[EN] Highlights What are the main findings? The EVACH driving simulator, equipped with custom hardware and data acquisition systems, achieved precise calibration of the braking and steering controls. Comparative analysis between naturalistic tests and virtual simulations showed that the simulator reproduces autonomous driving speeds with fidelity and stability. What is the implication of the main finding? The EVACH simulator provides a robust and reliable platform for investigating driver behavior and human-machine interaction in SAE 2 and SAE 3 scenarios. Its validated calibration and reliable virtual environment enable safe, cost-effective, and versatile experimentation for future autonomous driving research projects.Highlights What are the main findings? The EVACH driving simulator, equipped with custom hardware and data acquisition systems, achieved precise calibration of the braking and steering controls. Comparative analysis between naturalistic tests and virtual simulations showed that the simulator reproduces autonomous driving speeds with fidelity and stability. What is the implication of the main finding? The EVACH simulator provides a robust and reliable platform for investigating driver behavior and human-machine interaction in SAE 2 and SAE 3 scenarios. Its validated calibration and reliable virtual environment enable safe, cost-effective, and versatile experimentation for future autonomous driving research projects.Abstract The integration of autonomous vehicles (AVs) into road transport requires robust experimental tools to analyze the human-machine interaction, particularly under conditions of system disengagement. This study presents the primary controls calibration and virtual scenario validation of the EVACH autonomous driving simulator, designed to reproduce the SAE Level 2 and Level 3 driving modes in rural road scenarios. The simulator was customized through hardware and software developments including a dedicated data acquisition system to ensure the accurate detection of braking, steering, and other critical control inputs. Calibration tests demonstrated high fidelity, with minor errors in brake and steering control measurements, consistent with values observed in production vehicles. To validate the virtual driving rural environment, comparative experiments were conducted between naturalistic road tests and simulator-based autonomous driving, where five volunteers participated in the preliminary pilot test. Results showed that average speeds in the simulation closely matched those recorded on real roads, with differences of less than 1 km/h with minimum standard deviation and confidence values. These findings confirm that the EVACH simulator provides a stable and faithful reproduction of autonomous driving conditions. The experimental platform offers valuable support for current and future research on the safe deployment of automated vehicles., This work was funded by MCIN/AEI/10.13039/501100011033/ and by FEDER A way of making Europe, with reference PID2021-127183OA-I00.

[EN] Concerns regarding the safety of partial automated vehicles (AVs) remain prevalent, especially in complex roadway environments. AVs incorporate Advanced Driving Assistance Systems to perform the dynamic driving task; however, unexpected disengagements of the systems still occur due to various contextual and infrastructural factors. Among these, two-lane rural roads¿with their geometric design and operational challenges¿represent a critical setting. Notably, a high number of disengagements are concentrated on horizontal curves.
This study analyzes naturalistic disengagement data from two SAE Level 2 AVs operating on different segments of two-lane rural roads. The horizontal curves were characterized across geometrical and operational variables (radii, curvature change rate (CCR), curve direction, speed or visibility). These variables served as inputs to an artificial neural network (ANN) model designed to predict disengagement occurrences.
The ANN, a feedforward multilayer perceptron with one hidden layer, was trained using backpropagation. Performance was validated with K-fold cross-validation, and accuracy assessed via cross-entropy loss and confusion matrices. A Monte Carlo-style simulation tested robustness by generating multiple confusion matrices from randomized data partitions to evaluate classification stability. The results highlight CCR and lane width as key predictive factors. The calibrated ANN demonstrated robust classification (accuracy=87.8%, sensitivity= 92.7%, specificity= 85.9%) in identifying curve segments with a higher likelihood of disengagement.
This study provides road administrations a new neural network derived empirical formula to identify potential AV disengagement zones. By identifying risk-prone areas, authorities can consider targeted measures¿such as enhanced signage or driver alerts¿ to support safer and more efficient automated driving in rural settings., This research is part of the project Assessment of the Automated Vehicles Performance considering the physical road infrastructure
and human perception (EVACH) with Grant PID2021-127183OA-I00 funded by MCIN/AEI/ 10.13039/501100011033 and by ERDF
A way of making Europe .

[ES] Este Trabajo Fin de Máster analiza la geometría de carreteras convencionales y la operatividad que los vehículos automatizados (VA) de nivel SAE 2 tienen en ella, aplicando la metodología a un tramo representativo de la CV-35, CV-50 y la CV-345 en la provincia de Valencia, España. El estudio combina un análisis de la normativa y la utilización de herramientas de aprendizaje automático como son las Redes Neuronales Artificiales (ANN) para identificar las condiciones geométricas que determinan la conectividad de los sistemas automatizados y proponer medidas de mejora que aumenten dicha conectividad, aumentando así tanto la seguridad como el confort de la conducción.
La investigación parte de un análisis geométrico de los elementos de la carretera (curvas, rectas y clotoides), para posteriormente entrenar la red neuronal artificial con datos reales de circulación. Mediante la utilización de la ANN se obtiene una ecuación predictiva capaz de clasificar las curvas, identificado como elemento más relevante a la hora de desconexión del sistema, en tres categorías: conectadas, desconectadas o inciertas, a partir de las variables que mostraron mayor influencia, las cuales fueron la tasa de cambio de curvatura (CCR) y el ancho de carril, parámetros que definen la exigencia geométrica sobre el control lateral del vehículo y el margen disponible para maniobrar.
Posteriormente, se aplica la ecuación predictiva únicamente a los tramos de la CV-35, donde se analizaron 122 curvas, de las cuales 21 resultaron conectadas, 62 desconectadas y 39 se situaron en la franja de incertidumbre entre los dos umbrales de decisión, lo que refleja que las carreteras actuales no están preparadas para una automatización nivel SAE2.
A partir de este diagnóstico se proponen intervenciones geométricas selectivas, las cuales consisten en el suavizado de curvas y en la aplicación de sobreanchos de carril en zonas críticas. Estas actuaciones permiten encadenar subtramos de conexión garantizada de más de un kilómetro, mejorando además el cumplimiento normativo. Como medida complementaria, se propone la implantación de señalización vertical específica que advierta al usuario de los subtramos sin conexión, anticipando la transición a conducción manual y reforzando la seguridad vial.

[EN] This paper presents an on-road study examining the interaction between Level 2 automated vehicles (AVs) and the geometric features of two-lane rural roads. These roads constitute the largest percentage of the road network in most countries. Their complex geometric features and lack of design consistency present a significant challenge for the operation of automated vehicles (AVs). Although AV systems aim to reduce driver workload and improve road safety, their performance remains limited on these roads, often resulting in automated system disengagements that require driver intervention. Such disengagements highlight emerging challenges for road administrations regarding road readiness and the safe integration of AVs into existing infrastructure.
Based on an on-road study using two instrumented AVs, this study investigates how geometric features of Homogeneous Road Segments (HRS), including Curvature Change Rate (CCRHRS), average radius, tangent length, and average parameter of vertical curves, affect driving system engagement rates. A Bayesian approach was used to quantify the relationship between road geometry indices and AV disengagements. The findings reveal that the primary factor affecting engagement status is CCRHRS, showing that higher values of CCRHRS (>400 gon/km) are associated with AV engagement dropping below 25%, highlighting a direct impact of geometric complexity on AV reliability. The other geometric indices have a moderate positive effect. The findings underscore that road geometry significantly affects AV reliability.
These findings have significant implications for infrastructure planning and policy. By identifying geometric thresholds that compromise AV functionality, this research supports the definition of Operational Road Sections (ORS) aligned with Operational Design Domains (ODD). The proposed model offers a practical tool for road administrations to assess road suitability for AV deployment, inform upgrading priorities, and establish design guidelines that facilitate the safe integration of AV technologies into existing rural networks. As automation advances, aligning infrastructure design with AV capabilities becomes essential for equitable and effective transport policy., This research is part of the project Assessment of the Automated Vehicles Performance considering the physical road infrastructure and human perception (EVACH) with Grant PID2021-127183OA-I00 funded by MCIN/AEI/10.13039/501100011033 and by ERDF A way of making Europe .

[EN] The integration of autonomous vehicles (AVs) into road transport requires robust ex-perimental tools to analyze human–machine interaction, particularly under conditions of system disengagement. This study presents the development and validation of the EVACH autonomous driving simulator, designed to reproduce SAE Level 2 and Level 3 driving modes in rural road scenarios. The simulator was customized through hardware and software developments, including a dedicated data acquisition system to ensure accurate detection of braking, steering, and other critical control inputs. Calibration tests demonstrated high reliability, with minor errors in brake and steering control meas-urements, consistent with values observed in production vehicles. To validate the virtual driving rural environment, comparative experiments were conducted between natu-ralistic road tests and simulator-based autonomous driving. Results showed that average speeds in simulation closely matched those recorded on real roads, with differences of less than 1 km/h and significantly lower variability. These findings confirm that the EVACH simulator provides a stable and faithful reproduction of autonomous driving conditions. The platform represents a validated and versatile tool for evaluating driver workload, takeover performance, and human–machine interaction, offering valuable support for current and future research on the safe deployment of automated vehicles.