Baseline Analysis of Driver Performance at Intersections for the Left-Turn Assist and Intersection Movement Assist Applications

Stevens, Scott; Lam, Andy H.; Bellone, Jeffrey; Azeredo, Philip; Mui, Emily; Guglielmi, John; Medri, Marisol · 2020 · ROSA P / United States. Department of Transportation. National Highway Traffic Safety Administration

archive: archived pipeline: cataloged verified

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

Summary

This study addresses the need for improved design of Left-Turn Assist (LTA) and Intersection Movement Assist (IMA) applications, which warn drivers of potential collisions when crossing intersection paths. The primary motivation is to reduce false and nuisance alerts by establishing a robust understanding of baseline driver behavior. By characterizing normal driving metrics, developers can better distinguish between safe maneuvers requiring no alert and hazardous situations requiring intervention. The research focuses on six specific intersection scenarios, including left turns across opposite or lateral traffic, right turns into path, and straight crossings. The methodology comprised three components: a literature review, an analysis of naturalistic driving data, and an analysis of crash data. Researchers queried two naturalistic driving databases—the Safety Pilot Model Deployment in Ann Arbor, Michigan, and the Driver Adaptation study in Washington, DC—to identify 772 baseline events involving 107 drivers. These events were validated through manual video review. To provide a counterpoint for collision scenarios, the study analyzed 194 severe crashes from the Crashworthiness Data System (CDS) where both vehicles had event data recorders. The primary metric analyzed was "gap length," defined as the time from when a driver began crossing until an oncoming vehicle reached the conflict point. Additional metrics included speed, acceleration, steering angle, and factors such as age, gender, distraction, and intersection geometry. Key findings revealed distinct differences between baseline and crash scenarios. In baseline driving, the average accepted gap length was 3.6 seconds, with over 94% of accepted gaps exceeding the average rejected gap of 3.4 seconds. In contrast, crashes occurred with significantly shorter adjusted gap lengths, averaging 2.5 seconds. Gap acceptance varied by scenario, with Straight Crossing Paths–Left (SCP-L) yielding the largest baseline gaps (7.1 seconds) and Left Turn Across Path–Opposite Direction (LTAP-OD) the smallest (5.0 seconds). Gender analysis indicated a medium effect, with men accepting smaller gaps (5.4 seconds) than women (6.5 seconds), though no gender effect was found in crash data. Age, lighting, and weather showed no significant effects in baseline data, likely due to sample size limitations. Drivers accelerated from stopped positions at similar rates, reaching speeds between 11.5 and 31.2 mph depending on the scenario. The significance of this work lies in providing empirical data to refine LTA and IMA algorithms. By establishing baseline metrics for gap acceptance and vehicle dynamics, the study offers developers specific thresholds to minimize nuisance alerts, such as those triggered when drivers inch forward for visibility. The findings highlight that intersection geometry and scenario type significantly influence driver behavior, suggesting that warning systems must be context-aware. This baseline analysis supports the creation of more effective, user-accepted safety technologies by grounding alert logic in real-world driving patterns rather than theoretical assumptions.

Key finding

Drivers accepted significantly larger gaps on average (3.6 seconds) during normal baseline driving compared to the adjusted gap lengths observed in crash scenarios (2.5 seconds).

Methodology

naturalistic

Sample size: 966

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. Discovered via bulk_ingest_rosap on 2026-05-23 (6 acquisition events logged).

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

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

• gap acceptance
• lane changing
• signaling behavior
• looked but failed to see
• temporal
• rail grade crossings

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: behavioral performance data, crash risk outcomes
• Methodological Resource: dataset resource