An Interactive Learning Framework for Understanding Infrastructure Health Monitoring and Leveraging Machine Learning for Safety Improvement

Abstract

Crash mitigation in modern transportation systems requires more than just reactive safety strategies. It demands a comprehensive understanding of how infrastructure health conditions directly influence crash occurrence and severity patterns. As road networks continue to deteriorate and traffic complexity increases, ensuring roadway safety increasingly depends on integrating advanced predictive modeling capabilities with real-time infrastructure health monitoring systems that can detect safety-critical conditions before they contribute to accidents. The first manuscript presents a machine learning framework to predict crash injury severity using real-world crash data and roadway characteristics. By applying models such as Artificial Neural Networks (ANN), Light Gradient Boosting Machine (LightGBM), Random Forest (RF), K-Nearest Neighbors (KNN), and Ordinal Logistic Regression (OLR), the study quantifies the potential injury reduction benefits of infrastructure improvements. These data-driven predictions are validated against established Crash Modification Factors (CMFs), offering a practical method to prioritize safety investments based on injury prevention impact. Recognizing that infrastructure condition directly influences safety outcomes, the second manuscript introduces a vehicle-infrastructure-integrated digital twin platform that enables real-time monitoring of structural health (e.g., strain and vibration). This educational system combines physical sensors, Raspberry Pi microcontrollers, and a live dashboard to simulate intelligent infrastructure capable of interacting with traffic environments. The platform not only supports hands-on learning but also highlights the critical role of continuous infrastructure monitoring in proactive crash mitigation.