Urban planning and public health management have historically relied on static datasets, such as census records or infrequent travel surveys. However, the proliferation of GPS-enabled devices, smart card systems, and mobile signaling has shifted the focus toward high-resolution, dynamic mobility data. By analyzing the "digital breadcrumbs" left by millions of individuals, researchers can now reconstruct city-wide movement patterns to optimize infrastructure, predict individual behavior, and mitigate environmental or health risks.
Reconstructing Movement from Noisy Signals
The raw data captured by urban sensors is rarely clean or complete. A central challenge in this field is transforming sparse or noisy observations into high-fidelity trajectories. For example, Meegahapola et al. demonstrate in Inferring Accurate Bus Trajectories from Noisy Estimated Arrival Time Records that vehicle paths can be reconstructed using only publicly available Estimated Time of Arrival (ETA) data. This approach democratizes urban analysis by bypassing the need for private, restricted fleet logs.
Beyond simple location tracking, researchers aim to identify the intent or mode of travel. Zong et al. utilize Support Vector Machines to distinguish between walking, cycling, and motorized transport in Identifying Travel Mode with GPS Data Using Support Vector Machines and Genetic Algorithm. Similarly, Moosavi et al. propose segmenting trajectories into "behavioral dissimilarity" signals to automatically discover driving patterns like lane changes without requiring pre-labeled training data, as detailed in Discovery of driving patterns by trajectory segmentation.
Predicting Future States and Social Interdependence
Once trajectories are reconstructed, the focus shifts to prediction. Early work by Froehlich in Route Prediction from Trip Observations showed that the inherent repetitiveness of human travel allows algorithms to match the initial segments of a trip to historical routes, predicting a driver's destination long before they arrive.
However, individual movement is not isolated. De Domenico et al. argue in Interdependence and predictability of human mobility and social interactions that social connections are a primary driver of mobility. By leveraging the patterns of correlated individuals, they significantly improve forecasting accuracy over models that only consider a single person's history. This social dimension is also explored in tourism; Demessance et al. use Hidden Markov Models to predict tourist sequences based on social media reviews in Hidden Markov Model to Predict Tourists Visited Places.
Urban Dynamics and Public Health
Large-scale mobility datasets provide a lens through which to view city-wide crises. During the COVID-19 pandemic, researchers applied network theory to these datasets to identify vulnerabilities. Deng et al. utilized SafeGraph GPS data in Percolation of temporal hierarchical mobility networks during COVID-19 to find critical links that, if severed, could isolate clusters and contain disease spread. In Delineating Urban High-Risk Zones of Disease Transmission: Applying Tensor Decomposition to Trajectory Big Data, Lu et al. applied tensor decomposition to mobile signaling data to identify high-risk zones based on spatiotemporal community overlaps.
Mobility data also reveals socioeconomic disparities that static data misses. Bashan et al. found that traditional residence-based pollution assessments underestimate exposure. In Mobility-driven estimate reveals elevated air pollution exposure and socioeconomic disparities beyond residence-based approaches in Boston, the authors show that individuals experience significantly higher pollution during their daily commutes, with these exposures falling disproportionately on specific demographic groups.
Balancing Utility and Privacy
As the granularity of mobility data increases, so do privacy concerns. The ability to identify a home address from telematics data, as explored by Grumiau et al. in Address Identification Using Telematics: An Algorithm to Identify Dwell Locations, highlights the sensitivity of this information.
To address this, researchers are developing privacy-preserving frameworks. Park et al. proposed FedGeo in FedGeo: Privacy-Preserving User Next Location Prediction with Federated Learning, which uses federated learning to train location prediction models on-device. This ensures that sensitive trajectory data never leaves the user's phone, maintaining the utility of urban mobility analysis while safeguarding individual privacy.
Key Foundational Work
Spatial analysis of dynamic movements of Vélo'v, Lyon's shared bicycle program (2009) — contributed 2 concept elements to this direction.
Route Prediction from Trip Observations (2008) — contributed 1 concept elements to this direction.