This article introduces a dataset that investigates the physiological responses of drivers when using advanced driver assistance systems (ADAS) in real-world traffic conditions. The study, conducted in the Federal District, Brazil, involved seven drivers in controlled driving sessions. The time of day and the days of the week were standardized to ensure comparable traffic conditions. The data collection was centered on ADAS Level 2 systems, specifically the Lane Keeping Assist System (LKAS) and the Forward Collision Warning System (FCWS). The dataset includes five physiological signals: respiration, heart rate, galvanic skin response (GSR), leg muscle activity, and brain activity. These signals were continuously acquired using a dedicated instrumentation system installed in the vehicle. Given the complexity of collecting data under real traffic conditions, the acquisition sessions generated a large volume of raw data. Considerable post-processing was conducted to identify and segment portions of the signals with sufficient integrity for subsequent analysis. The dataset is structured as time-stamped raw signal spreadsheets, each corresponding to a specific driver and direction of the pre-established route (outbound and return). Such organization enables researchers to navigate the dataset easily, explore specific segments of interest, and conduct comparative analyses across participants and varying traffic conditions. The dataset is relevant to researchers in biomedical signal processing, driver state monitoring, intelligent transportation systems, and human-machine interaction. It may be used by academic laboratories investigating physiological responses during driving tasks, as well as by engineers and developers working on advanced driver assistance systems (ADAS), including automotive manufacturers and ADAS technology suppliers. The dataset, which includes synchronized physiological and vehicle dynamics data collected under real traffic conditions may contribute to the study of human responses during semi-automated driving, supporting research and development of driver-centered mobility technologies.

Accurate and efficient lane detection is required for the effective operation of autonomous vehicles and advanced driver-assistance systems need it for their functioning. Traditional methods frequently use computationally demanding trigonometric operations, which are particularly challenging during line detection using the Hough Transform. One of the focused areas of this study is a real-time lane detection framework based on CORDIC technology. The CORDIC system operates as an iterative process that performs fundamental mathematical functions to enable its implementation on embedded platforms, whereas the matrix Mactor usage of CORDIC iterative methods presents an alternative to regular sine and cosine computations. The proposed pipeline implementation combines region-of-interest masking with Canny edge detection, modified Hough Transform, and CORDIC methods to detect multiple straight lane lines. The CORDIC implements the polar line equation using an iterative rotation method, thereby minimizing the computational requirements. The method can accurately extract multiple lane lines while significantly increasing the processing speed, as shown in the experimental results for both clear and rain-blurred highway images. The CORDIC-enhanced method shows advantages over standard algorithms through timing benchmarks, accumulator space visualizations, and performance metrics, which display the complete results of in-depth comparisons. This study demonstrates how hardware-oriented computation combined with algorithmic optimization enables real-time automotive applications and intelligent transportation systems to scale while achieving a correct rate of 98.72%.

As a fundamental technology in the field of autonomous driving, lane detection plays a critical role in various advanced driver-assistance systems, including lane keeping, lane departure warning, lane change assistance, and forward collision warning. However, due to vehicle load, lane lines are prone to damage, making lane line detection more difficult. This study proposes SALane, an improved row-based ultra-fast and lightweight lane line detection network aiming at balancing ultra-fast speed and accuracy. Specifically, the Inception-v3 network is employed as a feature extractor, where the input image is partitioned into a grid with significantly fewer cells than pixels via anchor points. This design enables efficient localization of lane lines directly on the grid. Furthermore, Otsu’s method is utilized to automatically determine an optimal image binarization threshold. Additionally, a hybrid approach combining random sample consensus and least-squares fitting is introduced to enhance the robustness and accuracy of lane line modeling. Cross entropy is also used to attenuate the effect of imbalance between lane lines and background categories, and a simple calculation method of lane marking degradation rate is proposed. On the CULane benchmark, SALane achieves a total accuracy of 79.6% at a speed of 60.2 frames per second, which outperforms the compared algorithms by 10% in accuracy and demonstrates a well-balanced performance between precision and speed. Further, this study explores the effect of the degree of lane marking degradation on the detection results, and the algorithm performance is stable when the damage category is less than 3. However, as the number of damage categories increases to 4 and 5, the performance of the algorithm decreases sharply.

Lane detection stands as a foundational capability for modern autonomous vehicles and advanced driver assistance systems (ADAS). Despite decades of progress, the problem is far from solved — road environments remain unpredictable, lane markings vary dramatically across geographies, and real-time constraints impose strict computational budgets on embedded hardware. This paper offers a structured review of the state of the art in deep learning-based lane detection, tracing the field from early convolutional segmentation approaches through the latest transformer-based architectures. We examine the progression from pixel-level classification frameworks to anchor-driven regression models, polynomial curve fitting strategies, and end-to-end detection transformers. We further survey the benchmark datasets and evaluation metrics that shape how the community measures progress, and critically assess persistent challenges including adverse weather, occlusion, non-standard lane markings, and the sim-to-real gap. Drawing on these findings, we identify directions that are likely to define the next generation of lane detection systems, including multi-task learning, domain-adaptive training, 3D lane reconstruction, and lightweight deployment on embedded automotive platforms.

This work presents a novel method to integrate Gaussian-process tire models into a Kalman filter for simultaneous vehicle velocity and sideslip angle estimation. These quantities enable advanced driver assistance systems and impact the performance of autonomous driving software. Most estimation approaches use vehicle model-based Kalman filters, relying on accurate system identification. Considering tire parameters, this can be challenging in the case of limited excitation (longitudinal or lateral slip). Gaussian-process tire models omit this by additionally estimating model uncertainty. Current state-of-the-art Gaussian-process tire models only consider unidirectional excitation. We develop a Gaussian-process tire model to additionally consider combined excitation and tire temperature. Next, we integrate the developed model into an Unscented Kalman Filter for vehicle dynamics state estimation and compare the results with different Magic Formula models. The study shows how covariance adaptation based on Gaussian-process tire models can decrease the impact of model mismatch due to missing excitations in the tire fitting procedure. All studies are evaluated on real-world data acquired with a full-scale autonomous Dallara EAV24 Super Formula race car during the Abu Dhabi Autonomous Racing League on the Yas Marina Formula 1 circuit. Our C++ code is available at https://github.com/TUMFTM/GPTireUKF.

The long-wave infrared band, which at room temperature covers the infrared radiation of humans and objects, has significant applications across various fields including wireless communication, national defense, military, biomedical, and advanced driver assistance systems. Metalens provides a pathway to lightweight, compact, and integrated solutions for infrared imaging and sensing systems, marking an inevitable trend in future development. This study presents a design for a high numerical aperture of 0.89 in a polarization-insensitive all-chalcogenide metalens operating at 10 µm, utilizing the commercially available chalcogenide glass material As2Se3 via a transmission phase approach. Building upon this, we have achieved, for the first time, a high numerical aperture of 0.84 for an all-chalcogenide broadband LWIR achromatic metalens operating in the 9.5–10.5 µm range, with significantly improved focusing performance through the application of particle swarm optimization algorithms. The superior performance of the all-chalcogenide LWIR metalens, combined with the advantages of chalcogenide glass over traditional LWIR materials such as Si or Ge—namely, lower cost, reduced optical loss, and a smaller thermo-optic coefficient—suggests it has significant potential for broader applications.

The rapid introduction of Advanced Driver Assistance Systems (ADAS) poses a unique challenge for older drivers, who often face barriers in adopting these technologies. This study evaluates the effectiveness of a practical, simulator-based training concept designed specifically for drivers aged 50+. The empirical analysis of a research group of 25 people, focused on verifying four research hypotheses regarding the simulator's suitability, trust calibration, user awareness, and training utility. The results confirmed that the high-fidelity simulator is an appropriate training environment for this demographic; analysis of the Revised Simulator Sickness Questionnaire (RSSQ) revealed a statistically significant reduction in symptoms during the adaptation process, validating the physical feasibility of the training (H1). The intervention led to a measurable increase in trust towards ADAS, with a strong effect size, confirming positive behavioral adaptation (H2). Furthermore, participants demonstrated raised awareness of system benefits, primarily identifying enhanced safety and speed control (H3). The proposed training model achieved high internal consistency and received positive subjective usability ratings (H4). These findings support the deployment of simulator-based practical training as an effective tool for preventing digital exclusion among older drivers.

Road traffic accidents continue to be a leading cause of fatalities worldwide, with human error contributing to nearly 90% of incidents. Advanced Driver Assistance Systems (ADAS), particularly Automatic Emergency Braking (AEB), have significantly improved vehicle safety but remain expensive for widespread adoption in developing regions. This paper presents the design and development of a low-cost Intelligent Automatic Braking System (ABS) using ultrasonic sensing and embedded control. The system employs real-time distance measurement and a threshold-based decision algorithm implemented on an Arduino platform. A multi-stage braking strategy is introduced to enhance safety and comfort. The proposed system is evaluated through experimental testing and compared with existing AEB and Automatic Preventive Braking (APB) approaches. Results demonstrate a response time below 50 ms and reliable obstacle detection, making the system suitable for low-speed urban applications.

Harsh Road conditions in developing nations, particularly India, present a formidable challenge for Advanced Driver Assistance Systems (ADAS). Traditional computer vision-based systems often struggle with lighting variations, high dust levels, and waterlogged roads. This paper presents a specialized, low-cost solution using an array of RCWL-1655 waterproof ultrasonic sensors combined with high-speed servo actuators to create a dynamic "scanning" phased array. We propose a differential distance algorithm that identifies potholes, speed-breakers, and minor road bumps with high fidelity. The system utilizes an ESP8266-based edge-computing architecture to provide low-latency classification. This paper provides an in-depth mathematical analysis of the scanning geometry, signal processing filters, and the physical constraints of ultrasonic propagation at vehicle speeds. Experimental results demonstrate a 96.2% detection accuracy for potholes and a 93.5% accuracy for speed-breakers at urban speeds, providing a robust, low- cost candidate for ADAS integration in unpredictable environments.

Vulnerable road users in informal urban environments confront a distinct set of hazards that standard computer vision datasets are ill-equipped to represent: artisanal speed bumps constructed without regulatory compliance, deteriorated road markings, and the mototaxi-a three-wheeled motorized vehicle that constitutes the primary informal transport mode in intermediate Andean cities yet is absent from all major international repositories. This paper presents QHAWAY-from Quechua qhaway, a transitive verb meaning "to look; to observe"-an Advanced Driver Assistance System (ADAS) predicated on instance segmentation, monocular distance estimation via the pinhole camera model, and Time-to-Collision (TTC) computation, developed for the road environment of Ayacucho, Peru (2761 m a.s.l.), a city recognised by UNESCO as a Creative City of Crafts and Folk Art since 2019. A hybrid dataset comprising 25,602 images with 127,525 annotated instances across 12 classes was assembled by combining an original local collection of 4598 images (10,701 instances) captured through four complementary acquisition methods across the five urban districts of the Huamanga province with three established international datasets (BDD100K, BSTLD, RLMD; 21,004 images, 116,824 instances). A three-phase progressive training strategy with monotonically increasing resolution (640, 800, and 1024 pixels) was evaluated as an ablation study. A multi-architecture comparison spanning YOLOv8L-seg and the YOLO26 family (nano, small, large) identified YOLO26L-seg as the best-performing model, attaining mAP50 Box of 0.829 and mAP50 Mask of 0.788 at epoch 179. The integration of ByteTrack multi-object tracking with the pinhole equation D=(Hreal×f)/hpx delineates operational risk zones aligned with the NHTSA forward collision warning standard (danger: <3 m; caution: 3-7 m; TTC threshold ≤ 2.4 s). The system sustains processing rates of 19.2-25.4 FPS on an NVIDIA RTX 5080 GPU. A systematic field survey established that 96% of the audited speed bumps fail to comply with MTC Directive No. 01-2011-MTC/14, constituting the first quantitative record of informal road infrastructure non-compliance in the Andean region. Validation was conducted under naturalistic driving conditions without staged scenarios. Grad-CAM explainability analysis, encompassing three complementary visualisation algorithms (Grad-CAM, Grad-CAM++, and EigenCAM), confirmed that model attention concentrates consistently on safety-critical objects.

Automated vehicles enhance road safety and efficiency through Advanced Driver Assistance Systems (ADAS) like Adaptive Cruise Control (ACC). However, the reliance on onboard sensors makes ACC vulnerable to False Data Injection (FDI) attacks, which can compromise safety and stability. To address the lack of formal guarantees in conventional systems, this paper proposes a safety-aware, Lyapunov-based ACC framework resilient to FDI attacks. The framework integrates a Lyapunov function with Control Barrier Functions (CBF). The Lyapunov function synthesizes stable control actions by simultaneously capturing tracking dynamics and real-time estimation of both inter-vehicle distance and the FDI attack. Concurrently, the CBF ensures the actual inter-vehicle distance remains above a predefined safety threshold by constraining the control action space. A Quadratic Programming (QP) formulation coordinates these objectives, resolving the trade-off between safety enforcement and stability preservation. Simulation results demonstrate that the proposed framework effectively ensures stability, maintains safety, and accurately mitigates FDI attacks in ACC systems.

Comparative safety evaluation of ADAS-Equipped electric and gasoline vehicles using real-world crash data.

Modeling vehicle-cyclists' interactions to support automated driving and advanced driving assistance systems

Signal temporal logic-based neural network for driving task generation in Advanced Driver Assistance Systems

Lane detection remains a high-priority research area for Advanced Driver Assistance Systems (ADAS) and Autonomous Driving (AD). To address the limitations of current mainstream convolutional neural networks in capturing global context and the elongated structure of lanes, we study a lane detection algorithm based on a dual attention mechanism using Transformer architecture (DATLD) for connected and autonomous vehicles. On the other hand, a new infrared sensor for onboard use that is not affected by light is used for lane line image acquisition to detect lane lines throughout the day. Firstly, an enhanced dense connected Backbone network is developed to improve channel feature extraction capability. Secondly, a novel Transformer architecture (CSVIT) based on a dual attention mechanism is designed to conduct global modelling from the two orthogonal dimensions of space and channel. Improve the accuracy of lane detection under different weather, environment and time conditions. Finally, the accuracy of DATLD on the TuSimple data set was improved to 95.71%, and the FPS was 1030, satisfying the lane detection real-time requirement of at least 30 FPS. The F1 score on the CULane dataset increased to 66.12%. We have verified the validity and feasibility of DATLD in the Autonomous-rail Rapid Transit (ART)Center, thus ensuring driving safety.

Traffic sign (TS) detection and recognition are essential components of advanced driver assistance systems (ADAS), contributing to safer and more reliable driving. However, deploying deep learning–based vision models on embedded platforms is challenging due to constraints in computational power and energy consumption. In this work, a comparative deployment of you only look once version 7 (YOLOv7) and YOLOv7-tiny deep learning algorithms is conducted on embedded NVIDIA platforms, namely Jetson Nano and Jetson Xavier NX, to evaluate their suitability for real-time TS detection. Following the detection stage, a convolutional neural network (CNN) is integrated to perform TS recognition, enabling a complete detection–recognition pipeline. Experimental results show that YOLOv7-tiny achieves higher detection precision of 97%, while providing better speed and computational cost on resource-constrained devices, with Jetson Nano reaching 18.8 frames per second (FPS) and, on Jetson Xavier NX reaching 43 FPS. The integrated CNN model ensures reliable classification of detected TS with an accuracy of 99.54%. This work highlights the trade-offs between precision, speed, and power consumption and provides practical guidance for selecting detection and recognition architectures for embedded ADAS applications.

Abstract While the existing object detection models have robust feature learning capability, there were challenges when deployed on embedded devices for traffic information detection due to the large number of parameters and heavy computational demand. In this regard, an MLETI-P (Multi-scale lightweight and efficient traffic information after pruning) model was developed in this paper. Firstly, four scales of object detection layers were designed to obtain multi-scale characteristics and enhance the recognition ability of traffic participants. Secondly, a lightweight MLE detection head was designed by coupling the 3DLEBlock with parameter sharing, effectively reducing the parameter's count and FLOPs (Floating Point Operations). Thirdly, a multi-scale neck structure was proposed to reduce the parameter count while maintaining a high feature extraction capability. Finally, a layered adaptive magnitude pruning method was introduced to obtain MLETI-P, which reduces the overall parameter count by pruning the redundant channels in the model. In addition, the comparison experiments with different pruning rates were performed to trade off the precision and the number of model parameters better. The ablation results showed that compared with the YOLOv8n model, the accuracy of MLETI-P was improved by 1.7%, but the model parameters, the FLOPs, and the model size were only 48.3%, 48.1%, and 51.9%, respectively. In the KITTI dataset and BDD100K dataset, compared with the existing advanced lightweight models, MLETI-P showed superior performance than the others. The MLET-P model was also embedded in an experimental car, the mainboard of which was the NVIDIA Jetson Xavier NX. The tested average FPS (frames per second) was 24.2, which indicated that the MLET-P model could perform smoothly on embedded systems. It is anticipated that MLETI-P could be integrated into autonomous vehicles to obtain traffic information in the future.

Abstract Path planning is a major area in mobile robotics and autonomous vehicles allowing the interpretation of the circumstances and the generation of safe, obstacle-free trajectories from the starting point to the destination. Light Detection and Ranging (LiDAR) sensors perform an essential role in environmental localization and mapping for autonomous systems. The proposed framework employs a RPLiDAR-A1 360° sensor for detailed environmental mapping and obstruction identification. This article proposes a novel heuristically scaled weighted A* algorithm on the Cartesian occupancy grid mapping to attain an optimal route in static environments. Additionally, an optimization of the key node selection technique is developed to smooth the trajectory by removing unnecessary nodes and reducing computational cost. Furthermore, a safer zone creation strategy is implemented to strengthen navigation reliability by triggering a warning for path intersections within unsafe regions. Experiment results reveal that the proposed path planning algorithm in the Cartesian grid map generates the shortest path length of 277.88mm and eliminates enormous unnecessary turns. In terms of computational efficiency, it reduces the number of extended nodes up to 25 and finds an optimal path within 1.1ms. These findings demonstrate the significance of the proposed algorithm in generating optimal paths as well as reducing the computational complexity of autonomous pathplanning applications.

In highly automated driving systems (ADSs), drivers’ ability to resume manual driving remains a road safety issue. However, to the best of our knowledge, there is no existing computational model to predict which drivers require more than the 4 seconds mandated by United Nations Regulation No. 157 to regain manual control. To address this challenge, we developed a Random Forest model that predicts takeover time using measurable human factors. Three controlled driving simulator experiments were conducted in which participants engaged in distinct tasks—texting, drinking, and traffic monitoring—before responding to a takeover request. During the experiments, we collected human factor features, including gaze behavior, age, and scores, from the self-reported driving behavior questionnaire (K-DBQ). The Random Forest classifier achieved 77% accuracy. Recursive feature elimination selected 10 dominant predictors; notably, engaging in non-driving-related tasks, reduced on-road gaze, and older age were significantly associated with longer takeover times. Although K-DBQ scores were not directly correlated with takeover time, their inclusion improved model robustness, consistent with ensemble learning from weak yet complementary signals. The proposed model can be integrated into advanced driver assistance systems (ADASs) to proactively identify drivers likely to exceed the 4-second takeover window, support targeted interventions, and enhance human-centered transition safety in ADSs.

A performance-efficient compact, highly selective 24GHz SIW-based bandpass filter (BPF) is proposed for automotive radar and advanced driver-assistance systems (ADAS). The key contribution is a single-layer SIW-DGS-CPW co-design in which SIW cavities provide a high-Q passband, an open-rectangular DGS introduces transmission zeros to steepen the skirts and reinforce the stopband, and a CPW feeding transition improves matching and practical integration. Implemented on a substrate of Rogers RO4003C with [Formula: see text] and thickness of 0.508mm, the prototype occupies [Formula: see text], achieving a pronounced miniaturization while maintaining strong spectral selectivity. Full-wave simulations and measurements confirm a center frequency of 23.97GHz and a 450-MHz 3-dB bandwidth, with return loss better than 24dB and an in-band insertion loss of 1.6-2.0dB ([Formula: see text]). The filter exhibits sharp roll-off with a measured 40-dB rejection within ± (0.8-1.2) GHz from [Formula: see text], and 30-35dB suppression across 20-30GHz. A compact equivalent-circuit model captures the passband behavior and transmission zeros. Thermal analysis (25-105°C) shows only a slight downshift (~ 30MHz) with minimal performance degradation, supporting automotive reliability. Compared with prior 24-GHz BPFs, the proposed co-integration simultaneously improves skirt selectivity and wide stopband suppression within a compact footprint.