Smart intersection monitoring systems face critical limitations when relying on single-sensor modalities, with cameras degrading under poor visibility and radar struggling with sparse point clouds. This work presents a novel multi-modal sensor fusion framework that integrates radar and camera detections through adaptive range-dependent strategies for robust vehicle detection across diverse operational conditions. We develop a radar detection pipeline that leverages drone-supervised learning to transform sparse radar points into precise 2D bounding boxes, achieving 93.0% precision through feature consistency filtering and motion-aware association. A geometry-aware camera-to-bird’s-eye-view transformation architecture combines homography priors with residual neural networks to enable direct spatial comparison between modalities, achieving 0.494 average IoU with lane-aware geometric correction. The fusion framework implements four distinct combination strategies—weighted IoU, range-adaptive, envelope, and intersection fusion—evaluated against 19,612 ground truth bounding boxes across six range bins. Results demonstrate universal fusion superiority with IoU improvements ranging from 6.1% to 28.3% across all operational ranges, achieving 30.4% overall ground truth win rate and 22.6% average IoU enhancement. Range-based analysis reveals distinct operational zones: camera dominance at close range (0-40m), fusion optimization in transition zones (40-80m), and progressive radar emergence through fusion support (80-100m), leading to radar effectiveness beyond 100m. The range-adaptive fusion strategy achieves peak IoU of 0.708 at 60–80m, highlighting intelligent sensor integration as key for intersection safety and guiding deployment strategies for next-generation transportation systems.

Vulnerable road users (VRUs) such as pedestrians and cyclists frequently engage in distracted behaviors (e.g., phone or headphone use) that elevate crash risk. General-purpose vision–language models (VLMs) struggle to capture these subtle, small, and context-dependent cues, and most camera pipelines still rely on frame-based analyses that ignore temporal context. We present IntersectionPAR, a lightweight vision–language framework that couples an Interactive Attribute Encoder with region-level captioning to recover safety-relevant semantic attributes (gender, phone/headphone use, mobility mode and aids) together with kinematic cues in a bird’s-eye view. We further introduce a temporal risk module that computes a continuous Situational Awareness Score (from phone, pose, and gaze proxies) and a Dynamic Risk Score that fuses awareness with spatial context (safe vs. unsafe zones), aligning system output with the specific moments practitioners need to act. Our dataset contains 12,273 annotated images drawn from~200 intersection-crossing scenarios under varied lighting and weather. Under a standardized zero-shot protocol against a broad suite of modern VLMs, IntersectionPAR attains keyword-based accuracy of 0.6123 with balanced precision/recall (0.7936/0.7864) and an F1-score of 0.7719 while requiring only ~9 GB of VRAM(>40 GB for several baselines). Qualitative analyses show strong behavior on safety-critical cues (e.g., cane and mobility-aid detection) and illustrate how the temporal risk module elevates risk precisely when low-awareness VRUs enter the roadway. Statistical analyses of 200 crossing events further reveal behavioral impacts of phone use (e.g., longer waiting time and increased path deviation). By pairing accurate, attribute-aware perception with low computational cost and an actionable temporal risk signal, IntersectionPAR supports real-time intersection safety monitoring and proactive interventions to mitigate VRU risk.

We introduce a real-time crash safety system capable of predicting crashes 5-10 minutes in advance, using advanced machine learning techniques. The system integrates primary, severity, and secondary crash predictions into a unified framework, employing a multi-layered prediction strategy that enhances proactive traffic safety management by providing robust and actionable alerts. It uses a comprehensive segment-specific approach, utilizing well-established machine learning models suited to the unique characteristics of different road segment types, such as basic, merge, diverge, and ramp. Additionally, route-specific models capture regional variations in traffic patterns and environmental factors, ensuring that the system adapts to the unique characteristics of different routes. Performance evaluations demonstrate that the system achieves high sensitivity (0.798-0.918) and low false alarm rates (0.088-0.208), depending on the segment type. The use of the continuity concept ensures the consistency of high crash likelihood situations, which improves the reliability of predictions. The results show that the system can accurately predict crashes and provide proper warnings. It can make traffic operators to take proactive actions such as adjusting traffic management measures. By predicting both the occurrence and severity of crashes, as well as the likelihood of secondary crashes, the system enhances the overall efficiency and effectiveness of traffic crash management. This comprehensive approach to crash prediction represents a significant advancement in proactive traffic management and road safety enhancement.

Signalized intersections are a critical component of road safety, yet rear-end crashes due to the yellow and red clearance phases remain a persistent problem. Traditional methods for setting these phases rely on kinematic models and video-based trajectory analysis, which suffer from high costs, weather/lighting limitations, and an inability to isolate signal timing impacts from confounding factors. This study introduces a novel framework for optimizing yellow and red clearance durations using Automated Traffic Signal Performance Measures (ATSPM). Analyzing over 446,000 traffic signal cycles across multiple intersections, the study examines rear-end conflicts during these phases and estimates the impact of signal timing adjustments using a causal forest. The results show that even small changes in yellow or red clearance durations can significantly affect rear-end conflict rates, with variations depending on intersection characteristics, traffic direction, and time of day. Unlike current standards, the proposed framework identifies context-specific optima, demonstrating how adaptive signal timing strategies can enhance safety without compromising traffic flow. This research expands the application of ATSPM beyond mobility optimization, offering a scalable and data-driven framework to improve intersection safety.

Large-scale real-time crash prediction is critical to traffic operation and management, but very challenging, even for machine learning models because the observation data are not independent and identically distributed (non-IID) due to its spatial heterogeneity. Data discontinuity, computational inefficiency, and higher false alarm rates are some other challenges in real-world and real-time crash prediction. To address these issues, we propose a generic framework, which blends spatial ensemble learning and knowledge distillation. Spatial ensemble learning leverages individual segment-level modeling to alleviate the non-IID issue for more accurate crash prediction, while knowledge distillation reduces the model size and improves false alarm rate. We justify the effectiveness of the proposed method using real-world and real-time scenario and comparing the results with the up-to-date benchmark models. Our method successfully improves sensitivity and false alarm rate to 90.35% and 24.21%. With this more accurate prediction model, we analyze the prediction variability across this region. We found that: 1) across segments, false alarm rate exhibits more heterogenous pattern than sensitivity; 2) rear-end crashes are more accurately predicted compared to sideswipe/angle crashes; 3) urban segments show better prediction performance compared to rural segments. 4) developing models with desired accuracy requires special attention in higher traffic fluctuation segments. These observations are very useful to develop more accurate prediction model and traffic safety decision making. To the best of our knowledge, this is one of the pioneering studies to integrate spatial ensemble learning and knowledge distillation to predict large-scale real-time crashes and apply it to analyze crash prediction variability.

Identifying vehicle maneuvers in the context of Connected Vehicles (CV) system brings huge potentials to enhance traffic safety. However, this process requires various advanced sensors, which are either available for luxury vehicles or expensive to install. Differently, smartphone is a more feasible choice with high penetration rate and various built-in sensors. Among the existing studies of applying the smartphone to detect vehicle maneuvers, most treated the detection as a classification problem without considering the real-world application. For example, the smartphone was fixed. Too many descriptive features were generated from the sensor data. To alleviate these problems, this paper developed a vehicle maneuvers detection system using a common smartphone with GPS, gyroscope, accelerometer, and magnetometer sensors. We first released the constraints on the smartphone’s position through a coordination system reorientation method. Then, simply filtered sensor data were directly used. A stacked-LSTM model was built to detect the vehicle maneuvers considering the time-dependency of the sensor data. This paper compared the performance of the proposed system with previous studies and various machine learning methods, including LightGBM, K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Random Forest. Extensive experimental results indicated that the proposed system accurately detected different vehicle maneuvers with an average F1-score of 0.98, precision of 0.97, and recall of 0.98, which outperformed the counterparts. Moreover, the model can be easily transferred to different drivers and locations. The system is robust and suitable for the real-time application as it requires simple processing of smartphone sensor data.