Enhancing AV Traffic Safety through Pedestrian Detection, Classification and Communication

Chase, R. Thomas; Feng, Jingyu; Hollar, Seth; Karimoddini, Ali; Wright, Waugh · 2022 · ROSA P / North Carolina. Department of Transportation. Research and Analysis Group

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 research addresses the safety challenges associated with Connected and Autonomous Vehicles (CAVs), specifically focusing on the breakdown of traditional two-way communication between pedestrians and vehicles. Traditional interactions rely on non-verbal cues like eye contact, which are absent in autonomous systems. The study aims to enhance AV traffic safety by improving pedestrian detection capabilities, particularly in occluded scenarios, and developing effective external communication methods to convey vehicle intent to pedestrians. The project employed a three-pronged methodology involving vehicle development, algorithmic testing, and human factors surveys. First, the researchers expanded the EcoPRT prototype autonomous shuttle system, integrating advanced pedestrian detection systems and external lightbar communication features. Second, they developed and tested multiple pedestrian detection methods using traditional datasets and a newly created dataset featuring occluded pedestrians. This included a body part-based detection method identifying heads, arms, and legs to handle partial occlusions, as well as sensor fusion techniques combining LiDAR and camera data. Third, a survey was conducted to evaluate how well pedestrians interpret static and dynamic lightbar patterns intended to signal CAV intent. The results demonstrated significant improvements in pedestrian detection accuracy and reduced latency. The improved detection method was successfully incorporated into the EcoPRT vehicle and showed potential for application in infrastructure-based detection systems. The body part-based method effectively improved detection rates for partially occluded pedestrians, addressing a critical challenge in cluttered environments. Additionally, the project produced a database of occluded pedestrian images for future training. Regarding communication, survey respondents struggled to correctly identify messages conveyed by lightbars in complex environments where multiple vehicle movements were possible, such as intersections with turning options. However, interpretation accuracy improved significantly in more constrained environments with fewer potential movement vectors. The significance of this work lies in its contribution to the standardization of AV-pedestrian interaction. By providing open data and algorithms, the project supports transparent policy-making and regulatory development, contrasting with proprietary private sector approaches. The findings highlight that while technical detection challenges can be mitigated through sensor fusion and specific algorithmic adjustments, effective communication requires context-aware signaling. The study suggests that lightbar designs must account for environmental complexity to avoid pedestrian confusion, emphasizing the need for standardized communication protocols to ensure safety in mixed-mode transportation settings.

Key finding

Pedestrians struggled to correctly identify lightbar messages in complex scenarios with potential turning movements, although identification improved in more constrained environments.

Methodology

mixed_methods

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.

• ehmi external hmi
• vru facing ehmi
• vru conspicuity