M-STAR: Multi-Scale Spatiotemporal Autoregression for Human Mobility Modeling

TL;DR

M-STAR tackles long-term human mobility trajectory generation by introducing a coarse-to-fine framework that explicitly models multi-scale spatiotemporal patterns. It combines a Multi-scale Spatiotemporal Tokenizer (MST-Tokenizer) to compress trajectories into residual tokens and a STAR-Transformer to autoregressively generate finer scales conditioned on coarser context and movement attributes. Across two real-world weekly datasets, M-STAR delivers superior fidelity and 15x–30x faster generation than diffusion-based baselines, while supporting downstream tasks like next-location prediction and epidemic simulation with strong privacy safeguards. The approach demonstrates the value of structured multi-scale modeling for scalable, realistic, and privacy-conscious mobility synthesis.

Abstract

Modeling human mobility is vital for extensive applications such as transportation planning and epidemic modeling. With the rise of the Artificial Intelligence Generated Content (AIGC) paradigm, recent works explore synthetic trajectory generation using autoregressive and diffusion models. While these methods show promise for generating single-day trajectories, they remain limited by inefficiencies in long-term generation (e.g., weekly trajectories) and a lack of explicit spatiotemporal multi-scale modeling. This study proposes Multi-Scale Spatio-Temporal AutoRegression (M-STAR), a new framework that generates long-term trajectories through a coarse-to-fine spatiotemporal prediction process. M-STAR combines a Multi-scale Spatiotemporal Tokenizer that encodes hierarchical mobility patterns with a Transformer-based decoder for next-scale autoregressive prediction. Experiments on two real-world datasets show that M-STAR outperforms existing methods in fidelity and significantly improves generation speed. The data and codes are available at https://github.com/YuxiaoLuo0013/M-STAR.

Figures (8)

• Figure 1: High-level overview of M-STAR’s coarse-to-fine trajectory generation across spatial and temporal scales.
• Figure 2: Overview of the M-STAR framework: a coarse-to-fine trajectory generation process that first tokenizes input trajectories into Multi-Scale Spatiotemporal Tokenizer (MST-Tokenizer), followed by autoregressive next-scale prediction using STAR-Transformer.
• Figure 3: The pipeline of the Multi-Scale Residual Quantization module, illustrating the encoding and decoding phases.
• Figure 4: Visualization of community-level origin-destination flows in Shenzhen
• Figure 5: Comparison of trajectory diversity between real and generated datasets. The Diversity Error measures the relative difference in the number of unique trajectories between generated and real data.