Multi-agent Simulation Model Using Multi-layer Human Mobility Quantification for Prediction of Epidemics

Epidemic modelling is a critical field of study for understanding and mitigating the spread of infectious diseases, especially in complex environments. Traditional approaches often struggle to reconcile the high resolution of human mobility with the computational demands of large-scale simulation. Specifically, there is a lack of integrated frameworks that simultaneously handle high-fidelity human dynamics and social structures across different scales and dimensions. This thesis introduces a multi-scale modelling framework designed to bridge this gap, enhancing the predictive precision and operational utility of epidemic simulations.

The research delivers three primary contributions. First, a high performance pedestrian model towards epidemic simulation, which enables fine-grain strategies in dense urban environment without sacrificing computational scalability, help determine the limits of NPI during small-scale transmission. Second, the development of two regional epidemic simulation models to systematically trace infection routes across expansive geographic regions. By uncovering potential high-risk transmission nodes, pathways and super-spreaders, these offer actionable insights into region-specific epidemic dynamics. Finally, the thesis proposes an integrated approach to incorporate social network into agent-based models. This captures the impact of human interaction beyond simple physical proximity, examining community-driven transmission and the corresponding strategy from a different perspective.

Collectively, these frameworks demonstrate how integrating human mobility can overcome the limitations of traditional epidemic simulation approaches. The findings reveal that the spread of the disease is not merely a product of proximity, but a complex interplay of social mobility and connectivity. The collective output of this thesis provides a robust, actionable method for designing targeted interventions and improving rapid-response strategies.