TRDMU: An End-to-End trajectory-road discrepancy-aware Map Update Framework

Abstract

Accurate digital map maintenance is essential for modern navigation systems, where vehicle trajectories offer automated updating potential through their extensive coverage and continuous collection. Analyses from major mobility service providers indicate that missing roads introduced by urban renewal or expansion are the most prevalent map errors, making automated missing road updating a critical requirement for reliable and safe navigation systems. Conventional methods sequentially infer the geometry of missing roads and then reconstruct the topology by predicting connection point locations; such multi-stage updating pipelines inevitably cause error propagation and exhibit poor performance in complex structures such as parallel-roads and intricate intersections. To address these limitations, we propose TRDMU, an end-to-end framework for map updating that jointly tackles: (i) missing road geometry inference through a dual-view encoder and context-aware cross-view fusion module that integrates local and global contextual information to capture cross-view discrepancies; and (ii) precise connection point localization through topology-enhanced connectivity modeling, which constructs a grid-level candidate connection point graph from the existing map, augments each candidate with a fused context embedding, and then applies a topology enhanced connectivity embedding module to propagate contextual topology and learn connectivity attention weights among candidates. Extensive evaluation on real-world datasets demonstrates that TRDMU outperforms state-of-the-art methods, achieving a 6.72% improvement in GEO F1-score on a large-scale industrial trajectory dataset while maintaining robust performance across diverse urban road environments.

Keywords

trajectory-based map updating / missing road detection / trajectory-road discrepancy

RIGHTS & PERMISSIONS

Higher Education Press 2026

Just Accepted

This article has successfully passed peer review and final editorial review, and will soon enter typesetting, proofreading and other publishing processes. The currently displayed version is the accepted final manuscript. The officially published version will be updated with format, DOI and citation information upon launch. We recommend that you pay attention to subsequent journal notifications and preferentially cite the officially published version. Thank you for your support and cooperation.