Movement similarity is a hot topic and the foundation of a plethora of methodologies in movement analysis. In this thesis, movement similarity is classified into two categories. The first category, explicit movement similarity, is defined as the closeness of trajectories in space and time. It is directly quantified along the path of individuals using a wide variety of trajectory similarity measures, such as Fréchet distance and dynamic time warping. On the other hand, the second category, implicit movement similarity, is defined as the consistency in movement and behavioral patterns of an individual or a group of moving entities. It can serve as a high-level representation of movement patterns of an individual or a group of moving entities and can be applied to solve many types of movement-related problems, such as trajectory prediction and interpolation. Surrounding the topic of movement similarity, this thesis investigates two methodologies based on the explicit and implicit movement similarity, respectively, to demonstrate their applications in movement analysis. Explicit movement similarity in this thesis is utilized to unravel the associations between migration paths and underlying environmental correlates influencing movement choices of migratory turkey vultures (Cathartes aura) in North America. Multiple commonly used trajectory-similarity measures including Fréchet distance, dynamic time warping (DTW), Hausdorff distance, longest common subsequence (LCSS), and edit distance are integrated into a hierarchical clustering approach to identify variations in turkey vultures’ migration path choices over multiple seasons. At each hierarchy, the optimal clustering setting, i.e., a distance metric together with the number of clusters, is selected automatically based on the silhouette coefficient. Using 15 years of tracking data of turkey vultures during their fall and spring migration seasons, seasonal clusters are identified and then annotated with environmental variables for Kolmogorov-Smirnov (KS) test and Jensen-Shannon distance (JSD) calculation to examine the variation between clusters and the background and variation between clusters. In terms of the application of implicit movement similarity, this thesis proposes a trajectory interpolation model with an encoder-decoder architecture based on gated recurrent units (GRUs) to interpolate trajectory gaps (missing values). The proposed model is able to read a trajectory containing a gap in both chronological and reverse chronological orders. The information obtained from these two directions is fused to learn the implicit movement similarity contained in that trajectory, which is later used to reconstruct the complete trajectory with the original gap filled in. The proposed interpolation method is validated using turkey vulture migration trajectories. Interpolation results demonstrate that the proposed method is capable of capturing implicit movement similarity from trajectories for interpolation purposes, since without which, some gaps would be difficult to interpolate accurately using traditional interpolation methods. Movement similarity is a promising field in computational movement analysis and is often used as a foundation of other machine learning and modeling methodologies such as trajectory classification, behavioral model detection, and movement prediction. On the foundation of explicit and implicit movement similarity, researchers can build multifarious models for diverse moving entities, such as animals, humans, vessels, and vehicles.

Background

Results

Conclusions

Human beings have a fundamental need for social interaction according to Maslow’s hierarchy of needs. Human interaction contributes to individual psychological development and well-being, and enhances social cohesion more broadly. Among others, two scientific fields that are interested in human interaction in space and time progressed in separate paths but emerged from similar inspirations and aim at similar end goals. The first is activity analysis with a focus on developing models of activity participation and travel behavior using travel surveys. These are commonly referred to as activity-based approaches in travel demand analysis. The second is the study of human movement in space and time using geometric relationships that evolve over time and space derived from the observations of moving individuals, a tradition followed mostly by geographic information scientists. Travel behavior researchers recognize human interaction as one of the most important determinants affecting people's daily time allocation and travel. However, inter-individual interactions, particularly inter-household ones, in travel behavior research have received less attention while these interactions contribute to a significant amount of travel. Traditional travel surveys are usually limited in sample size and observation period due to expensive costs. They also often lack sufficient information regarding inter-household interactions. Besides human interactions when individuals engage in activities at specific locations, it is also crucial to consider potential interactions in shared spaces, such as when their activity spaces or paths intersect during movement. In recent years, the increased availability of diverse and fine-grained movement data offers an unprecedented opportunity to study human movement and interaction at an unparalleled level of fine granularity. However, there is a shortage of advanced computational methods to trace and understand human interactions present in fine-grained and large-volume movement data. It also remains unclear to what extent the knowledge gained from individual movement data can contribute to our understanding of collective behaviors across different socio-geographical contexts. The primary goal of this dissertation is to develop and evaluate new computational methods for identifying and quantifying spatiotemporal patterns of human daily activities and interactions, using both conventional travel diaries and massive movement data of individuals. By doing so, this dissertation explores important questions to understand heterogeneity in people’s daily activity-travel patterns and interactions across different socio-geographical spaces.