Comparison of safety and kinematic patterns of automated vehicles turning left in interaction with oncoming manually driven vehicles

Junghans, Marek; Krauns, Florian; Sonka, Adrian; Böhm, Michael; Dotzauer, Mandy · 2021 · OpenAlex-citations

DOI: 10.5507/tots.2021.003

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates the safety and kinematic differences between automated vehicles (AVs) and manually driven vehicles (MVs) executing conditionally tolerable left turns in interaction with oncoming traffic. The research is motivated by the challenge of developing AVs that are both safe and efficient in complex urban environments, where current systems often exhibit overly conservative behavior. The authors aim to determine if an SAE Level 3 AV behaves differently than human drivers regarding speed, acceleration, gap acceptance, and surrogate safety measures, providing insights for algorithm optimization. The study utilized real-world data collected at the Application Platform Intelligent Mobility (AIM) Research Intersection in Braunschweig, Germany, between May and June 2019. Video-based sensors tracked trajectories of road users, which were anonymized and fused in real-time. The automated vehicle used was TEASY 3, equipped with laser scanners, radar, cameras, and V2X communication capabilities. The dataset comprised 27 left-turn maneuvers by the AV and 70 matched baseline maneuvers by MVs. Matching criteria ensured comparable traffic conditions, queue positions (first or second), and oncoming traffic lane usage. The analysis focused on kinematic factors (speed, acceleration) and situational factors (gap acceptance, position in queue). Statistical comparisons employed non-parametric tests (Mann-Whitney-U and Kruskal-Wallis-H) due to violations of normality and homoscedasticity assumptions. Key metrics included Post Encroachment Time (PET), accepted/non-accepted time gaps, and distance/time to the conflict point during acceleration and deceleration phases. The results indicate that the AV exhibited significantly more conservative driving behavior than MVs. The median PET for the AV was 6.35 seconds, compared to 1.73 seconds for MVs, indicating that the AV maintained a much larger safety margin from oncoming traffic. The AV never accepted a time gap to intersect oncoming traffic, whereas MVs accepted gaps with a median of 1.54 seconds. Regarding kinematics, the AV required significantly more time to reach the conflict point during braking maneuvers compared to MVs, though the distance covered before braking did not differ significantly. The AV also showed less variance in its kinematic responses, suggesting more consistent but cautious control. Queue position influenced PET values for both vehicle types, with second-position vehicles exhibiting higher PETs than first-position vehicles, but the disparity between AV and MV remained substantial across positions. The findings suggest that while the AV achieves a high level of safety with very conservative PET distributions, its behavior is markedly different from human drivers, particularly in gap acceptance and intersection traversal speed. This conservative approach, while safe, may impact traffic efficiency. The study concludes that these insights are valuable for tuning AV algorithms to balance safety with efficiency, potentially allowing AVs to adopt more human-like gap acceptance strategies without compromising safety, thereby improving overall traffic flow in urban intersections.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• situational awareness