Measures and Models of Mobility in Physical and Virtual Environments

点击翻译按钮获取中文摘要

Models of how people and animals move through space inform engineers designing cities and services, biologists studying animal behavior, and computer game designers attempting to build more believable worlds. Gathering the data for these models is commodified, with several companies offering technologies and services to capture movement data on target populations. These models are almost always based on spatial features - concise mathematical or algorithmic encapsulations of trajectory properties. Statistical models that differentiate populations based on movement properties are directly tied to features, which make up the independent variables under evaluation. Synthetic movement models use features algorithmically to determine when movement patterns should change, or to validate that the algorithms produce appropriate distributions for trajectory features. However, our library of movement features is incomplete, leading to incomplete models. For example, random walker models can create realistic distributions of trip length and dwell time, but produce paths which are erroneous upon inspection, leading to incorrect models of cell tower usage or the spread of disease. We will enable more realistic mobility models through the discovery, characterization and validation of novel features encapsulating mobility patterns for both real and virtual spaces, through the analysis of trajectories as strings and spatial histories as graphs. To facilitate and accelerate the development of new features we will also codify and deploy standard datasets and testing mechanisms to evaluate feature quality. Developing new measures can be a difficult process, as it requires both data analytic and domain expertise, supported by empirical data and mathematical rigor. My research group has extensive experience in analyzing digital games and measuring human spatial behavior, and has published and validated several spatial features over the last three years. Graduates from my research group are employed in data analytics in industry and academia. Through collaborators we have access to terabytes of mobility data. This proposal outlines a program of research which creates curated standardized datasets, and clear proscribed tests of efficacy and generalizability to drive innovation in features of spatial histories and trajectories. Leveraging innovations in graph and string-based analysis of trajectories, we will provide novel, proven, and reliable features to researchers and practitioners.