| Summary: | In an increasingly mobile world, traffic safety poses stark realities. In 2022, roadway incidents in the U.S. claimed over 38,824 lives, surpassing the fatality rate of the COVID-19 pandemic within the country. Despite these alarming statistics, understanding the overarching patterns of traffic safety presents a complex challenge due to myriad influencing factors such as driver behavior, roadway network geometry, weather conditions, and vehicle design.
This study revisits traffic safety from the perspective of statistical physics, positing universal temporal and memory effects to delve into the internal structure of traffic by exploring higher-order statistics. The examination of the internal structure enables the uncovering of near-miss incident risks in congested traffic flow—risks positively correlated with collision risks derived from historic accident records. By integrating the complex dynamics of traffic flow, the near-miss risk is ascertained from the crowdsourced velocity measurements of vehicles, thereby offering a computationally efficient framework with potential for real-time implementation.
We apply this framework to extensive velocity datasets collected anonymously across multiple states in the U.S., enabling the derivation of the spatial distribution of expected near-miss risk on a large scale. Moreover, we assess and compare the reliability and robustness of these networks, merging graph theory with our physics-inspired near-miss risk approach. Our findings consistently reveal patterns across different states, facilitating the identification of the most and least reliable/robust networks. This framework lays the foundation for a real-time, proactive maintenance of roadway networks, a major stride towards creating a safer transportation infrastructure.
|
|---|