.png){kind=icon}
.png){kind=icon}
Return
DuGTRL: Dual-view grid-based trajectory representation learning framework integrating spatiotemporal semantics
.png){kind=icon}
PRE
AI
Y
俞
Z
陈
郑
L
DOI：10.1016/j.eswa.2026.133896.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• Models grid-based trajectories as both spatiotemporal images and static graphs. • Fuses local and global patterns via parallel CNN and GAT encoders. • Eliminates reliance on external road networks via semantic feature injection. • Integrates Contrastive Learning and MLM for robust self-supervised training. • Achieves SOTA performance on travel time estimation and classification tasks.
Keywords:
Trajectory representation learning
Dual-view encoder
Grid trajectory
Cross-view fusion
Self-supervised learning
Journal
IF:
7.5
Papers:
2.9W
Citations:
10.2W
Organization
Cited Papers