Fleeting Decisions and Risks in Pedestrian Road-Crossing Behavior: Building Insight with Next-Generation Data, Models, and Platforms

New York University, New York NY

Investigators

Abstract

In this project, the Principal Investigator will examine road-crossing behavior over very small moments of space and time, by observing and studying real-world road-crossing sites. The conditions observed will then be recreated in a motion capture studio that will allow collection of very detailed data that can be used to build what-if models for training, planning, and design. The decision to cross a road can take mere fractions of a second and centimeters of action, but over this small window of space and time, people are at risk of injury and death in collisions with vehicles. Despite decades of work to design interventions that might promote safe crossing behavior, many people continue to engage in unsafe crossing, among them our most vulnerable - pedestrian youth and seniors. The aim of the project is to uncover the, often cursory, decisions and risks that people invoke when they move at the roadside. The focus on the microcosm of road-crossing will enable identification and measurement of subtle signals from body language, expression, social cues in crowds, gaze behavior, and footfall that can convey both covert and overt insight into crossing intent. These signals and the road-crossing behaviors that they relate to will inform actionable understanding of road-crossing activities that can be used to improve the design of crossing infrastructure, that can be used in training and education, that might enhance signaling schemes, and that can be used to inform Advanced Driver Awareness Systems now being deployed in many vehicles. The project make use of a coupled research instrument that allows for coded real-world observation to be paired with advanced sensor-based instrumentation in motion capture studios, and with what-if modeling based on human automata that can facilitate the exploration, identification, and measurement of subtle signals of pedestrian decision-making and risk-taking while crossing. The technical focus of the project is to identify, capture, measure, and frame human-environment signals of road-crossing behavior in ways that can be used for broader pedestrian sensing and detection. The project will generate a robust library of data, including qualitative observations, sensor data, and model output that can be reused by communities in the socio-behavioral sciences and the computing sciences. These data, and the models they will power, will form the backbone of an immersive Virtual Geographic Environment that can be used for training and education.