Spatio-Temporal Trajectory Foundation Model - Recent Advances and Future Directions
Sean Bin Yang, Ying Sun, Yunyao Cheng, Yan Lin, Kristian Torp, Jilin Hu
TL;DR
The paper addresses the need for robust spatio-temporal trajectory foundation models (TFMs) by formalizing a trajectory as $T=\langle (x_1,y_1,t_1),\dots,(x_n,y_n,t_n)\rangle$ and surveying recent advances across data modalities and learning paradigms. It proposes a taxonomy that separates data-modality oriented methods from foundation-model learning methods, covering contrastive, generative, hybrid, and causal approaches with representative systems such as TrajRL, MM-Path, LightPath, START, and TrajCL. The key contributions include a structured overview of single- and multi-modality trajectories, critical analysis of each paradigm’s strengths and limitations, and a forward-looking agenda highlighting responsible AI, scalability, and multi-modal data integration. This work advances spatio-temporal general intelligence by clarifying methodological options, identifying open challenges, and outlining directions for sustainable, transferable TFMs with potential impact on travel, traffic management, and urban planning.
Abstract
Foundation models (FMs) have emerged as a powerful paradigm, enabling a diverse range of data analytics and knowledge discovery tasks across scientific fields. Inspired by the success of FMs, particularly large language models, researchers have recently begun to explore spatio-temporal foundation models (STFMs) to improve adaptability and generalization across a wide spectrum of spatio-temporal (ST) tasks. Despite rapid progress, a systematic investigation of trajectory foundation models (TFMs), a crucial subclass of STFMs, is largely lacking. This tutorial addresses this gap by offering a comprehensive overview of recent advances in TFMs, including a taxonomy of existing methodologies and a critical analysis of their strengths and limitations. In addition, the tutorial highlights open challenges and outlines promising research directions to advance spatio-temporal general intelligence through the development of robust, responsible, and transferable TFMs.