Trajectory Network modeling is an advanced approach at the intersection of spatial epidemiology, network science, and human mobility analysis. It focuses on epidemic spreading driven by explicit individual movement pathsβsuch as pedestrian trajectories, commuting routes, or vehicular flowsβrather than assuming static contacts or homogeneous mixing. By transforming raw trajectory data into dynamic networks, this framework captures how repeated co-presence, crowding, and route overlap shape transmission risk. Trajectory Networks are especially powerful for analyzing localized outbreaks and designing mobility-aware intervention strategies.
ββββββββββββββββββββββββββββββββββββββββββββ
𧱠Compartmental Structure and Flow
Trajectory Networks typically underpin Agent-Based Models, where individuals are represented as moving agents whose disease states evolve through contact events defined by their paths. Despite the microscopic resolution, the disease progression follows a classical compartmental structure.
Susceptible agents (Sα΅’)
Individuals who are not infected but are vulnerable to infection. As agents move along their trajectories, they may encounter infectious agents. Exposure depends on proximity, duration of contact, and local crowding along the path.
Infectious agents (Iα΅’)
Individuals capable of transmitting the pathogen while moving through the network. Their contribution to transmission depends on how frequently and how long they co-locate with susceptible agents along shared trajectories.
Recovered agents (Rα΅’)
Individuals removed from the transmission process due to immunity or recovery. They continue to move through the network but no longer contribute to infection spread.
The core flow is Sα΅’ β Iα΅’ β Rα΅’, where the transition from susceptible to infectious is driven by contact events generated by overlapping trajectories in space and time.
ββββββββββββββββββββββββββββββββββββββββββββ
π Mathematical Formulation (Trajectory-Network-Adapted ODE System)
Although trajectory networks are often simulated using stochastic, rule-based algorithms, their expected dynamics can be approximated by a node-level compartmental system. For an agent or location i, the governing equations are:
dSα΅’/dt = β Ξ»α΅’ Sα΅’ β ΞΌ Sα΅’
dIα΅’/dt = Ξ»α΅’ Sα΅’ β Ξ³ Iα΅’ β ΞΌ Iα΅’
dRα΅’/dt = Ξ³ Iα΅’ β ΞΌ Rα΅’
Here, ΞΌ represents the natural mortality rate and Ξ³ the recovery rate. The key innovation lies in the force of infection Ξ»α΅’, which is explicitly determined by the trajectory network.
The force of infection for agent i is defined by the probability of acquiring infection from infectious agents encountered along its path:
Ξ»α΅’ = 1 β (1 β Ο)α΅
where Ο is the probability of transmission per contact, and k is the number of infectious contacts encountered by agent i during its trajectory over a given time interval. This formulation highlights how microscopic movement patterns directly determine macroscopic infection rates.
ββββββββββββββββββββββββββββββββββββββββββββ
π’ Table 1. Parameter Definitions
| Parameter | Definition |
|---|---|
| Ξ | Recruitment or birth rate of agents |
| ΞΌ | Natural mortality rate |
| Ξ³ | Recovery or removal rate |
| Ο | Probability of transmission per contact |
| k | Number of infectious contacts encountered along a trajectory |
| Ξ»α΅’ | Force of infection for agent or location i |
ββββββββββββββββββββββββββββββββββββββββββββ
π Table 2. Typical Parameter Ranges (General Viral Disease Context)
| Parameter | Typical Range | Interpretation |
|---|---|---|
| ΞΌ | 0.0002 β 0.002 per day | Low natural mortality |
| Ξ³ | 0.05 β 0.10 per day | Infectious period of approximately 10β20 days |
| Ο | 0.01 β 0.05 | Low to moderate transmission probability per contact |
| k | Highly variable | Depends on crowd density, route overlap, and mobility intensity |
ββββββββββββββββββββββββββββββββββββββββββββ
π Applicability and Limitations
Applicability
Trajectory Network models are best suited for epidemics driven by fine-scale human movement, such as respiratory viral diseases spreading in urban environments, transportation hubs, campuses, or large facilities. They are particularly valuable for evaluating localized non-pharmaceutical interventions, including crowd control, route redesign, staggered schedules, and mobility restrictions.
Model Strengths
By explicitly encoding individual trajectories, these models capture heterogeneous contact patterns, time-varying networks, and stochastic effects that are invisible to aggregate compartmental models. The evolving network structure naturally represents hosts moving in space and forming transient contacts, enabling realistic simulation of localized transmission hotspots.
Key Limitations
Trajectory-based models are computationally intensive, with costs scaling rapidly as population size and trajectory resolution increase. They are also highly data-dependent, requiring detailed, high-resolution mobility data that may be difficult to obtain or anonymize. Parameter estimation and model interpretation can therefore be challenging, particularly for large-scale populations.
ββββββββββββββββββββββββββββββββββββββββββββ
π Selected References
Pechlivanoglou, T., Li, J., Sun, J., Heidari, F., and Papagelis, M. Epidemic spreading in trajectory networks.
Zino, L., and Cao, M. Analysis, prediction, and control of epidemics: A survey from scalar to dynamic network models.
Chen, K., Jiang, X., Li, Y., and Zhou, R. A stochastic agent-based model to evaluate COVID-19 transmission influenced by human mobility.
Riley, S. Large-scale spatial transmission models of infectious disease.
Macdonald, G. The epidemiology and control of malaria.