Improving Crash Frequency Prediction from Simulated Traffic Conflicts Using Machine Learning Based Microsimulation

2026-06-12 · Source: cs.AI updates on arXiv.org · Field: Transportation & Mobility — Artificial Intelligence & Machine Learning, Transportation Infrastructure, Data Science & Analytics · Depth: Expert, quick

Summary

A study investigated improving crash frequency prediction by integrating machine learning (ML)-based behavior models into traffic microsimulation. Researchers conducted simulations for five real-world signalised intersections in Leeds, UK, comparing a standard rule-based model with a state-of-the-art ML model. Vehicle trajectories were analyzed using a two-dimensional Time-to-Collision metric to identify simulated conflicts, which were then processed with Extreme Value Theory to predict crash frequency. The ML model's conflicts produced crash predictions consistent with actual crash data, demonstrating its effectiveness without requiring location-specific calibration. In contrast, the rule-based model failed to provide meaningful predictions. While ML models realistically reproduce conflicts, directly using ML-generated simulated crashes for prediction proved ineffective, suggesting a current limitation in generating realistic crash events.

Key takeaway

For traffic safety engineers evaluating new road infrastructure designs, you should integrate machine learning-based behavior models into your microsimulations. These models can significantly improve crash frequency predictions from simulated conflicts, aligning with real-world data without extensive location-specific calibration. However, avoid directly using ML-generated simulated crashes for prediction, as current models are not yet capable of producing realistic crash events. Focus on conflict analysis as a more reliable indicator.

Key insights

ML-based traffic microsimulation improves crash frequency prediction from simulated conflicts without location-specific calibration.

Principles

• ML models can learn human driving behavior from trajectory data.
• Realistic conflict dynamics are crucial for accurate crash prediction.
• Direct ML-generated crashes are not yet realistic for prediction.

Method

Traffic microsimulation uses a two-dimensional Time-to-Collision metric to identify conflicts from vehicle trajectories, then applies Extreme Value Theory to model these conflicts for crash frequency prediction.

In practice

• Use ML models for conflict generation in safety studies.
• Validate ML-simulated conflicts against real-world crash data.
• Avoid direct use of ML-generated crashes for prediction.

Topics

• Traffic Microsimulation
• Crash Frequency Prediction
• Machine Learning Models
• Surrogate Safety Measures
• Extreme Value Theory
• Time-to-Collision

Best for: AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• Learning collision risk proactively from naturalistic driving data at scale — GSSM proactively learns collision risk from naturalistic driving data, outperforming baselines in accuracy and timeliness.
• Calibrating Urban Traffic Simulation from Sparse Road Observations via Genetic Optimization — Urban traffic simulation calibration is achievable with minimal data by inferring job distributions via genetic optimization.
• Artificial Intelligence for Modeling and Simulation of Mixed Automated and Human Traffic — AI methods are crucial for accurately simulating complex mixed human and autonomous traffic behaviors.
• Learning of Robot Safety Policies via Adversarial Synthetic Scenarios — Adversarial gamification efficiently discovers robot safety edge cases by pitting a hazard-creating Red Team against a policy-refining Blue Team.
• Low-Cost System for Automatic Recognition of Driving Pattern in Assessing Interurban Mobility using Geo-Information — A low-cost system uses an ANN and geo-information to recognize driving patterns, improving safety and efficiency.
• Multi-Agent Reinforcement Learning for Safe Autonomous Driving Under Pedestrian Behavioral Uncertainty — Co-training SDCs with personality-driven MARL pedestrians significantly improves safety by exposing vehicles to realistic, uncertain human behaviors.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.