Applied Soft Computing (2023)

Key idea. PositionNet learns to correct urban GNSS errors with a CNN that reads residual maps — turning the spatial pattern of pseudorange residuals into a learned position correction.

Impact. Shows that data-driven methods can complement model-based 3D-mapping-aided GNSS, an early thread in the lab’s AI-for-positioning direction.