Towards A Foundation Model For Trajectory Intelligence
This work addresses trajectory intelligence for applications like urban planning or mobility analysis, but appears incremental as it adapts existing pre-train/fine-tune methods to this domain.
The authors tackled the problem of creating a foundation model for trajectory intelligence by training a large model on over 2 billion real-world user check-ins, achieving effective learning of underlying patterns that enabled application to three downstream tasks.
We present the results of training a large trajectory model using real-world user check-in data. Our approach follows a pre-train and fine-tune paradigm, where a base model is pre-trained via masked trajectory modeling and then adapted through fine-tuning for various downstream tasks. To address challenges posed by noisy data and large spatial vocabularies, we propose a novel spatial tokenization block. Our empirical analysis utilizes a comprehensive dataset of over 2 billion check-ins generated by more than 6 million users. Through fine-tuning on 3 downstream tasks we demonstrate that our base model has effectively learned valuable underlying patterns in raw data, enabling its application in meaningful trajectory intelligence tasks. Despite some limitations, we believe this work represents an important step forward in the realization of a foundation model for trajectory intelligence.