Abstract
Micro-mobility usage has exploded over the last decade. Based on data from the National Association of City Transportation Officials (NACTO), between 2010-2022, there have been over 730 million trips on shared bikes and e-scooters in the United States and Canada alone. Micro-mobility options appeal to riders because they provide mobility that is flexible, sustainable, and cost-effective. However, as micro-mobility usage continues to soar, so do concerns regarding rider safety. Between 2017 and 2021, injuries connected to micro mobility vehicles spiked 127% to 77,200. Within this time period, e-scooter users experienced the greatest increase in injuries and fatalities. Unfortunately, due to the novelty of these transportation platforms, specifically e-scooters, there is a lack of policy and infrastructure governing their existence in our transportation systems, aiding the aforementioned safety issue. The focus of the research and analysis included in this thesis is to introduce and validate a novel framework to evaluate the impact of roadway features and conditions on e-scooter riders' behavior. The work herein will set the groundwork for future studies to inform and promote e-scooter safety. This novel framework integrates advanced computer vision and human sensing techniques to identify where objects and conditions on the road may impact e-scooterists' behavioral responses. Specifically, this research demonstrates the merits of the proposed framework through conducting preliminary analysis on how certain road situations, such as passing a pedestrian, passing a bus, encountering an occupied crosswalk, and other common situations impact e-scooterist’s behaviors using established metrics such as gaze entropy, variability, and percentage of road center fixations. In summary of our analysis, which included 10 hours of e-scooter data, we find that the situation wherein a rider switches from the "bike lane to [the] crosswalk", which commonly occurs at intersections, the gaze transition entropy and gaze variability in their eye-tracking data are the highest among all situations at 28.68 bits and 209.96 pixels respectively. This is most likely due to the difficulty of switching roadway infrastructures and navigating an intersection at the same time. We also find that the situation "road fixture," which includes the navigation of speed bumps/tables, manhole covers, and potholes, has the highest percentage of road center fixations at 76%. This can be attributed to the riders need to focus on the obstacle positioned within their path of travel to safely navigate it.
