Content area
Abstract
The most recent statistics in the United States are showing an increasing rate of construction worker fatalities due to vehicles crashing into roadway work zones, peaking at over 900 deaths in 2021. Research in transportation safety has mainly investigated these accidents in relation to the behavior and awareness of vehicle motorists, leaving the role of roadway workers’ behavior not well understood. Meanwhile, recent commercial roadway work zone intrusion alert (WZIA) systems have been deployed to raise alarms (e.g., sounds, flashing lights) to warn workers of traffic hazards (e.g., speeding car) and prevent accidents, but suffer from slow worker response times the farther away they stand from a WZIA’s stationary alarm source (e.g., loudspeaker/lamp). Wearable warning devices (e.g., smartwatch) have the potential to address the limitations of current WZIA systems by ensuring workers always perceive alarms on their bodies. As wearable warning devices become adopted in roadway work zones, workers may still experience alarm fatigue due to repeated exposure to alarms of constant attribute-values, reducing their responsiveness towards traffic hazards over time.
To address these challenges, prior research in the author’s research team initiated an integrated traffic and virtual reality (VR) platform that simulates realistic roadway work zones with surrounding traffic flow. Utilizing the platform to collect data on human roadway workers’ behaviors and alarm reactions, the major contributions of this dissertation include: (1) a deep learning model that can accurately predict roadway workers’ trajectory based on their construction activities and proximity to traffic vehicles, (2) a identified list of alarm attributes (e.g., modality) and values (e.g., “haptics and sounds”) that informs the design of future wearable warning devices for roadway worker safety applications, and (3) a reinforcement learning-based intelligent control system for fine-tuning alarm attribute-values to minimize alarm fatigue in worker reactions.