Human mobility data underpins many location-based services and urban applications, but collecting real trajectories can be expensive or impractical. Synthetic mobility generation offers an alternative, producing artificial trajectories that mimic real patterns. However, the abstract notes that recent methods have focused primarily on trajectory statistics, potentially missing higher-level structures.
The new work introduces a framework that goes beyond these statistics by using evolving user-location graphs. This graph-based approach is designed to maintain community consistency, meaning the generated trajectories should reflect the same group-level movement patterns and social or spatial communities observed in real data.
Because the abstract is brief, the specific mechanisms and evaluation results are not detailed here. The paper's contribution appears to be in shifting the objective from purely statistical fidelity to also preserving community structure, which could make synthetic data more useful for urban planning and network analysis.