Highway security assessments have historically relied on police-reported accident statistics. This statistic is commonly thought-about the “gold commonplace” as a result of it instantly correlates to the variety of deaths, accidents, and property harm. Nevertheless, counting on historic crash information for predictive modeling poses important challenges. As a result of such information is actually a “lagging” indicator. Moreover, as a result of crashes are statistically uncommon occasions on highways and native roads, it could actually take years to build up sufficient information to ascertain a sound security profile for a selected street phase. This sparsity, mixed with inconsistent reporting requirements throughout areas, complicates the event of strong danger prediction fashions. Proactive security evaluation requires “main” measures. This can be a proxy for crash danger, which correlates with security outcomes however happens extra steadily than crashes.
“From Lag to Lead: Inspecting Onerous Braking Occasions as a Dense Indicator of Segmental Crash Danger” evaluates the effectiveness of arduous braking occasions (HBEs) as a scalable surrogate measure of crash danger. HBE is when the automobile’s ahead deceleration exceeds a sure threshold (-3m/s²), which is interpreted as an evasive maneuver. In contrast to proximity-based surrogates akin to crash time, which regularly require the usage of mounted sensors, HBE is sourced from related automobile information, facilitating network-wide evaluation. By combining public crash information from Virginia and California with anonymized and aggregated HBE data from the Android Auto platform, we established a statistically important constructive correlation between crash incidence (no matter severity stage) and HBE frequency.
