Map-Aware Super-Resolution GPS Trajectory Reconstruction via Machine Learning

This research aims to realize a novel system for reconstructing high-resolution GPS trajectory data from truncated or low-resolution inputs, addressing the challenge of balancing data utility with privacy preservation in mobility applications. The system integrates transformer-based encoder-decoder models with graph convolutional networks (GCNs) to effectively capture the temporal dependencies of trajectory data and spatial relationships in road networks. By leveraging these techniques, the system restores fine-grained trajectory details lost due to data truncation or rounding, which are common practices for privacy protection.

Published Paper

• Yonekura, H., Ozeki, R., Rizk, H., & Yamaguchi, H. (2024, October). Restoring Super-High Resolution GPS Mobility Data. In Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Geo-Privacy and Data Utility for Smart Societies (pp. 19-24).　https://dl.acm.org/doi/abs/10.1145/3681768.3698501