Applying AI to Data Sources to Improve Driver-Pedestrian Interactions at Intersections
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 report addresses the critical issue of pedestrian safety at intersections, which account for 40% of transportation crashes in the US. Motivated by an 80% increase in pedestrian fatalities between 2009 and 2021, the study aims to improve driver-pedestrian interactions using Artificial Intelligence (AI) and data analytics. The research is structured into three components: optimizing traffic signals via reinforcement learning, identifying rare "corner case" crashes using unsupervised machine learning, and analyzing determinants of nighttime crash severity. The first component employs a decentralized Dyna Q-Learning algorithm within a SUMO simulation of the Shallowford Road corridor in Chattanooga, Tennessee. The AI agent manages traffic signals by balancing vehicular flow and pedestrian safety, using delay metrics as rewards. Pedestrian data was approximated based on Highway Capacity Manual standards due to a lack of historical crossing data. The second component analyzes 1,000 fatal pedestrian-vehicle crashes from the 2020 Fatality Analysis Reporting System (FARS). It utilizes text mining and K-means clustering to identify extreme "corner cases" by isolating clusters with high weights of critical variables. The third component investigates nighttime crash injury severity using Random Forest algorithms and ordered logit models to determine contributing factors. Results from the traffic signal optimization indicate that the AI agent can reduce total vehicular delay by approximately 24% while ensuring pedestrian safety, though this comes at the cost of a 21% increase in average pedestrian waiting time. The system proved capable of prioritizing pedestrian service even when assigned lower priority than vehicles, demonstrating a tunable balance between safety and efficiency. In the crash analysis, the study identified nine specific "corner cases" (1% of the dataset) characterized by a combination of poor visibility, severe weather, dark lighting, and impaired behaviors. Specifically, 89% of these corner cases involved intoxicated pedestrians who failed to obey traffic signals, and 56% involved drivers under the influence. The nighttime severity analysis revealed that alcohol impairment, foggy weather, elderly pedestrians, speed limits of 50–55 mph, and motorists failing to yield are significant contributors to severe injuries. The findings suggest that AI-driven traffic control systems can effectively integrate pedestrian safety into optimization models, offering a scalable solution for urban intersections. Furthermore, identifying specific corner cases provides actionable insights for Automated Vehicle (AV) developers and road safety practitioners to design infrastructure and technologies that mitigate risks in extreme scenarios. The study underscores the necessity of addressing visibility and impairment factors to achieve the "Vision Zero" goal of eliminating traffic fatalities.
Key finding
AI-driven traffic signal optimization can significantly reduce vehicular delays while maintaining pedestrian safety, and rare fatal pedestrian crashes are predominantly triggered by combinations of environmental and behavioral factors such as darkness, impairment, and rule violations.
Methodology
mixed_methods
Sample size: 1000
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).
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | rosap | — | — | 2 | 2026-05-23 |
| archive | success | — | — | — | 1 | 2026-05-23 |
| extract | success | cached | — | — | 2 | 2026-06-10 |
| clean | success | — | — | — | 1 | 2026-06-01 |
| chunk | success | — | — | — | 1 | 2026-06-01 |
| embed | success | — | — | — | 1 | 2026-06-02 |
| enrich | success | — | — | — | 1 | 2026-05-23 |
| promote | success | — | — | — | 1 | 2026-05-23 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 3 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 19 | 2026-06-11 |
| verify | success | — | — | — | 2 | 2026-06-10 |
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.
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: crash risk outcomes