Mobility studies on humans, vehicles, and animals involve trajectory data, which allows for location tracking over time per entity. Trajectory data can help with analyzing and understanding the location information of an entity or a crowd, the flow of entities in a particular location or segment, and patterns of human activities (for example, walking or driving). However, these movement analytics are influenced by heterogeneous context data such as the underlying landscape (for example, the road network), the surrounding points of interest (for example, gas stations, drive-throughs, and so on), and the temporal context (for example, time of day, business hours, and so on). Meaningful insights from trajectory data are revealed when it is integrated with these rich sets of contextual data.
The authors identify a list of algorithms frequently used in human mobility analyses, and suggest the research problems that emphasize the role of the contextual data: (1) an outlier detection algorithm; (2) a context-aware interpolation algorithm; (3) the fine-grained segmentation of trajectory data with contextual data to build better mobility models; (4) a trajectory clustering algorithm for better differentiation and grouping of moving entities; (5) traffic prediction models and flow analysis techniques; (6) co-operative and competitive movement models; (7) anonymization algorithms and analyses to protect the risks of inference attacks; and (8) computing platforms to ingest, integrate, and analyze the diverse context data with the trajectory data, to facilitate reusability and high performance.
This position paper nicely summarizes the current human mobility research problems, as well as warns researchers that mobility-related algorithms and approaches require an integrated contextual algorithm library and computing platforms to gain more meaningful insights and intelligence. The paper lacks concrete examples, the current state of the art, a discussion of the challenges and barriers, and action plans for pursuing solutions to the described problems, such as whether the context-aware library and computing platforms will be designed by one entity or organically evolve.
Application developers working with mobility data in any vertical industry will find this paper valuable, as it presents possible models of incorporating contextual data for value generation.