Johns Hopkins develops AI tool to predict traffic accident rates

Johns Hopkins University researchers have developed SafeTraffic Copilot, an AI tool that uses large language models to predict how traffic infrastructure changes will affect accident rates at specific intersections. The system represents a significant advancement in traffic safety technology, offering policymakers data-driven insights to reduce crashes in an era when Maryland highway fatalities have risen from 466 in 2013 to 621 in 2023.

How it works: SafeTraffic Copilot was trained on descriptions of more than 66,000 accidents, incorporating road conditions, blood alcohol levels, satellite images, and on-site photography to make predictions.

• The AI can evaluate both individual and combined risk factors, providing detailed understanding of how various elements influence safety.
• Unlike traditional machine learning models, the generative AI offers “what-if” capabilities, allowing users to predict outcomes from specific changes like adjusting traffic light timing from 20 to 30 seconds.
• The system provides confidence scores with its predictions, addressing the “black box” problem that has previously deterred AI use in high-risk traffic safety applications.

Key findings: The model reveals that alcohol and aggressive driving are the most dangerous factors, contributing to three times more crashes than other causes.

What the researchers are saying: “These are complex events affected by numerous variables, like weather, traffic patterns, roadway design and driver behavior,” said senior author Hao “Frank” Yang, an assistant professor of civil and systems engineering at Johns Hopkins.

• “With SafeTraffic Copilot, our goal is to simplify this complexity and provide infrastructure designers and policymakers with data-based insights to mitigate crashes.”
• “Generative AI has a big potential to improve the trustworthiness of accident prediction,” Yang continued.

The big picture: Maryland currently uses traditional machine learning for traffic safety assessment, but Yang’s approach offers superior adaptability and predictive capabilities.

• “With machine learning, if a sample is not similar to your training samples, you cannot generate a prediction. Generative AI can give clear ‘what-if’ capabilities,” Yang explained.
• The system can be customized for different states or cities and adapted to international traffic conditions simply by providing descriptive paragraphs about local differences.

Global applications: Yang envisions expanding the research internationally, particularly to countries with different traffic patterns and cultural driving behaviors.

• “In South Asian countries like Taiwan or the Philippines, most crashes are motorcycle crashes, and the way they drive is different as well,” he said.
• “With previous models, incorporating drivers’ behavior or cultures is impossible.”

Why this matters: With 381 people killed on Maryland highways so far this year and fatalities steadily rising over the past decade, SafeTraffic Copilot could provide crucial tools for reducing traffic deaths through evidence-based infrastructure decisions.