Open Collections

Abstract

Empirical quantification of how autonomous driving will affect road safety, particularly whether current road designs can accommodate autonomous vehicles (AV), remains under-researched. This research addresses the gap by proposing a three-dimensional (3D) risk assessment framework that integrates reliability theory and mobile Light Detection and Ranging (LiDAR) scans, focusing on how sight distance limitations interact with vehicle autonomy. Using data from 308 curves along a rural highway in British Columbia, Canada, the framework was applied in three phases. First, a voxel-based 3D LiDAR method was developed to estimate available sight distance (ASD) in complex terrain, with results compared against traditional two-dimensional (2D) methods. Second, three vehicle types were defined to represent different automation levels, including human-driven vehicles (HDV), transition-stage AVs, and fully developed AVs, followed by a reliability-based risk assessment comparing ASD with stopping sight distance (SSD) required by these vehicles. The resulting probability of non-compliance (Pnc) served as a quantitative measure of design risk from insufficient sight distance. Finally, sensitivity analyses were conducted to explore how operational parameters and sensor configurations influence highway design risk levels. The research found that the 3D method provided a more accurate, location-sensitive evaluation of ASD, while the 2D method often overestimated sight distance, especially on combined horizontal and vertical curves. The 3D-based risk assessment indicated an overall reduction in risk with increasing automation, although some cases showed higher risk for fully developed AVs. Segment- and curve-level analyses showed that fully developed AVs face higher risks on sharp curves with limited visibility due to assumptions of strict speed compliance and comfort-based braking rates. Sensitivity analysis showed that increasing deceleration rates can substantially reduce AV risks, while raising sensor height offers limited benefits. This integrated framework highlights the value of combining LiDAR technology and reliability theory for estimating Pnc as a 3D risk index. P𝚗c can guide manufacturers in adjusting operating parameters (e.g., speed and braking rates) or provide targeted system training at high-risk locations, while also helping road agencies prioritize design improvements to support road infrastructure readiness and safer transitions to full driving automation.