A Universal Model for Human Mobility Prediction

TL;DR

The paper tackles unifying human mobility prediction across two modalities: individual trajectories and crowd flows. It introduces UniMob, a diffusion-transformer framework with a multi-view mobility tokenizer and a bidirectional Individual-Collective Alignment mechanism to learn shared spatiotemporal patterns across modalities. The model uses a joint noise predictor and modality-specific predictors, optimized with I2C, C2I, and prediction losses, achieving superior performance on real-world Shanghai and Senegal datasets and robustness to noise and data scarcity. This universal approach demonstrates strong practical value for urban planning and emergency response by enabling accurate, joint predictions of micro- and macro-mobility dynamics without being limited to a single data modality.

Abstract

Predicting human mobility is crucial for urban planning, traffic control, and emergency response. Mobility behaviors can be categorized into individual and collective, and these behaviors are recorded by diverse mobility data, such as individual trajectory and crowd flow. As different modalities of mobility data, individual trajectory and crowd flow have a close coupling relationship. Crowd flows originate from the bottom-up aggregation of individual trajectories, while the constraints imposed by crowd flows shape these individual trajectories. Existing mobility prediction methods are limited to single tasks due to modal gaps between individual trajectory and crowd flow. In this work, we aim to unify mobility prediction to break through the limitations of task-specific models. We propose a universal human mobility prediction model (named UniMob), which can be applied to both individual trajectory and crowd flow. UniMob leverages a multi-view mobility tokenizer that transforms both trajectory and flow data into spatiotemporal tokens, facilitating unified sequential modeling through a diffusion transformer architecture. To bridge the gap between the different characteristics of these two data modalities, we implement a novel bidirectional individual and collective alignment mechanism. This mechanism enables learning common spatiotemporal patterns from different mobility data, facilitating mutual enhancement of both trajectory and flow predictions. Extensive experiments on real-world datasets validate the superiority of our model over state-of-the-art baselines in trajectory and flow prediction. Especially in noisy and scarce data scenarios, our model achieves the highest performance improvement of more than 14% and 25% in MAPE and Accuracy@5.

Figures (9)

• Figure 1: The transition from single model to universal model.
• Figure 2: The overview architecture of UniMob, which consists of four modules: (1) Multi-view Mobility Tokenizer, (2) Bidirectional Individual-Collective Alignment, (3) Joint Noise Predictor, (3) Mobility Predictor.
• Figure 3: Four model variants based on whether parameters are shared and whether two types of mobility data are used during testing.
• Figure 4: Flow prediction with noisy data on Shanghai and Senegal datasets.
• Figure 5: Trajectory prediction with noisy data on Shanghai and Senegal datasets.