VEHITS 2026 Abstracts

Area 1 - Connected Vehicles

Full Papers

Paper Nr:	32
Title:	Vehicle Security Operations Center: Future-Proofing ISO/SAE 21434 and UN Regulation No. 155 for Modern Vehicles
Authors:	Jenny Hofbauer, Manfred Vielberth and Kevin Mayer
Abstract:	The integration of traditional Information Technology (IT) systems and increased connectivity has exposed modern vehicles to a broader range of threats, prompting the introduction of cybersecurity standards and regulations, such as ISO/SAE 21434 and UN Regulation No. 155. The automotive industry adopted an analogous approach to traditional enterprise IT, by employing a Vehicle Security Operations Center (VSOC) to provide regulation-compliant cybersecurity services specialised in protecting vehicle fleets and their ecosystem on the road. This research examines whether a VSOC compliant with ISO/SAE 21434 and UN Regulation No. 155 sufficiently protects modern vehicles against emerging and future threats, and how it can be evolved. The structure of a VSOC is established based on literature and a survey to define a framework of capabilities. Shortcomings in ISO/SAE 21434 and UN Regulation No. 155 are identified by mapping roles, tools, and processes to an established enterprise IT Security Operations Center (SOC). As a result, differences in maturity and underrepresented traditional IT infrastructure are considered in the automotive context, and applicable recommendations are derived.
Download

Short Papers

Paper Nr:	28
Title:	Automotive Security Architectures for Plug & Charge with a Central Trusted Platform Module
Authors:	Stephan Zitzlsperger, Mahboubeh Tajmirriahi, Abhishek Subedi, Simon Rudhart, Daniel Trick, Martin Schramm and Christian Plappert
Abstract:	The rapid adoption of Electric Vehicles (EVs) and their integration into ecosystems with uncertain security underscores the need for robust cybersecurity such as zero-trust architectures. Plug and Charge (PnC), defined in ISO 15118-20, is vulnerable without hardware-based protection. This risk increases in zonal vehicle architectures, where multiple Electronic Control Units (ECUs) share resources, making centralized security critical. We propose a centralized Trusted Platform Module (TPM) 2.0 as a hardware trust anchor for PnC. Unlike local solutions, our architecture consolidates cryptographic operations within a single TPM 2.0 on a High Performance Controller (HPC), enabling secure key generation, storage, and usage for multiple ECUs. The methodology integrates one TPM 2.0 into the EcoG-io/ISO15118 framework and emulates Electric Vehicle Communication Controller (EVCC) and Supply Equipment Communication Controller (SECC) using Raspberry Pi devices. By generating and managing keys inside the TPM, the approach eliminates private key exposure and ensures ISO 15118-20 compliance. The solution offers enhanced security through a single hardware root of trust and simplified scalability for zonal architectures. Compared to local TPMs, one centralized TPM 2.0 lowers hardware costs, streamlines maintenance, and enables resource sharing.
Download

Paper Nr:	30
Title:	Replay Cyber-Attack Detection and Isolation in Vehicle Lateral Dynamics with Event-Based Communication
Authors:	Fatemeh Tohfeh, Ali Eslami and Khashayar Khorasani
Abstract:	This paper investigates the problem of replay cyber-attack detection and isolation in vehicle lateral dynamics with event-triggered communication. To reduce communication load, event-triggered communication schemes are considered on the output channels between the vehicle plant and the Electronic Control Unit (ECU). By utilizing the notion of auxiliary systems operating in parallel with the lateral dynamics, a detection framework is developed to identify inconsistencies in both received measurements and inter-event timing. In addition, an isolation mechanism is proposed to distinguish replay cyber-attacks from general false data injection cyber-attacks. Sufficient conditions are provided to ensure boundedness of the closed-loop system and to prevent Zeno behavior, making the proposed approach suitable for real-time vehicular applications.
Download

Paper Nr:	31
Title:	Evaluating the Impact of Vehicle-To-Pedestrian Communication on Pedestrian Safety and Behavior at Urban Intersections Using Virtual Reality
Authors:	Zeyu Mu, Ismet Goksad Erdagi, Milan Knezevic, Marko Vukojevic, Aleksandar Stevanovic and B. Brian Park
Abstract:	As connectivity technologies become more prevalent, ensuring safe and intuitive interaction between pedestrians and vehicles remains a critical challenge, especially at urban intersections. This study developed a virtual reality (VR) simulation framework to investigate how varying levels of vehicle-to-pedestrian (V2P) communication affect pedestrian behavior during intersection crossings under different pedestrian crossing time conditions. Participants experienced three communication scenarios: no communication, partial communication (auditory warning), and full communication (warning with intent sharing), while crossing an intersection with short or long crossing times. Results indicate that communication reduces collisions and increases pedestrian yielding behavior, particularly in time-constrained (short crossing time) conditions. Female participants exhibited more cautious behavior and lower walking speeds, leading to fewer collisions than males. Additionally, pre- and post-experiment surveys revealed that participants felt safer, better informed, and more comfortable in communication-enabled scenarios, with increased willingness to adopt and pay for such systems. These findings highlight the value of V2P connectivity in enhancing pedestrian safety and underscore the potential of VR as a tool for evaluating human responses to emerging vehicle communication technologies.
Download

Paper Nr:	70
Title:	Cooperative Maneuver Coordination for Prioritization of Public Transport and Emergency Vehicles
Authors:	Matthias Nichting, Thomas Lobig, Claas-Norman Ritter, Nicolai Steinke, Stephan Sundermann, Daniel Goehring, Julian Pfeifer, Johann Nikolai Hark, Ilja Radusch, Jorin Kouril, Christopher Schahn and Bernd Schaeufele
Abstract:	This paper describes the implementation and evaluation of maneuver coordination between normal vehicles, public transport vehicles, and emergency vehicles on the road. An implicit coordination method based on the exchange of the maneuver coordination message (MCM) is investigated. Different use cases of non-trivial traffic situations are defined, and experiments with real vehicles are conducted. In addition, the special challenges posed by the use of different automation stacks in the vehicles participating in the experiments are discussed. The evaluation of the experiments shows the effectiveness of the method for the coordination of traffic in case of conflicts between the planned trajectories of different vehicles and vehicle types. The results are interpreted as a proof of concept with further research required.
Download

Area 2 - Intelligent Transport Systems and Infrastructure

Full Papers

Paper Nr:	61
Title:	IMAP: Enhancing First- and Last-Mile Accessibility through Public Transport-Aware Intermodal Micro-Mobility Allocation
Authors:	Rania Swessi, Zeineb El Khalfi and Mohamed Mosbah
Abstract:	Shared micro-mobility has become an essential component of modern urban transportation, offering convenient options for short-distance travel. It helps bridge the so-called “first and last mile” gap-connecting passengers from their origins to public transport stops and supporting onward travel beyond them. Such integration encourages public transit use, decreases dependence on private vehicles, and helps mitigate urban congestion. However, this is often challenged by the unavailability of micro-mobility vehicles at the right locations, specifically near public transport stops, and at the right times. In this paper, we propose a short-term micro-mobility demand prediction framework based on a hybrid graph-based GCN–TCN–Transformer architecture designed to capture spatial dependencies and temporal patterns. The model integrates local public transport demand together with high-impact exogenous factors, including demographic, functional, meteorological, and temporal features, which are known to have a strong influence on shared mobility demand. The effectiveness of the proposed framework is validated through comprehensive experiments conducted on recent real-world operational datasets collected in New York City in 2025. This evaluation includes comparisons with state-of-the-art baseline models, showing up to a 10.7% MSE reduction over the strongest baseline model, and an analysis of input–output temporal settings. It also includes an ablation study assessing the contribution of each model component and a detailed examination of the prediction results.
Download

Short Papers

Paper Nr:	25
Title:	Road Network Dynamic Criticality Assessment Based on Floating Car Data
Authors:	Iñaki Cejudo, Harbil Arregui, Eider Irigoyen, Thomas Dimos, Josep Maria Salanova Grau and Evdokimos Konstantinidis
Abstract:	Traffic congestion is one of the biggest challenges metropolitan areas face, affecting economic efficiency and urban quality of life. Urban growth and evolution alter traffic dynamics; therefore, constant monitoring and plans for decision-making are necessary. Understanding how road status changes over time is crucial for making these informed decisions. This work presents a method for analysing the dynamic criticality of the road network in Thessaloniki, leveraging OpenStreetMap and vehicle GPS data collected over several days. A network graph is built, and Betweenness Centrality is calculated at different moments throughout a week, taking into account the dynamic nature of traffic, to identify the most vulnerable roads. Additionally, ”what-if” strategies methodology is applied to simulate scenarios and identify reserve roads. The results provide a detailed analysis of the network, identifying critical zones and roads vulnerable to traffic congestion. This study offers a methodology and insights for traffic management strategy planning and urban infrastructure enhancement.
Download

Paper Nr:	33
Title:	A Data-Driven Framework for Climate-Aware Asset Failure Explanation in Urban Tram Systems
Authors:	Mohamed Amine Ayachi, Mohamed Mosbah, Akka Zemmari, Margot Quantin and Pauline Gautier
Abstract:	Increasingly frequent heat waves, intense rainfall and strong wind gusts are likely to accelerate degradation of urban tramway assets in large metropolitan areas. This paper proposes a data-driven framework to quantify and predict weather-related failures in the Bordeaux tram network in south-western France. The study integrates multi-year maintenance logs from the operator Keolis Bordeaux Métropole Mobilités with meteorological, operational and geographical datasets to characterise asset condition and climate exposure. After an expert-based criticality assessment of key subsystems, the analysis focuses on switches and crossings, identified as the most climate-sensitive components. Text-mining methods are used to distinguish weather-related from non-weather-related failures and to derive root-cause categories. A machine-learning pipeline combining several classifiers, nested cross-validation and SHAP-based explainability is then developed to model failure occurrence under varying weather and operating conditions. The framework highlights the relative contributions of precipitation, wind, local topography and operational factors to failure risk, and lays the groundwork for a predictive tool to support climate-aware maintenance planning. The best-performing classifier (Random Forest, 24-hour window) achieves an F1-score of 0.7503 under nested cross-validation, demonstrating the framework’s ability to discriminate weather-related from human-factor failures.
Download

Paper Nr:	34
Title:	Socioeconomic Factors and Private Car-Ownership: A Correlation Study in the Helsinki Metropolitan Area
Authors:	Juha Koskela, Flaithrí Neff and Markus Sihvonen
Abstract:	Finland’s goal of achieving carbon neutrality by 2035 requires a substantial reduction in private car dependency, particularly in metropolitan regions where sustainable transport alternatives already exist. This research aims to inform transport policy by examining spatial variation in household car-ownership across the Helsinki Metropolitan Area (HMA) and identifying the socioeconomic factors that explain differences. Household car ownership density was analysed at the postal-code level using 2023 data from Statistics Finland, with four predictors used in analyses: median household income; the proportion of young adults aged 18-34; the share of households with children; and gender composition. Multivariate linear regression, spatial diagnostics, clustering, and cartographic analysis were employed to assess relationships between these predictors and cars-per-household. The resulting model demonstrates high explanatory power (adjusted R² = 0.857) with no evidence of residual spatial autocorrelation. Cluster analysis identified three spatial profiles within the HMA: 1) low-car, young-adult-dominated; 2) high-income, family-oriented suburban areas with high car-ownership; and 3) mixed transitional areas with intermediate characteristics. Overall, the findings show that socioeconomic structure explains a large share of spatial variation in household car-ownership - despite the availability of public transport - highlighting the need for mobility policies that account for differing life-stage constraints and heterogeneous transport needs.
Download

Paper Nr:	71
Title:	Providing Secure Information Exchange for Transportation Management in a Decentralised Platform
Authors:	David Gray Marchant, Tim Clausing, Victor Tvrdy, Arne Lamm, Wonhee Lee, Falk Bethke, Oliver Steensen-Bech Haagh, Michael Kirkedal Thomsen, Kamer Kaya and Cansu Tanrikulu
Abstract:	This paper introduces the Maritime Connectivity Platform (MCP), as a means of achieving a decentralised, secure, and reliable data space both within the maritime domain and beyond. This is achieved by outlining three MCP core components: identity management, trust enforcement, and service discoverability. Several instances of these components can be networked together to provide a robust, decentralised system that is nevertheless secure and provides fine grain control of authentication and authorisation. Due to its decentralised nature the platform is intended as a starting point to be added to by subsequent users, but core services have been provided to ensure an already usable and meaningful data space. This is all intended to support the increasing need and desire for digitisation within the global maritime sector, but also additional transportation sectors such as inland waterways and truck or rail transportation. The system is currently in an advanced stage of development, with demonstrators having been constructed for the core components and several technical specifications already accepted by the relevant governing authorities.
Download

Paper Nr:	75
Title:	Adaptive Multimodal Signal Control Using Real-Time Cyclist Platoon Detection
Authors:	Sarah Salem and Axel Leonhardt
Abstract:	This paper presents a real-time framework for cyclist platoon prioritization at signalized intersections, integrating fuzzy C-means (FCM)–based platoon detection with adaptive fuzzy signal control. Building on prior work that validated the detection approach, this study focuses on its closed-loop integration into signal control and its multimodal operational impacts. Detected platoon attributes: estimated time of arrival (ETA), size, and cohesion, are embedded directly into the controller to enable demand-responsive priority decisions. A demand sweep ranging from 120 to 1,200 cyclists per hour is evaluated in a two-intersection VISSIM test network using COM-based real-time data exchange. The results reveal three distinct operational regimes. Under low demand, the controller remains inactive and converges to fixed-time behavior, ensuring stability and avoiding unnecessary interventions. At medium demand, priority activation increases, producing substantial cyclist delay reductions while introducing moderate vehicle impacts, reflecting a transitional trade-off regime. At high demand, sustained priority operation leads to cyclist delay reductions of up to 40–45%, while vehicle delay impacts stabilize. A sensitivity analysis of the preference threshold highlights its role in governing multimodal trade-offs. Furthermore, incorporating a bounded cohesion term improves control stability and efficiency under higher demand conditions. The proposed approach demonstrates demand responsiveness and robustness, offering a scalable strategy for cyclist-aware adaptive signal control in urban networks.
Download

Paper Nr:	85
Title:	Toward Bus-Based Mobile Environmental Monitoring: A Route Optimization Framework and Sensor Prototype for Astana, Kazakhstan
Authors:	Aigerim Mansurova, Sabina Saleshova, Didar Yedilkhan, Aigul Adamova and Aliya Nugumanova
Abstract:	This paper proposes a bus-based mobile sensing platform for comprehensive urban environmental monitoring, using Astana, Kazakhstan as a case study. The system mounts multi-parameter sensor modules measuring PM2.5, PM10, NO₂, SO₂, O₃, CO, temperature, humidity, and noise on public buses selected through a weighted greedy route optimization algorithm. The algorithm maximizes coverage of points of interest (POIs) categorized as emission sources and human exposure sites over the city's 87-route bus network. Experiments conducted for k = 3, 5, 10, and 15 routes show that k=10 provides the best trade-off between coverage and redundancy, achieving coverage of 42% of the total weighted points of interest, covering 16.6% of the city area, with a moderate overlap of 37% between routes. The proposed system offers a cost-effective and scalable complement to fixed monitoring infrastructure, with direct applicability to other cities facing similar monitoring constraints.
Download

Paper Nr:	54
Title:	A Practical Pipeline for Urban Arterial Simulation Calibration
Authors:	Rahul Sengupta, Ryan Casburn, Tushar Patel, Jeremy Dilmore and Sanjay Ranka
Abstract:	Accurate calibration of microscopic traffic simulation models of urban arterials is essential for developing and testing optimal traffic signal control strategies. This paper presents a semi-automated calibration pipeline for the SUMO (Simulation of Urban Mobility) microscopic traffic simulator, using Automated Traffic Signal Performance Measures (ATSPM) loop detector data with sparse probe trajectory data. We discuss signal timing calibration using Ring-and-Barrier (NEMA) control logic, as well as traffic flow calibration. We demonstrate techniques for extracting behavioral parameters directly from sparse trajectories, including time headway calibration, over-speeding, and lane-changing parameters. We hope this work helps provide a practical guide to calibrating different aspects of a SUMO simulation scenario.
Download

Paper Nr:	56
Title:	Predicting Station-Wise Remaining Arrival Times from GPS-Based Train Trajectories Using a Linear Regression Approach
Authors:	Mohamed Samy Massoum, Bruna G. Palm, Carolina Bergeling, Henrik Fredriksson and Mattias Dahl
Abstract:	This paper presents a methodology for predicting remaining train arrival times along a predefined railway path using sectionally sampled GPS-based observations collected between consecutive stations. Station-level remaining arrival times are derived from multiple train trips and analyzed using a linear regression model, enabling assessment of both prediction accuracy and temporal consistency. The approach provides continuous predictions of remaining arrival times at each station along the path rather than focusing solely on the final destination. Results demonstrate that median remaining arrival time errors decrease with increasing sampling density, an effect attributed to greater sample availability that captures a larger proportion of the underlying train dynamics and timetable adherence. Under these conditions, the proposed methodology achieves a station-based average median remaining arrival time error of less than 91 seconds. Across 100 repeated out-of-sample evaluations, the results indicate reproducible performance and acceptable coarseness even when the sampling density is substantially reduced. This continuous, station-level prediction capability enables early detection of delays and potential timetable conflicts, providing minute-level predictive accuracy suitable for decision support for train dispatchers.
Download

Paper Nr:	88
Title:	Overcoming Standard Lock-in in C-ITS through Profile-Driven Governance
Authors:	Tanja Pavleska, Massimiliano Masi, Giovanni Paolo Sellitto and Helder Aranha
Abstract:	This paper introduces a methodology to support automated governance in complex, standards-based system architectures. In many critical sectors, long-term interoperability is hindered by the absence of governance frameworks that span the lifecycle of evolving standards. Static compliance alone often leads to standard lock-in, increased management overhead, and fragmentation across stakeholders. To address this, we propose an architecture-driven approach inspired by practices from domains such as healthcare, energy, and Industry 4.0. The methodology builds on the IHE architectural model, which enables modular system design and embeds governance rules directly into machine-readable architectural specifications. By organizing systems into interoperable building blocks with well-defined interfaces, the approach isolates changes and facilitates automated verification of compliance. The methodology is demonstrated through a proof-of-concept use case in the Cooperative Intelligent Transport Systems (C-ITS) sector, illustrating its applicability to real-world, multi-stakeholder environments. This method offers a scalable and sustainable path for maintaining interoperability over time, reducing reliance on manual coordination and enabling consistent policy enforcement across diverse systems and vendors.
Download

Area 3 - Intelligent Vehicle Technologies

Full Papers

Paper Nr:	13
Title:	Parking-Substitute Cruising Strategy Optimization: Reducing Congestion with Deep Reinforcement Learning
Authors:	Chenhe Liu and Farhana Choudhury
Abstract:	The growing demand for parking infrastructure consumes valuable land and significant expenditure. Autonomous vehicles can mitigate this by engaging in empty cruising as a substitute for parking (parking-substitute cruising), though this may increase congestion. This study proposes an algorithm-driven approach, an independent Deep Q-Network (IDQN), to optimize the cruising route of each autonomous vehicle (AV). Unlike conventional methods relying on tolling or centralized traffic control, our vehicle-based model enables each AV to plan its own route, reducing dependence on top-down management and enhancing flexibility and personalization. The method optimizes parking-substitute cruising without discouraging its use, as it guides route planning rather than imposing speed limits. Trained on a Parkville (Melbourne) network using real-world trip distribution data, the learned policy outperforms both greedy and random baselines. Finally, we offer theoretical insights to help decision makers balance congestion and lateness risk through congestion-related parameters.
Download

Paper Nr:	35
Title:	Improving EKF Consistency in 3D Multi-Object Tracking via Heteroscedastic Detection Uncertainty
Authors:	Cornelius Schröder, Felix Fent and Markus Lienkamp
Abstract:	Reliable uncertainty estimates are essential for safe autonomous driving, as perception, tracking, prediction, and planning form a tightly coupled pipeline in which overconfident intermediate outputs can lead to unsafe decisions. While prior work incorporates detection uncertainty into multi-object tracking, the resulting state covariance estimates are often insufficiently calibrated for downstream use. We extend CenterPoint to estimate heteroscedastic aleatoric uncertainty and, optionally, epistemic uncertainty via a Multiple-Input Multiple-Output formulation without sampling-based inference. Predicted per-detection variances are integrated into an Extended Kalman Filter by constructing time-varying measurement covariances without any further processing. Experiments on the nuScenes dataset show that this straight-forward approach to heteroscedastic uncertainty propagation substantially improves the calibration of tracking output uncertainties, while tracking accuracy and precision improve modestly when detector-level velocity estimates are available. These results highlight that the primary benefit of uncertainty-aware tracking lies in improved filter consistency, yielding state covariance estimates that are more informative for downstream prediction and planning.
Download

Paper Nr:	48
Title:	Enhancing Functional Reliability of Autonomous Vehicle Safety Monitoring in Curves
Authors:	Junnan Pan, Mohak Mansharamani, Prodromos Sotiriadis, Vladislav Nenchev and Ferdinand Englberger
Abstract:	Active safety systems that operate alongside a higher-level autonomy stack and perform autonomous braking require reliable obstacle detection within the vehicle’s motion corridor. Many runtime-constrained monitoring approaches simplify road geometry, producing blind spots on high-curvature roads where obstacles could not be detected. In this paper, we eliminate these blind spots by aligning the monitoring region with a locally framed, predicted motion corridor. We compute a predicted corridor boundary and filter incoming lidar point clouds in real time using an azimuth-sorted visible boundary and a moving window. In a case study, our method avoids curvature-induced blind spots while maintaining performance on straight segments.
Download

Paper Nr:	51
Title:	Towards Uncertainty-Calibrated Traffic Flow Prediction Using Combinatorial Graph Neural Networks
Authors:	P. Kirthan, M. Anbazhagan, Radha Reddy and Harrison Kurunathan
Abstract:	Urban traffic prediction is a challenging task, as real-world sensor streams often contain noise, incomplete measurements, or failures that affect model stability. In this regard, we introduce a spatio-temporal Graph Neural Network (GNN) framework that explicitly models uncertainty. The architecture combines diffusion-style graph convolutions to model multi-hop spatial interactions, WaveNet-inspired dilated convolutions to capture temporal patterns across multiple scales, and a compact bidirectional Long Short-Term Memory (BiLSTM) to aggregate long-range dependencies. Uncertainty is quantified using Monte Carlo dropout with two output branches, enabling separate estimation of the aleatoric and epistemic components and yielding well-calibrated confidence intervals alongside standard point predictions. The proposed approach achieves an MAE of 0.439 and an RMSE of 1.03 on the benchmark California Performance Measurement System - Bay Area Traffic Dataset (PEMS-BAY), with an aleatoric uncertainty of 0.550 and an epistemic uncertainty of 0.256. It outperforms Spatio-Temporal Graph Convolutional Network (ST-GCN), Diffusion Convolutional Recurrent Neural Network (DCRNN), and Graph WaveNet, especially under sensor-dropout testing conditions, thereby demonstrating suitability and resilience for real-time intelligent transportation applications.
Download

Paper Nr:	64
Title:	Effect of Hitch Angle Error Control on the Path Tracking Performance of an Articulated Vehicle Using a Linear Quadratic Integral Controller
Authors:	Mike Haemers and Taehyun Shim
Abstract:	This paper presents the development of an LQI-based path tracking controller for a vehicle–trailer application, in which a single-track vehicle–trailer model and three error states (lateral position, heading, and hitch angle) are used in the control system. The performance of this controller is compared with an LQI-based tracking controller that considers only vehicle dynamics and two error states (lateral position and heading). The tracking performance of the vehicle–trailer model with these two controllers was evaluated on straight and curved roads through simulations. The simulation results indicate that the conventional lane-keeping controller, developed for vehicle lateral position and heading control, can handle vehicle–trailer tracking under normal operating conditions with known vehicle parameters. The controller that considers trailer dynamics and hitch angle control performs more robustly under parameter uncertainty compared to the vehicle-only lane-keeping controller.
Download

Paper Nr:	65
Title:	Measuring Braking Behavior Using Vehicle Tracking and Camera-to-Satellite Homography Rectification
Authors:	J. P. Fleischer, Tanchanok Sirikanchittavon, Chonlachart Jeenprasom, Nooshin Yousefzadeh, Sanjay Ranka and Mohammed Hadi
Abstract:	This paper presents an open-source software application for analyzing traffic camera footage, focusing on vehicle behavior and braking events at signalized urban highways. The core innovation is a robust ground-plane homography estimation that links fixed traffic camera views to satellite orthoimagery. This process rectifies the camera’s oblique perspective, ensuring that pixel distances accurately represent real-world distances. This enables the acquisition of features such as vehicle trajectory, speed, deceleration, and braking severity without the need for camera calibration. The pipeline employs the MAGSAC++ estimator to build the homography, converting YOLO11 object detections into a rectified top-down coordinate system. All detection and trajectory data are stored in a ClickHouse database for subsequent analysis. A real-world case study at two signalized intersections in Key West, Florida, showcased the system’s capabilities. Across two days of daytime footage, braking activity at the higher-volume intersection peaked around 4 PM at approximately 57.5 events per hour, while the second intersection peaked around 10 AM at roughly 15.5 events per hour. The spatial analysis revealed that most braking events initiated upstream, with mild and moderate braking mostly occurring 30–45+ meters away from the stop bar and severe braking distributed throughout, but particularly concentrated in lanes with higher interaction and merging activity. The findings highlight the significant potential of this centralized safety information system to support connected vehicles, facilitating proactive traffic management, crash mitigation, and data-driven roadway design and safety analysis.
Download

Paper Nr:	66
Title:	Vision-Language Integration for Safe and Natural In-Cabin Interaction in the Driverless Minibus
Authors:	Qazi Hamza Jan, Dawood Khan and Karsten Berns
Abstract:	Driverless minibuses are increasingly deployed in pedestrian-priority and traffic-calmed environments, where safe and intelligible interaction with passengers is critical for trust and acceptance. While existing research has primarily focused on external interaction with pedestrians, comparatively little attention has been given to in-cabin interaction between autonomous vehicles and their passengers. In the absence of a human driver, passengers lack situational explanations and safety guidance, which can lead to uncertainty, stress, and unsafe behavior such as standing while the vehicle is in motion. This paper proposes a vision–language–based in-cabin interaction framework that integrates real-time cabin perception with large language models (LLMs) to generate context-aware safety instructions. An in-cabin camera is used for understanding the scenario, while vehicle motion states are obtained from the autonomous driving system. These inputs are encoded into a structured prompt that enables the LLM to assess risk levels and produce short, constrained spoken commands aligned with predefined safety policies. The LLM operates as a controlled language generation module within a safety-constrained interaction pipeline, including rule-based filtering and fallback mechanisms. The system is evaluated on a driverless minibus platform using representative safety-critical and preventive scenarios, such as standing passengers during driving and emotionally agitated behavior. Experimental results demonstrate that the proposed multimodal prompting approach can generate appropriate, concise, and risk-adaptive instructions grounded in both vehicle dynamics and passenger behavior. The findings indicate that vision–language integration can enhance passenger safety, transparency, and user experience in autonomous public transport.
Download

Paper Nr:	67
Title:	Integrating SiL, HiL and OTA Sensors: A Testbed Architecture for Modular Software-Defined Vehicles
Authors:	Niklas Rahenbrock
Abstract:	The increasing complexity of automotive software, combined with the demand for shorter development cycles, presents significant challenges in the development and validation of safety-critical functions, particularly in Software-Defined Vehicles (SDVs). This paper investigates how a simulation and test environment can support the modular development and early-stage evaluation of software systems across various stages of an SDV lifecycle. To address this, a flexible test environment is proposed that enables a stepwise transition from fully virtual Software-in-the-Loop (SiL) testing to more realistic setups using Hardware-in-the-Loop (HiL) and Over-The-Air (OTA) sensors. The environment integrates standardised interfaces such as the Open-Simulation-Interface (OSI) and a message-oriented middleware to ensure modularity and scalability. Central components include an adapter module for seamless integration of the System under Test (SuT) and a component for automated deployment and testing. A proof-of-concept implementation demonstrates the feasibility of hybrid simulation setups and automated test execution. The results show reduced integration effort, seamless migration between SiL and HiL, and effective validation with both simulated and physical sensors, enabling early-stage evaluation of modular SDV.
Download

Paper Nr:	74
Title:	Uncertainty Evaluation to Support Safety of the Intended Functionality Analysis for Identifying Performance Insufficiencies in ML-Based LiDAR Object Detection
Authors:	Milin Patel and Rolf Jung
Abstract:	Machine Learning (ML)-based LiDAR 3D object detectors in automated driving produce false detections, missed detections, and localisation errors under adverse weather and reduced visibility. Detection errors arising without hardware or software faults constitute performance insufficiencies under ISO 21448, Safety of the Intended Functionality (SOTIF), and the standard requires identification of the triggering conditions responsible. The prescribed analysis methods assume a design specification, but ML-based LiDAR object detectors have no design specification because the mapping from point clouds to bounding boxes is learned from training data. This paper proposes an uncertainty evaluation methodology that uses disagreement among deep ensemble members to separate correct from incorrect detections. Ensemble disagreement and performance insufficiencies arise from insufficient training data coverage of the operating condition. The methodology evaluates whether three uncertainty indicators derived from ensemble disagreement (mean confidence, confidence variance, and geometric disagreement) separate correct from incorrect detections. The evaluation produces outputs mapped to ISO 21448 analysis activities: discrimination metrics, triggering condition rankings by false positive share, frames flagged for investigation, and acceptance gates reporting coverage and false acceptance rate. A case study using simulated ensemble predictions across 22 weather configurations shows that geometric disagreement achieves the strongest separation, with acceptance gates that retain only true detections at reduced coverage. The observed separation arises because false detections produce spatially inconsistent bounding boxes across ensemble members where no physical object constrains the predicted position, while true detections remain spatially consistent.
Download

Paper Nr:	78
Title:	BAM-ACS: Bayesian Adaptive Mixtures with Adaptive Covariance Scaling for High-Integrity Localization for Autonomous Vehicle
Authors:	Elias Maharmeh, Paulo Resende and Fawzi Nashashibi
Abstract:	High-integrity localization is essential for autonomous vehicles operating in safety-critical environments. Conventional integrity monitoring methods rely on internal localization system covariance, which often becomes overconfident during low observability. In this paper, we propose BAM-ACS, a Bayesian Adaptive Mixture with Adaptive Covariance Scaling for high-integrity localization. The localization system employs tightly-coupled LiDAR-inertial odometry using an error-state extended Kalman filter, fused with GNSS measurements in a loosely-coupled architecture for global drift correction. For integrity monitoring, we extract two physical observables: kinematic discrepancy from the IMU and the filter-estimated state inconsistency and geometric discrepancy from scan-matching quality. These drive a Bayesian mixture model that probabilistically switches between nominal and fault modes, adaptively scaling uncertainty during dead-reckoning. A novel adaptive update prevents GNSS measurements from producing false confidence by inflating covariance when innovations are inconsistent. Validation on three UrbanNavDataset environments confirms our method’s effectiveness. BAM-ACS achieves zero integrity risk (lateral and longitudinal) across all scenarios, outperforming the baseline: medium-urban (13.2% long., 16.5% lat.), deep-urban (12% long., 17.1% lat.), and low-urban (90.2% long., 23.6% lat.). The algorithm runs in real time without latency.
Download

Paper Nr:	84
Title:	Geometry-Consistent Vectorized HD Map Learning: Cross-Domain Validation from nuScenes to CARLA
Authors:	Iyad Abuhadrous, Fawzi Nashashibi and Benazouz Bradai
Abstract:	Vectorized HD map learning enables direct prediction of structured lane geometries from multi-camera inputs and has become a key component of modern autonomous driving systems. Recent approaches such as MapTR demonstrate strong performance on real-world datasets like nuScenes, but their robustness under cross-domain deployment remains insufficiently understood. This work investigates the behavior of vectorized BEV map models when transferred from real-world data to a synthetic driving environment. Although sensor metadata and coordinate formats appear compatible, direct deployment from nuScenes to CARLA reveals systematic structural degradation, including polyline fragmentation, curvature discontinuities, and spatial misalignment in bird’s-eye-view (BEV) space. Through a detailed audit of the projection chain and coordinate transformations, we identify inconsistencies in geometric conventions as a key source of instability. To address this issue, we introduce a geometry-consistent CARLA pipeline that enforces explicit coordinate contracts across intrinsics, extrinsics, and BEV transformations, and generates MapTR-compatible vector annotations with structured training and evaluation splits. Experimental results show that geometry-consistent fine-tuning substantially improves cross-domain performance, restoring structural coherence and reducing geometric error. These findings demonstrate that validating the geometric projection pipeline is critical for reliable cross-domain deployment of vectorized HD map models, and provide a reproducible foundation for BEV-based map learning across heterogeneous driving environments.
Download

Short Papers

Paper Nr:	37
Title:	Flexible Mobile Test Infrastructure for Automated Driving, V2X Applications, and Scenario-Based Testing
Authors:	Kai Schultz, Martina Emde-Rajaratnam, Chirag Modi, Daniel Rau, Philipp Schorr, Jonas Vogt and Horst Wieker
Abstract:	Connected mobility offers a lot of advantages, it helps making traffic flow smoother, therefore more comfortable and environmentally friendly, and it is able to drastically increase safety for all road users. Especially with regard to automated driving it offers huge benefits, mainly by significantly increasing the field of view. A central aspect in this context are infrastructure-mounted sensors providing reliable information of safety critical areas. On the one hand, as these technologies are currently emerging, only marginal parts of the infrastructure are equipped with the respective technology. On the other hand, in order to accelerate the roll out of connected (automated) mobility, different use cases and scenarios have to be developed and tested. This work presents the development and construction of a mobile ITS-testing-station. Requirements were gathered, and based on this, a simple box trailer with a telescopic mast was converted to an autonomous, weatherproof, and highly flexible (with respect both to usage and setup) tool for testing infrastructure-based connected use cases and scenarios, especially when certain sensor configurations are needed. It is also shown, how the system is used in a project demonstration of an end-of-traffic-jam use case, proving all the before-mentioned advantages.
Download

Paper Nr:	39
Title:	Reliability of 3D Vehicle Damage Analysis in Determining Equivalent Energy Speed
Authors:	Pavlína Moravcová, Michal Křižák, Albert Bradáč and Kateřina Bucsuházy
Abstract:	Three-dimensional (3D) methods have significantly changed the way crashes and vehicle damage are documented in recent years. The study evaluates the accuracy and usability of 3D vehicle damage analysis for quantifying deformation energy, respectively equivalent energy speed (EES). The results of the 3D analysis were compared with reference EES values from crash tests and traditional 2D analysis (calculated using the CRASH3 algorithm). The study evaluates the strengths and limitations of 3D documentation, since the results indicate that both methods (2D and 3D analysis) are capable of providing accurate EES estimates. Within the analyzed dataset, most of the obtained deviations using the 2D method were within ±10%. Although 3D techniques generally provide highly detailed geometric data, the EES value may be underestimated by up to approximately 30% in cases where deformation is partially hidden by structural vehicle components or when the point cloud is insufficiently captured. The obtained data confirm that 3D damage analysis is a valuable tool for reconstructing traffic accidents, with the completeness and quality of capturing the actual damage being key to reliably determining the EES value.
Download

Paper Nr:	41
Title:	Improving Accuracy and Efficiency in HiL Testing for Modern Software-Centric Vehicle Platforms through Automated Script Generation
Authors:	Ryan James Lynch, Morteza Soleimani, Paul Weddell and Sulakshan Rajendran
Abstract:	The shift towards software-centric vehicle architectures demands scalable, efficient, and reliable Hardware-in-the-Loop (HiL) testing methodologies to validate increasingly complex active safety and automated driving functions. This paper presents and tests an automated method for creating closed-loop HiL test scripts, using a representative longitudinal control feature as a case study. The method uses Gherkin Behaviour Driven Development (BDD) requirements as executable specifications, processed via a Signal Parser Algorithm to generate signal abstraction files without manual scripting. These files are then run on a dSPACE HiL platform using Automation Desk, allowing large numbers of tests to be carried out without human supervision. In experiments, the automated approach reached 97.814% accuracy for variable path assignment and improved test success rates from 72.727% to 84.211% compared to manual scripting. Over 16,000 tests were completed in 19 hours, showing major gains in speed, coverage, and repeatability. The results show that this method provides enough realism for Software Component (SWC) validation while reducing engineering effort and enabling earlier defect detection. This work addresses key gaps in current HiL practices and gives vehicle manufacturers a practical way to save time, reduce costs, and shift testing earlier in the software development process.
Download

Paper Nr:	43
Title:	Driving through the Network: Performance and Workload under Latency and Video Impairments
Authors:	Ines Trautmannsheimer, Ahmed Azab and Frank Diermeyer
Abstract:	Teleoperation can extend the operational envelope of automated vehicles, but its effectiveness depends on communication latency and video quality. While both factors have been studied individually, their combined effects on operator workload and driving performance remain insufficiently understood. We conducted a fixed-base driving-simulator study (N = 25) using a 2×2 design with added latency (100/300 ms) and bitrate (500/2000 kbit/s), complemented by a best-case baseline. Effective roundtrip latency was measured for all conditions, revealing a baseline of approximately 413 ms and total delays of about 500–700 ms in the experimental cells. Multimodal measures captured driving performance, oculomotor behavior, physiology, and subjective workload. Both higher latency and lower bitrate increased operator load and modestly affected performance. Physiological measures, particularly heart rate and RR interval, were the most sensitive and showed sub-additive interaction effects, whereas interactions in performance and oculomotor measures were small or absent. Equivalence tests further indicated that the 300 ms/2000 kbit/s condition remained velocity-equivalent to the best-case baseline, whereas 300 ms/500 kbit/s did not. The findings suggest that latency and video quality can be treated as largely independent design levers and that physiological monitoring may support adaptive teleoperation interfaces.
Download

Paper Nr:	44
Title:	How Much LiDAR Field of View Is Enough? A Robustness and Sensor Placement Analysis for LiDAR-Based Localization
Authors:	Nijinshan Karunainayagam, Dominik Kulmer and Frank Diermeyer
Abstract:	LiDAR localization for autonomous vehicles (AV) is crucial for safe driving, especially where GPS is unreliable. Nevertheless, environmental conditions, mechanical damages and package drops may affect the performance of LiDAR sensors and thus the performance of the AV system. Additionally, complete LiDAR failures, especially in single-LiDAR setups might decrease the functional performance of the AV. Particularly, modules directly affected by the quality of the received point cloud like the localization might lose its functional performance. This can impact the safety of an AV enormously. To tackle this, this paper investigates how directional LiDAR field-of-view occlusions affect the robustness of LiDAR-based localization. For this, cropped point clouds with varying cropping opening angles and orientation are fed into a localization algorithm while evaluating its performance with selected metrics. Rather than focusing solely on accuracy degradation, both localization failure probability and distance-to-failure onset is analyzed additionally. The analysis provide actionable guidance for LiDAR sensor placement, redundancy and fault monitoring in safety-critical localization systems.
Download

Paper Nr:	50
Title:	openHDMap: High-Resolution Maps Created from Open Data for Self-Localization of Autonomous Vehicles
Authors:	Christoph Brückner and Lucila Patino-Studencki
Abstract:	Autonomous vehicles rely on accurate positioning to operate safely. LiDAR sensors can support this process through geo-referenced scan-to-map matching. One of the principal challenges in this type of localization is the efficient generation of high-definition point cloud data set (PCD) maps. This paper presents a pipeline for generating PCD maps with configurable and homogeneous point densities by integrating publicly available terrain, building, and vegetation models into a unified 3D representation. Evaluation results show consistent point distributions across different tiles and voxel resolutions, demonstrating the robustness of the resampling and map generation process. Quantitative analyses further indicate that the generated PCD maps capture the dominant static structures of the environment with sufficient geometric fidelity for robotics applications. The resulting maps are well suited for LiDAR-based self-localization in autonomous driving scenarios while relying solely on public data, reducing deployment costs.
Download

Paper Nr:	53
Title:	A Survey of AI Applications and Their Security Challenges for Autonomous Driving
Authors:	Leon Ramirez, Marco Leeske and Christian Plappert
Abstract:	Artificial Intelligence (AI) is increasingly being integrated into modern vehicles, enabling a wide range of new business models and advanced functionalities such as automated perception, localization, decision-making, and control. These AI-driven capabilities form the technological foundation of autonomous and connected vehicles. However, the adoption of AI also significantly increases system complexity and introduces novel cybersecurity threats, including adversarial attacks, data poisoning, and model manipulation across different architectural layers. In this paper, we provide a structured overview of AI-powered automotive use cases along the autonomous driving stack, ranging from perception and localization to decision-making and low-level control. Furthermore, we analyze the resulting security challenges and present state-of-the-art mechanisms to enhance the robustness and security of AI-based automotive systems, such as adversarial defenses, safety shields, and cross-layer verification techniques. Finally, we evaluate these mechanisms using a set of defined evaluation criteria, enabling a systematic comparison of their effectiveness, limitations, and applicability in safety-critical automotive environments.
Download

Paper Nr:	58
Title:	Safety-Savvy Intersection Control Harnessing Vision Language Model
Authors:	Abolfazl Afshari, Joyoung Lee and Yousef Mashal
Abstract:	Existing traffic signal control methods often rely on low-level numerical inputs or end-to-end optimization, limiting interpretability for safety-critical decisions. This study presents a vision-language-model-guided, event-driven traffic signal safety intervention framework implemented in a high-fidelity CARLA world model. Fixed cameras provide intersection views, and a fine-tuned vision-language model generates structured semantic outputs and bounded safety recommendations. A rule-constrained controller applies only conflict-free interventions, such as short pre-green holds or temporary all-red extensions, while preserving a fixed two-phase signal plan and immutable clearance times. Experiments under low, medium, and high demand show consistent reductions in pedestrian exposure relative to fixed-time control, with stable structured output and bounded inference latency. The results suggest that vision-language reasoning can serve as an interpretable semantic layer for safety-oriented signal intervention.
Download

Paper Nr:	63
Title:	Enactor: From Traffic Simulators to Surrogate World Models
Authors:	Yash Ranjan, Rahul Sengupta, Anand Rangarajan and Sanjay Ranka
Abstract:	Traffic microsimulators such as SUMO are widely used to evaluate road network performance under various “what-if” conditions. However, the behavior models controlling the actions of the actors are overly simplistic and fails to capture realistic actor-actor interactions and its impact on an actors behavior. Deep learning-based methods have been applied to model vehicles and pedestrians as “agents” responding to their surrounding “environment” (including lanes, signals, and neighboring agents). Although effective in learning actor-actor interaction, these approaches fail to generate physically consistent trajectories over long time periods, and they do not explicitly address the complex dynamics that arise at traffic intersections which is a critical location in urban networks. Inspired by the World Model paradigm, we have developed an actor centric generative model using transformer-based architecture that is able to capture the actor-actor interaction, at the same time understanding the geometry to the traffic intersection to generate physically grounded trajectories that are based on learned behavior from the data. Moreover, we test the model in a live “simulation-in-the-loop” setting, where we generate the initial conditions of the actors using SUMO and then let the model control the dynamics of the actors. We let the simulation run for 40000 timesteps (4000 seconds), testing the performance of the model on long timerange and evaluating the trajectories on traffic engineering related metrics. Experimental results demonstrate that the proposed framework effectively captures complex actor–actor interactions and generates long-horizon, physically consistent trajectories, while requiring significantly fewer training samples than traditional agent-centric generative approaches. Our model is able to outperform the baseline in most of the metrics as well as aggregate speed and travel-time metrics where our model beats the baseline by more than 10x on the KL-Divergence.
Download

Paper Nr:	73
Title:	A Reduced-Order Model-Driven Augmented Reality Digital Twin for Real-Time Vehicle Chassis Displacement Monitoring
Authors:	Alessio Cellupica, Mariagrazia Tristano, Marco Cirelli, Pier Paolo Valentini and Frank Naets
Abstract:	Timely acquisition of vehicle data is fundamental to improve passenger safety and monitor the operating conditions of the vehicle itself. However, offline data processing does not allow real-time intervention upon detection of a structural failure. This paper presents a sensor-driven reduced-order digital twin for real-time estimation of vehicle chassis displacements integrated with an augmented reality visualization layer. In the proposed framework, inertial measurements acquired from an onboard inertial measurement unit are processed to derive representative loading conditions through rigid-body dynamics and simplified vehicle modeling assumptions. The resulting displacement field is reconstructed in real time through a superposition-based formulation and transmitted to an augmented reality headset via a client–server architecture. A subscription–notification communication mechanism ensures low-latency data exchange and continuous synchronization between the physical vehicle and its digital counterpart. The framework provides a practical implementation of real-time digital twin technology for structural monitoring in automotive applications.
Download

Paper Nr:	82
Title:	Integrated Speed Harmonization and Perimeter Control for Congestion Mitigation
Authors:	Maha Elouni, Hesham A. Rakha, Monica Menendez and Hossam M. Abdelghaffar
Abstract:	Traffic congestion leads to increased travel times and frequent delays in urban transportation networks. This paper introduces a vehicle-centric approach for implementing perimeter controller. Specifically, a sliding mode perimeter controller was combined with speed harmonization to regulate the speed of vehicles entering a protected region, an area that must be protected from congestion. The developed controller was applied and evaluated on a medium-sized grid network inspired by downtown Washington, DC. Simulation results show that the developed controller reduced network-wide travel time and delay by 17.3% and 18.2%, respectively. It also reduced the network-wide total fuel consumption and CO2 emissions by 5.91% and 7.04%, respectively. These results demonstrate the effectiveness of the proposed control strategy in mitigating traffic congestion, improving overall traffic flow within urban networks and enhancing fuel efficiency and environmental sustainability.
Download

Paper Nr:	86
Title:	A Real-Time Vehicle-in-the-Loop Platform with Cooperative UAV Edge Computing and Passenger-in-the-Loop Integration
Authors:	G. Pepe, E. Pavanato, M. Laurenza, D. Spitaleri, E. M. Di Pietro, S. Milana and A. Carcaterra
Abstract:	To enable safe and scalable autonomous driving experimentation, this paper presents a novel Vehicle-in-the-Loop framework coupling a physical test vehicle with a high-fidelity, georeferenced digital environment for realistic closed-loop experimentation. The platform integrates three core components: a physical test vehicle with robotic actuation; a deterministic real-time control unit executing Behavioral Planning and Nonlinear Model Predictive Control (NMPC); and a high-fidelity virtual environment rendered in Unreal Engine, enabling interaction with dynamically generated scenarios while preserving physical consistency. A key contribution is a cooperative perception layer based on an Unmanned Aerial Vehicle (UAV), operating as an external sensing and edge-computing node. By processing visual data in real time, the UAV extends situational awareness beyond onboard sensor limitations. The platform also supports a Passenger-in-the-Loop configuration, combining real vehicle dynamics with immersive virtual reality for safe evaluation of human–machine interaction during autonomous maneuvers. Results demonstrate accurate trajectory tracking, reliable real-time performance of the NMPC controller, and consistent operation of the distributed perception pipeline under strict timing constraints.
Download

Paper Nr:	16
Title:	Traffic Light Control Model Using Reinforcement Learning for Traffic Optimization on Javier Prado Avenue, Lima
Authors:	Gonzalo Jurado, Marc Diaz and Rosa Felix
Abstract:	Traffic congestion in large urban corridors remains a persistent challenge, often aggravated by rigid signal plans that cannot adapt to real-time conditions. In the past, the solution to this problem has revolved with expanding roads, creating bypass or adding traffic lights, but those cannot adapt to the ever increasing amount of vehicles that use them. With smart traffic lights and artificial intelligence, dynamic solutions can be implemented to handle traffic flow, but in countries like Peru the speed that they get implemented is very slow. This paper presents an adaptive traffic-signal control model based on Reinforcement Learning (RL), implemented with Proximal Policy Optimization (PPO) in a 3D Unity simulation of Javier Prado Avenue in Lima, Peru. The environment incorporates randomized initial conditions, predefined shortest-path routes to emulate realistic vehicle dispersion and independent agent in each intersection. Performance was assessed through a synchronized A/B evaluation against a fixed-time baseline. Across the shared temporal window and evaluated paths, the RL-based configuration achieved reductions of 7.67% in total vehicle count and 3.50% in accumulated delay, along with a 5.26% increase in weighted average speed. These results indicate that reinforcement-learning-driven traffic control can enhance mobility efficiency and represents a promising direction for scalable intelligent transportation systems in Latin America.
Download

Paper Nr:	26
Title:	Achieving Reliable U-Shift Capsule Coupling: A Comparative Study of Geometric Steering Controllers and Safe Operating Zone Determination
Authors:	Raghava Santhan Mysore Pavan
Abstract:	This paper presents a simulation-based comparative study of two geometric controllers, Pure Pursuit and Stanley, applied to the coupling process of an Autonomous Modular Vehicle (AMV) known as U-Shift. The maneuver involves the diveboard reversing to dock and connecting with a transport capsule. The Hybrid A* planner considers driveboard and capsule geometric constraints to generate the reference path. The study compares the controllers’ performance and shows how changes in parameters and alignment between the driveboard and capsule affect coupling success rates without collisions. The paper further categorized the initial position and orientation of the diveboard into safe zones based on the mean lateral deviation from the reference path and proximity to collision boundaries. Based on the zones, the system can determine if a correctional maneuver is needed to bring the driveboard into a low-risk coupling state. The simulation results guide in designing coupling strategies to achieve more reliable AMV operations.
Download

Paper Nr:	38
Title:	Design of a Software Tool for Optimization of Road Reference Heading (RRH) for a Lane Departure Warning System
Authors:	Ethan Levi Ellison, Attiq Uz Zaman and M. I. Hayee
Abstract:	Lane Departure Warning Systems (LDWS) rely on accurate road trajectory estimates to correctly detect unintentional lane departures. Traditionally, image or sensor-based lane detection are used for LDWS. Performance with this type of LDWS implementation has been observed to drop significantly when lane markings are obscured. To address this limitation, this research focuses on optimizing Road Reference Heading (RRH) generation for the Global Positioning System (GPS) based LDWS previously designed by our group. An optimization tool was developed to visualize heading, differential heading, and the Lane Departure Detection (LDD) Algorithm outputs across multiple data sets. Using this tool enables the optimization of RRH through parameter adjustments e.g., smoothing window size, differential heading window size, and the curve identification threshold which can be adjusted to optimize section classification within RRH generation. This tool was used to optimize RRH on a different route in both directions and field testing was performed for evaluation. The field test results demonstrated increased reduction of false alerts while staying consistent with true lateral shift detection, validating the effectiveness of the optimization tool. The resulting technique supports a more adaptable approach to RRH generation, providing improved performance across various road and driving conditions, leading to zero false alerts.
Download

Paper Nr:	42
Title:	A Data-Driven Approach to Predict Fuel Rail Pressure Anomalies in Internal Combustion Engines
Authors:	Harleen Kaur Bagga, Mukund B. Nagare, Bhushan D. Patil, Hariharan Ravishankar, Vikram Melapudi and Abhijit Patil
Abstract:	Maintaining Optimal fuel rail pressure (FRP) control is essential in common rail diesel engines. It ensures efficient combustion, good fuel economy, and low emissions. Conventional Engine Control Unit (ECU) threshold-based detection methods often fail to identify subtle, early deterioration in FRP, resulting in delayed fault diagnosis and increased operational impact for fleet operators. This paper proposes a multi-parametric data-driven approach for the real-time detection of FRP anomalies under dynamic, real-world driving conditions. Real-time, high-resolution sensor data from the On-Board Diagnostics (OBD) interface is used to train predictive machine learning models. A Gaussian Mixture Model (GMM)-based outlier detection framework is developed and rigorously benchmarked against a baseline Linear Regression (LR) model trained on healthy data. GMM is found to be the most effective method for identifying common rail pressure anomalies. The approach has been validated on a fleet of 63 heavy-duty diesel vehicles with engine displacements of 12000–15000 cc and power outputs of 400–600 hp. Diagnostic Trouble Codes (DTCs) from the ECU have been used as ground truth for evaluating anomaly detection performance. The proposed GMM-based model achieves an average accuracy of 80% in forecasting FRP issues 3–10 days in advance, representing a 45% improvement in accuracy over LR. The proposed predictive framework enables enables proactive maintenance of common rail fuel systems.
Download

Paper Nr:	60
Title:	How to Assist the Perception of Automated Vehicles Remotely: An Extension and Proof-of-Concept of Perception Modification
Authors:	Tobias Kerbl and Frank Diermeyer
Abstract:	Despite the increasing maturity of automated driving technology, automated vehicles still encounter complex scenarios that they cannot resolve and leave them stuck. Such situations frequently arise from an incorrect perception of the driving environment. In such scenarios, teleoperation is widely considered a viable fallback solution to ensure mission completion. To specifically address false-positive object detections, the concept of perception modification was introduced. This work extends the concept beyond this use case to support the various perception tasks of automated vehicles. Based on an analysis of the perception task and existing perception-level assistance approaches, a set of viable modification modes is derived and incorporated into the extended perception modification concept. A proof-of-concept implementation assisting the perception module of Autoware is validated in closed-loop CARLA simulation and in real-world tests using the EDGAR research vehicle.
Download

Paper Nr:	68
Title:	A Framework for Integrated Behaviour Driven Development and Model Based Feature Design and Verification
Authors:	Sandeep Sandhu, Sulakshan Rajendran, Fernando Sarracini Jr and Morteza Soleimani
Abstract:	The rapid evolution of Software Defined Vehicles (SDVs) has shifted automotive innovation toward agile, software intensive features. These features are often distributed across zonal controllers or domain controllers and involve million lines of code. This makes traditional systems engineering and manual testing insufficient and impossible to test the features comprehensively. This paper proposes a new integrated framework which combines both the Behavior Driven Development (BDD) and Model Based System Engineering (MBSE) methodologies via multi loops to bridge the semantic gap between stakeholder and technical implementation following systems engineering V-cycle. By utilising Gherkin-based machine executable specifications with Behavioural Programming to drive formal MBSE Systems Modeling Language (SysML) behaviour models, the framework enables continuous verification upfront and throughout the V-cycle. This enhances requirements/specifications correctness and robustness within the feature software development cycle. The framework is evaluated using an automotive Remote keyless Entry (RKE) case study. Results demonstrate framework implementation and simulation of requirements and thus reducing test and development time.
Download

Paper Nr:	80
Title:	Local Motion Planning Based on Parallel Graph Search
Authors:	Rebecca Richter, Paul Roetzer, Thomas Rottmann, Vivien Wuwer and Matthias Gerdts
Abstract:	In this paper, we propose a simple yet robust method based on parallel graph search to tackle the task of two dimensional local motion planning. Following the discretize-then-optimize approach, we introduce a finite number of points in time and state space to approximate the continuous problem. It allows for explicit and therefore efficient evaluation of the arising constant control motion primitives. Furthermore, we derive assumptions on the local planning problem and its discretization that guarantee a multilayered, directed structure of the resulting feasibility graph. This allows exploitation of a parallelized graph search algorithm proposed in previous works. We numerically demonstrate efficiency and robustness of our algorithm in tasks arising from low-level helicopter flight as well as real-world autonomous driving.
Download

Area 4 - Data Analytics

Full Papers

Paper Nr:	17
Title:	Dynamic LEZ Simulation Tool: Traffic and Pollution Assessment
Authors:	Julen Macía, Iñaki Cejudo, Eider Irigoyen, Harbil Arregui and Itziar Urbieta
Abstract:	As urban mobility planning becomes more concerned about its environmental effects, one of the most widely applied interventions by administrations nowadays are Low-Emission Zones (LEZs), whose spatial and temporal extent are usually static. Nevertheless, pollution and traffic behaviour are dynamic, and studies have demonstrated the benefits from dynamic LEZ systems. In this work, an interactive tool for dynamic LEZ modelling is proposed. It is based on the microsimulation of traffic by SUMO, in any network and for any traffic demand provided. The main feature is that it takes into account LEZ characteristics determined by the user, and its perimeter can be modified during the simulation. Those changes will be reflected in the simulation in real-time. In that sense, pollution heatmaps are continuously generated, showing emissions from traffic directly affected by the LEZ. Moreover, this approach also allows to gain general metrics about the traffic state and pollution from the simulation scenario. Overall, this modular solution is valuable as a pre-emptive evaluation. Hence, all of this provides decision-making assessment to traffic managers on sustainable urban mobility regulations.
Download

Short Papers

Paper Nr:	18
Title:	Prediction of Traffic Crash Severity in the Maltese Islands
Authors:	Diana Cassara, Luana Chetcuti Zammit and Thérèse Bajada
Abstract:	Road traffic crashes represent a major global concern, impacting public safety, traffic congestion, and economic productivity. In Malta, the growing number of vehicles combined with a densely built environment, underscores the urgent need for predictive safety interventions. Research indicates that many road crashes exhibit recognisable patterns and are, to some extent, preventable. This work explores different machine learning techniques, to predict the severity of traffic crashes using training crash data in Malta. In this work, classification algorithms are developed to categorise crashes into four distinct severity classes, with promising prediction results. Furthermore, this work identifies high-risk zones and hotspots near critical infrastructures in the Maltese traffic network.
Download

Paper Nr:	19
Title:	Efficient Learning and Prediction of Variable Travel Times for a Vehicle Category in Different Geographies
Authors:	Fahad Rafique, Frank-Michael Schleif and Nitin Ahuja
Abstract:	This paper examines travel-time prediction across different geographic scales, focusing on regional and national levels in Switzerland. The national model covers the whole country, while the regional models only cover Zurich and Lucerne. This multi-scale approach addresses the distinct needs of transportation systems, where predictive accuracy, computational efficiency, and inference time are critical. The study evaluates the performance of a LightGBM predictive model across geographic areas using experiments with various data intervals, batching strategies, and spatial granularities. Regional models trained on local data achieved lower Root Mean Square Error (RMSE) values, indicating suitability for applications requiring high accuracy within specific regions. However, they showed limitations when exposed to unfamiliar travel patterns beyond their training scope. In contrast, the national model, trained on a broader dataset encompassing multiple regions, demonstrated greater flexibility across diverse traffic scenarios despite slightly higher RMSE values. It ef-fectively captured inter-regional travel trends, benefiting applications spanning wider areas, such as national logistics systems. Overall, this research provides an overview of the strengths and limitations of regional and national models, offering guidance on selecting them for transportation and logistics applications. It emphasizes the need for a balance between computational efficiency and predictive accuracy.
Download

Paper Nr:	36
Title:	Intelligent Ship Arrival Time Estimation with AIS Routes
Authors:	Paulo Silva, João Vitorino, Eva Maia and Isabel Praça
Abstract:	The optimization of the scheduling of shipping vessels plays a crucial role in port logistics. To improve operational efficiency, it is essential to correctly estimate a ship’s arrival time to a terminal when it is approaching a large port area. This approach is comprised of a data preparation stage for extraction of information related to shipping routes, a preprocessing stage for the extraction, cleaning and transformation of raw transmission data, originating in the Automatic Identification System (AIS), into trip data, a feature engineering stage for further trip information extraction, a feature analysis and selection stage and, finally, a model tuning and training stage to accurately predict ship Estimated Times of Arrival (ETAs). This approach underwent experimental validation with historical data from tanker and cargo ships, transmitted with the AIS, within the Galveston Bay Area channels and within the January of 2018 to April of 2020 time period. This process, with the aforementioned data, resulted in comparable results for a Random Forest, an XGBoost and a LightGBM models, verifying the importance of the extracted route information in this approach. These results reported very high prediction accuracy with an average error of approximately 20 minutes, making it suitable for large port areas.
Download

Paper Nr:	72
Title:	A Parallel Framework for Corridor Flow Calibration and Signal Optimization
Authors:	Rahul Sengupta, Ryan Casburn, Tushar Patel, Jeremy Dilmore and Sanjay Ranka
Abstract:	With increasing urbanization worldwide, road traffic congestion is a major issue. In cities, arterial corridors, that contain multiple successive signalized intersections, are vital. In these corridors, vehicles may enter and exit midway, making progression-oriented signal coordination essential. The focus is on reducing stops and delays while maintaining safety. This paper presents a simulation-based workflow for corridor calibration and signal timing optimization using SUMO microsimulation. It incorporates Ring-and-Barrier (NEMA) control, data-driven calibration using (a) ATSPM loop detector data, (b) sparse probe trajectory data, and (c) signal timing sheets. It presents a modular parallel simulation framework to scale up the evaluation of candidate timing plans. We apply Particle Swarm Optimization (PSO) to calibrate flows as well as to search over coordination parameters under operational constraints. As an example, we demonstrate that the system can reduce the mean corridor travel times relative to a baseline plan, by ∼10%.
Download

Area 5 - Smart Mobility and Sustainable Transport Services

Full Papers

Paper Nr:	49
Title:	Dynamic Graph Signal Processing for Anomaly Detection in Free-Floating Shared Mobility: Method, Benchmarks, and Real-World Evidences
Authors:	Svetlana Zubareva and Markus Lienkamp
Abstract:	Shared mobility systems such as free-floating e-scooters, bicycles, and cars generate large spatio-temporal data streams whose irregular patterns are difficult to detect using conventional methods. Existing anomaly detection approaches typically operate on independent time series or rely on static spatial statistics, limiting their ability to capture coordinated deviations across space and time. This paper proposes a dynamic graph signal processing (GSP) framework for anomaly detection in shared mobility systems. Mobility interactions are represented as time-varying, flow-weighted graphs, and aggregated usage measures are modeled as graph signals. By combining graph-spectral smoothing for spatial analysis with frequency-domain analysis of node-level time series, the framework separates spatially coherent anomalies from purely temporal irregularities. Rather than applying existing anomaly detectors to mobility data, the proposed approach introduces a unified formulation tailored to dynamic flow networks. The framework is evaluated on a large-scale real-world dataset of shared cars, bikes, and e-scooters in Munich (2023 - 2025), achieving an F1 score of 0.67 and recall of 0.85 against event-based proxy labels. Detected anomalies align with major events, holidays, and transport disturbances, demonstrating the method’s interpretability and practical relevance. The results indicate that incorporating network structure provides advantages over purely temporal or spatial baseline methods.
Download

Paper Nr:	55
Title:	Suspension Design for a Multi-Axle Electric Truck Moving over Cross-Country Terrain Using a Modified Half-Vehicle Model
Authors:	Mayank Khaparde, Shankar Krishnapillai and Piyush Shakya
Abstract:	Suspension design is an important factor affecting ride comfort and road-holding performance in multi-axle electric trucks operating on uneven and cross-country terrain. Most existing studies evaluate suspension behaviour assuming a constant vehicle speed, while actual driving conditions involve frequent acceleration and braking, which introduce additional inertial effects into the system. In this study, a modified half-vehicle model of a multi-axle electric truck is developed by incorporating longitudinal acceleration into the vertical and pitching dynamics. Considering the inertial forces produced during acceleration and deceleration leads to a more realistic suspension response and makes it possible to study dynamic effects that are normally neglected in conventional formulations. A non-linear suspension system is examined, and a multi-axle electric model is developed and simulated in MATLAB/Simulink across different longitudinal accelerations. Suspension performance metrics are assessed for multiple braking and acceleration scenarios over cross-country terrain. Pitch responses, including brake dive and squat, are observed when longitudinal acceleration is incorporated into the vertical vehicle dynamics, highlighting the shortcomings of constant-velocity assumptions. The findings show that accounting for inertial forces enhances the reliability of suspension performance prediction, offering clearer insight into the behaviour of heavy-duty electric trucks operating under varied loading environments and realistic driving conditions.
Download

Short Papers

Paper Nr:	20
Title:	Dynamic Road-Space Allocation: A Critical Review of Data, Methods and Readiness for Resilient Urban Transport Systems
Authors:	Eirini Stavropoulou, Lambros Mitropoulos and Annie Kortsari
Abstract:	Dynamic Road-Space Allocation (DRSA) has emerged as a promising approach for creating flexible, multimodal, and people-centred urban streets. While traditional Dynamic Lane Allocation (DLA) focuses primarily on motor-vehicle lanes, DRSA extends the concept to the entire street cross-section – sidewalks, bike lanes, curb space, and transit facilities – allowing infrastructure to adapt to temporal and spatial variations in demand. This study reviews and synthesizes existing DRSA tools and frameworks to assess their scope, capabilities, and operational readiness. Five tools were recorded using a structured form and systematically examined. Findings reveal that although DRSA shows strong conceptual potential, practical tools remain limited and vary widely in their methodological sophistication. Only one of the tools demonstrates the highest level of maturity, having been tested across multiple European cities. The remaining tools offer valuable modelling approaches but are constrained by simplification assumptions. Across all tools, common gaps relate to governance acceptance and regulatory constraints. The review highlights the need for next-generation DRSA tools that integrate real-time sensing, digital twin environments, and machine-learning techniques, while also embedding equity and accessibility considerations. Advancing these tools will be essential for scaling DRSA from research applications to common practice in urban street redesign.
Download

Paper Nr:	29
Title:	Development and Demonstration of Innovative Nonlinear State-of-Charge Estimation Techniques for EVs/HEVs
Authors:	Ehsan Sobhani-Tehrani, Nicolae Tudoroiu and Khashayar Khorasani
Abstract:	The State-of-Charge (SoC) of a battery is its available capacity expressed as a percentage of its rated capacity. The problem of SoC estimation is of paramount importance for successful commercialization of both Electric Vehicles (EV) and Hybrid Electric Vehicles (HEV). In this paper, we propose three nonlinear estimation techniques based on Extended Kalman Filtering (EKF), Unscented Kalman Filtering (UKF) and Neural Networks Filtering that can estimate the SoC of Nickel-Metal Hydride (Ni-MH) battery pack. We expect these techniques could also be extended to other battery chemistries as well. The novelty of this paper is in its implementation of UKF and neural networks filtering techniques for the SoC estimation of Ni-MH battery pack. These algorithms are shown to be clearly superior alternatives to the EKF technique that has been extensively developed in the literature.
Download

Paper Nr:	59
Title:	A Map-Based Predictive Speed and Energy Optimization Framework for Autonomous Electric Buses
Authors:	Doğacan Dolunay Acar, Ergün Can, Bora Duman, Duygu Kayaoğlu, Cemre Kavvasoglu, Alper Baykut, Ali Ufuk Peker and Kerem Par
Abstract:	This paper presents a map-based predictive speed and energy optimization framework for electric vehicles, aiming to improve energy efficiency, driving consistency, and operational range under varying road conditions. The proposed approach integrates high-resolution digital map information with a physics-based energy model to generate an optimized speed profile in advance. MATLAB-based simulations conducted on flat and hilly routes demonstrate clear energy efficiency gains compared to aggressive, normal, and comfort driving styles, with the optimized automated strategy achieving up to 18% reduction in consumed energy and up to 20% improvement in net energy. The framework is further validated through real-world experiments on operational routes, including automated bus deployments. Results indicate consistent regenerative energy behavior and measurable range improvements. Comparative evaluations against human driving and baseline autonomous strategies highlight the practical potential of the proposed method for automated public transportation and commercial fleet applications.
Download

Paper Nr:	83
Title:	Kinematic-Based State of Charge Estimation: Model Development and Real-World Validation
Authors:	Sofia Borgosano, Gianluca Canali, Andrea Di Martino and Michela Longo
Abstract:	Accurate State of Charge (SoC) estimation is critical for mitigating range anxiety and optimising energy management in Electric Vehicles (EVs). However, conventional estimation methods often require direct access to internal battery parameters, which are frequently restricted by proprietary Battery Management Systems (BMS). This paper presents a non-invasive SoC estimation model based exclusively on vehicle kinematic variables, velocity and acceleration, obtainable via standard smartphone sensors. By reconstructing the longi-tudinal dynamics and implementing an electromechanical model of a Permanent Magnet Synchronous Motor (PMSM) and a lithium-ion battery, the electrical power consumption is derived and integrated over time. The model was validated using an Opel Corsa-e through 18 experimental driving cycles on a 14 km real-world route. Results demonstrate a high degree of correlation with experimental data, achieving a Mean Absolute Error (MAE) of 0.9197% SoC. The proposed approach offers a robust sensor-fusion-based solution for real-time energy monitoring in third-party electric mobility applications.
Download