Seminar: Leveraging AI to Improve Safety for Personal Mobility Device Users

Presented by Prof. Reuben Tamakloe, Research Assistant Professor, KAIST

Abstract: Personal Mobility Devices (PMDs) have seen a remarkable rise in popularity, becoming a preferred mode of urban transportation. However, this surge has brought significant safety concerns, highlighted by increased PMD-involved crashes. Research shows that PMD user behavior and the physical environment, especially in urban areas, play crucial roles in these crashes. This emphasizes the need for a comprehensive investigation into the factors associated with injury outcomes in PMD-involved accidents. Notably, there is a research gap in analyzing the determinants behind fatal or severe PMD crashes, particularly those where the PMD rider is at fault. Additionally, it is unclear how micro-scale streetscape features influence the fatality of these crashes, making it difficult for safety researchers and urban planners to efficiently address the road safety problems associated with PMD users.

With advancements in computing power, safety researchers can now utilize state-of-the-art machine learning tools to understand the factors contributing to fatal/severe crashes and develop solutions to reduce their occurrence and severity. In this study, we apply unsupervised machine learning tools and computer vision to explore the factors associated with fatal or serious injuries in homogeneous groups of crashes caused by PMD riders. Furthermore, we use computer vision and automated machine learning tools to investigate the impact of streetscape quality features namely greenness, openness, enclosure, walkability, and imageability—developed based on micro-scale built environment features such as sidewalk proportion, vegetation proportion, and building/wall proportion—on the injury outcomes of PMD-involved crashes.

Our analysis revealed intriguing insights, indicating unique factor associations in homogeneous PMD-involved crashes connected with fatal/severe injury. Additionally, the study identified a strong link between these streetscape features and injury outcomes. From the non-linear factor impact analysis, opportunities to enhance PMD rider safety by improving the built environment were identified. Policymakers can apply these findings to improve PMD rider safety, contributing to greater environmental sustainability.