Citation

Wang S, Ma Y, Liu J, Yu B, Zhu F. Accid. Anal. Prev. 2022; 174: e106762.

Copyright

(Copyright © 2022, Elsevier Publishing)

DOI

10.1016/j.aap.2022.106762

PMID

35792477

Abstract

Emerging automated vehicle (AV) technology is being deployed on as-built roadways due to its promising safety improvements. However, realistic problems concerning whether and how perception sensor-based AVs can safely adapt to the existing roadway infrastructures remain to be well addressed due to a lack of consideration of the sensor's angular resolution and detection threshold. In this study, we aim to assess whether LiDAR-based AVs (LAVs) could safely adapt to as-built horizontal curved roads from the perspective of available sight distances (ASDs) through virtual simulations. In specific, i) numerous driving scenarios featuring the design speed (V(d): 40 ∼ 100 km/h), circular curve radius (R: limited minimum radius ∼ common minimum radius), LAV (with LiDAR technical parameters, e.g., number of channels, N(c): 32, 64, 128), and the front target vehicle were simulated in PreScan/MATLAB/Simulink co-simulation platform; ii) an ASD extraction algorithm was proposed considering the point threshold for detection (N(T)); iii) effects of V(d), R, N(c), and N(T) on the ASD were analyzed and polynomial models were adopted to capture relationships between the ASD, V(d), R at different N(c) and N(T); iv) the minimum speed against as-built sight obstructions along the roadside and the maximum speed against inadequate sight distance were proposed by comparing the ASD with the required stopping sight distance of human-driven vehicles and LAVs (level 3 ∼ 5), respectively; and v) speed limits (V(L)) against inadequate sight distances for level 3 ∼ 5 LAVs were proposed. The results indicate that: i) a larger R or V(d), fewer N(c), or a higher N(T) would cause a shorter ASD in general; ii) attention should be paid to the occlusion imposed by as-built roadside infrastructures even with more N(c) or/and a lower N(T), particularly to curved roads with more rigorous geometric design controls (e.g., small V(d)); and iii) level 3 LAVs struggle to adapt to as-built horizontal curved roads, and level 4 or 5 LAVs cannot assure adequate ASDs on high-type curved roads (e.g., large V(d)). These findings shall help road administrators make decisions on speed limits for LAVs on as-built curved roads.

Language: en

Keywords

Road safety; Automated vehicle; LiDAR; Available sight distance; Road geometry; Virtual simulation