TL;DR
This research establishes a statistically robust and deployable foundation for next-generation intelligent transportation systems by coupling Bayesian learning theory with edge computing design.
In recent years, vehicle and driver detection in real-world traffic scenarios has attracted increasing attention, with accurate driver identification being critical for traffic supervision and public safety. This paper proposes IYOLO, an enhanced YOLOv8-based framework for simultaneous detection and classification of vehicles, drivers, and passengers on highways, aiming to distinguish drivers from passengers and establish one-to-one vehicle-driver associations. The model leverages Cross Stage Partial to Fast (C2f) modules to reduce redundant computations and accelerate inference, and employs an optimized feature pyramid with multi-scale fusion to improve small-object detection. An adaptive Label Smoothing Regularization strategy enhances generalization and classification robustness, while Online Hard Sample Mining focuses learning on challenging samples during training, improving feature discrimination and overall performance under complex conditions. Extensive experiments on the PSD-HIGHROAD dataset demonstrate that IYOLO consistently outperforms state-of-the-art methods in detecting and classifying vehicles, drivers, and passengers, achieving superior accuracy and robustness across varying lighting, poses, and traffic conditions.
These findings demonstrate that RF–Bayesian provides a stable, interpretable, and computationally efficient framework for smartphone-based driver behavior classification, with practical relevance for telematics, fleet safety management, driver feedback systems, and intelligent transportation safety applications.
Advanced driver behavior analysis is a revolutionary strategy for improving driving behavior and preventing accidents. With the advancements of technology, it is now possible to capture real-time data on driving behavior, encompassing vehicle characteristics such as speed, acceleration, braking, negligent driving, and driver’s physiological parameters such as heart rate, drowsiness, etc. In this paper, we propose advanced data analytics to provide in-depth insight into drivers’ behaviors. One of the main objectives of this paper is to devise feature engineering techniques so that driving behavior can be determined in real time using an optimal number of features. This approach combines cutting-edge machine learning models with onboard sensors’ data to enhance vehicle safety, develop responsible driving behaviors, and eventually create a safer highway environment for all the stakeholders involved.
As a core interaction in the human-vehicle-road system, automated and refined detection of driving behavior has emerged as a crucial research direction in intelligent transportation systems and advanced driver assistance systems. Traditional post-event monitoring models that rely on manual or sensor-based methods are no longer able to meet the requirements of real-time and accurate risk identification. Therefore, this study proposes the You Only Look Once - Lightweight - BiFPN - ECA (YOLO-LBE) detection method. By integrating ghost convolution and GhostC2f modules to diminish computational complexity, the study employs a weighted bidirectional feature pyramid network, and further embed an ECA module to significantly enhance the precision and stability of driver behavior detection. Experimental findings demonstrate that the improved YOLOv8 model improves mAP@0.5 by 5.3%, FPS by 32.4%, Params by 34.4%, and FLOPs by 33.3% compared to YOLOv5s. This research method outperforms existing mainstream models in terms of accuracy, efficiency, and interference tolerance, providing reliable technical support for real-time driving behavior monitoring.
Runtime monitoring is essential for detecting impending hazards in autonomous driving systems (ADSs). However, existing ADS runtime monitors have fixed detection capabilities: rule-based monitors cover only manually specified hazards, while learning-based monitors depend heavily on their initial training data and may retain substantial prediction errors. We therefore propose EvoEye, which identifies the current monitor's errors, generates informative executions accordingly, and updates the monitor through self-evolution. To enable effective self-evolution, EvoEye combines a capable runtime monitor with targeted scenario acquisition. FusionMonitor learns cross-module temporal interactions for collision prediction, while BlindSpotEvolver converts current prediction errors into search guidance and uses density-aware mutation to acquire informative executions for subsequent monitor updates. We evaluate EvoEye on Baidu Apollo with CARLA in representative highway and urban scenarios. FusionMonitor improves frame-level Recall by up to 37.8 percentage points at a false positive rate of 0.05, with 2.49 ms latency and 2.8-4.2 seconds of median warning time. Under the same budget, BlindSpotEvolver outperforms uniform and violation-oriented sampling by up to 13.2 F1 points on previously missed unsafe contexts.
: Transportation has become an essential component of modern daily life, with continuous advancements aimed at reducing travel time and improving mobility. However, this increased convenience has also contributed to a rise in road accidents, often caused by driver distraction, fatigue, and age-related cognitive decline. These concerns have driven growing interest in Artificial Intelligence (AI)-based real-time driver monitoring systems designed to enhance road safety. Despite recent progress, several challenges remain, including limitations in detection accuracy, inadequate temporal reasoning, and high computational complexity. To address these challenges, we utilize the EfficientNetV2-S model for backbone feature extraction due to its high performance, compact model size, and fast inference speed. The model leverages Squeeze-and-Excitation (SE) attention to enhance feature learning capabilities; however, SE attention captures only channel information and overlooks important spatial features. To overcome this limitation, our framework incorporates a Driver-Monitoring Coordinate Attention (DM-CA) mechanism with modifications that encode features along both height and width directions. Beyond frame-level classification, the proposed system integrates temporal memory and reasoning to convert frame-level predictions into behavior-level insights. A risk-aware decision module evaluates the drivers state based on duration, context, and driving conditions, enabling goal-driven adaptive interventions such as warnings or alerts, supported by a feedback adaptation mechanism. We evaluated the framework on UTKFace, Fatigue, modified Fatigue, and 100-driver datasets, demonstrating its effectiveness in understanding driver distraction. Quantitative and qualitative analyses, including state-of-the-art comparison, cross-validation, statistical analysis, and computational complexity evaluation, highlight the system’s accuracy, modularity, and suitability for real-world deployment.