Electrical/Electronic Architectures (EEAs) are continuously evolving to meet newly emerging demands. In recent years, major drivers of this evolution have been the increasing software-defined nature of vehicles and the push toward automated driving. Key technologies such as edge-enhanced functions, vehicle-to-vehicle communication, and service-oriented architectures are therefore the focus of current research efforts. This paper presents a vision of how these technologies can be used to enable cooperation between vehicles, illustrated by using parked vehicles as edge nodes. These are typically seen as obstructions, as they significantly increase the risk of missing or misinterpreting vulnerable road users such as pedestrians or cyclists. Our proposed approach to counteract this problem is the use of the parked vehicles themselves as edge nodes that support object detection or even trajectory planning. Current research primarily considers smart traffic infrastructure, roadside units, and other vehicles as potential edge nodes. Including parked vehicles as edge nodes means that, instead of acting solely as obstacles, we leverage their built-in sensors to contribute to cooperative awareness. While such cooperation will enhance the safety of automated vehicles in urban areas, several challenges arise. In this paper, we discuss how data traceability, decision-making in the presence of conflicting information, and incentive mechanisms for owners of parked vehicles can be addressed. Based on these challenges, the paper outlines requirements for future cooperative architecture and highlights the role of edge-enhanced functions, Vehicle-to-Vehicle (V2V) communication, and service-oriented architectures in enabling fully automated driving.

Vehicle diagnostic systems are crucial for the normal operation of vehicles and their propulsion-related systems. Undetected unusual behaviour of such systems makes the vehicle diagnostic system unreliable. Current diagnostic systems, such as On-Board Diagnostics (OBD), are limited to monitoring only specific systems in order to make fault decision. However, various anomalies, including drastic performance drops, vehicle tampering, and changes in the driving environment, often go undetected during OBD system testing, validation, and inspection. This research presents a novel explainable OBD anomaly detection pipeline that is able to detect anomalies based only on OBD data snapshots during processes of OBD validation and inspection. The novel approach is implemented using combined dimension reduction and data clustering methodologies. First, the data is transformed into a latent space using t-distributed Stochastic Neighbor Embedding (t-SNE), where the general structure of the anomaly in the data can be exploited. Subsequently, clustering using Density-Based Spatial Clustering of Applications with Noise (DBSCAN) is applied to group similar normal data and identify anomalous patterns. The novelty of the solution is further extended with a feedback loop that suggests the root cause of OBD signals for individual anomalies using explainable AI (XAI) methodology, in this case Shapley additive explanations (SHAP). The proposed concept was verified and evaluated using real OBD snapshots with synthetically generated anomalies in two scenarios with different engine status, engine off and on, with an achieved accuracy of 92.89% and 96.45% for anomaly detection, respectively. The majority of anomaly causes in the form of specific OBD signals from propulsion- and emission-related systems were successfully explained using SHAP.

This study investigates the feasibility of utilizing thermal preconditioning of the passenger cabin of an electric coach during recuperation phases to save energy and reduce recharging phases on the battery system. The concept involves utilizing the cabin as a form of thermal storage, particularly during recuperating phases through running the Heating, Ventilation and Air Conditioning (HVAC) system on full power. The proposed approach addresses a limitation of current recuperation strategies: the restricted capacity of the High Voltage (HV) battery to absorb energy during braking. The overarching objective is to increase the effective energy yield from braking events and enhance overall system efficiency through targeted energy distribution and predictive control strategies, while limiting the need for constant battery recharging. Simulations based on real-world driving cycles of the operating routes of currently running diesel coaches. The results reveal that there is virtually no opportunity for meaningful energy savings for the tested routes. Overall, the findings suggest that more sophisticated electric coaches have little potential in saving energy through preconditioning, and the best results are achieved by simple controllers such as PID controllers. Future research on city buses could yield more promising results, as their frequent door openings lead to significant heat exchange with the environment, making thermal preconditioning strategies potentially more impactful.

The transition toward fully automated electric vehicles (AEVs) demands robust safety mechanisms capable of addressing unforeseen internal critical situations without reliance on human drivers. Conventional diagnostic systems remain constrained to predefined failure modes and cannot capture all safety-critical anomalies. This paper introduces an operational hypervisor framework, an integrated anomaly detection and fault-tolerant control advisor, designed to enhance unforeseen internal-system critical situations in AEVs. The proposed approach combines data-driven anomaly detection, including isolation forests, correlation analysis, and explainable AI (XAI) to capture anomalous patterns across heterogeneous AEV subsystems. Beyond anomaly detection, the operational hypervisor functions as a fault-tolerant advisory layer, attributing anomalies to their probable sources, quantifying their potential impact, and recommending context-aware corrective actions to ensure safe operation. To ensure robustness, the framework employs multistage data preprocessing, improving sensitivity to both shortterm and persistent anomalies. Designed for real-time execution, the hypervisor can be seamlessly integrated into the control architecture of AEVs, allowing interaction with decision-making and vehicle supervisory layers. Experimental evaluation with real occurrences of safety-critical AEVs anomalies verifies the system’s capability to extend detection beyond conventional diagnostic limits and provide interpretable feedback to controllers. By coupling explainable anomaly detection with anomaly handling, operational hypervisor advances the state of operational safety management in AEVs.

SUMMARY & CONCLUSIONSMost, if not all, safety-critical systems developed are cyber-physical systems (CPS), defined by the presence of a physical component and connectedness to other systems, with a clear trend towards inclusion of over-the-air update capabilities. From a certification standpoint, every change to the system, even in software, should lead to an evaluation of its safety impact. Model-based system engineering offers the possibility of connecting domain-specific models used in the development of the CPS with relevant safety analysis models. The goal is to automate the evaluation of the impact of changes. Changes, however, can only be propagated if the dependencies of the specific model in which the change is introduced are accurately represented. These are difficult to model, as formal descriptions only exist for a subset of dependencies, and some have been completely neglected in model-based systems engineering. Our work provides an overview of the various shapes that dependencies take within model-based safety analysis. The importance of modeling these dependencies is shown in an illustrative example, in which an FMEA, SysML, and Lifecycle model are investigated.

The increasing use of Large Language Models (LLMs) offers significant opportunities across the engineering lifecycle, including requirements engineering, software development, process optimization, and decision support. Despite this potential, organizations face substantial challenges in assessing the risks associated with LLM use, resulting in inconsistent integration, unknown failure modes, and limited scalability. This paper introduces the LLM Risk Assessment Framework (LRF), a structured approach for evaluating the application of LLMs within Systems Engineering (SE) environments. The framework classifies LLM-based applications along two fundamental dimensions: autonomy, ranging from supportive assistance to fully automated decision making, and impact, reflecting the potential severity of incorrect or misleading model outputs on engineering processes and system elements. By combining these dimensions, the LRF enables consistent determination of corresponding risk levels across the development lifecycle. The resulting classification supports organizations in identifying appropriate validation strategies, levels of human oversight, and required countermeasures to ensure safe and transparent deployment. The framework thereby helps align the rapid evolution of AI technologies with established engineering principles of reliability, traceability, and controlled process integration. Overall, the LRF provides a basis for risk-aware adoption of LLMs in complex engineering environments and represents a first step toward standardized AI assurance practices in systems engineering.

The increasing capabilities of machine learning models, such as vision-language and multimodal language models, are placing growing demands on data in automotive systems engineering, making the quality and relevance of collected data enablers for the development and validation of such systems. Traditional Big Data approaches focus on large-scale data collection and offline processing, while Smart Data approaches improve data selection strategies but still rely on centralized and offline post-processing. This paper introduces the concept of Fast Data for automotive systems engineering. The approach shifts data selection and recording onto the vehicle as the data source. By enabling real-time, context-aware decisions on whether and which data should be recorded, data collection can be directly aligned with data quality objectives and collection strategies within a closed-loop. This results in datasets with higher relevance, improved coverage of critical scenarios, and increased information density, while at the same time reducing irrelevant data and associated costs. The proposed approach provides a structured foundation for designing data collection strategies that are aligned with the needs of modern machine learning algorithms. It supports efficient data acquisition and contributes to scalable and cost-effective ML development processes in automotive systems engineering.

With the increasing demand for sustainable transportation in the face of challenges such as climate change and urbanization, optimizing the energy efficiency of Electric City Buses (ECBs) is essential. This study employs explainable artificial intelligence techniques, specifically SHapley additive expansion (SHAP), to assess the influence of factors such as vehicle speed, acceleration, and braking on the energy consumption of the drivetrain. The data is segmented into distinct scenarios, including acceleration, starting, curves, uphill, and downhill driving. In driving conditions like curves or uphill and downhill routes, the brake pedal position, alongside the accelerator position and vehicle speed, emerged as key factors impacting drivetrain consumption. Secondly, the study delves into analyzing driving behavior during bus stop entries and leaving instances, employing methods like Deep Autoencoder-based Clustering (DAC) and Self-Organizing Map (SOM). This analysis identified groups with energy-efficient and energy-inefficient driving behaviors, with certain clusters showing high acceleration and low braking use, particularly during nighttime or low-traffic conditions. These insights highlight the potential for energy savings by promoting smoother, more consistent driving styles, especially as electric buses approach or depart from bus stops.

Autonomous robotic systems require spatio-temporal understanding of dynamic environments to ensure reliable navigation and interaction. While Vision-Language Models (VLMs) provide open-world semantic priors, they lack grounding in 3D geometry and temporal dynamics. Conversely, geometric perception captures structure and motion but remains semantically sparse. We propose SNOW (Scene Understanding with Open-World Knowledge), a training-free and backbone-agnostic framework for unified 4D scene understanding that integrates VLM-derived semantics with point cloud geometry and temporal consistency. SNOW processes synchronized RGB images and 3D point clouds, using HDBSCAN clustering to generate object-level proposals that guide SAM2-based segmentation. Each segmented region is encoded through our proposed Spatio-Temporal Tokenized Patch Encoding (STEP), producing multimodal tokens that capture localized semantic, geometric, and temporal attributes. These tokens are incrementally integrated into a 4D Scene Graph (4DSG), which serves as 4D prior for downstream reasoning. A lightweight SLAM backend anchors all STEP tokens spatially in the environment, providing the global reference alignment, and ensuring unambiguous spatial grounding across time. The resulting 4DSG forms a queryable, unified world model through which VLMs can directly interpret spatial scene structure and temporal dynamics. Experiments on a diverse set of benchmarks demonstrate that SNOW enables precise 4D scene understanding and spatially grounded inference, thereby setting new state-of-the-art performance in several settings, highlighting the importance of structured 4D priors for embodied reasoning and autonomous robotics.

Cyber-Physical Systems integrate mechanical, electrical, and software components, resulting in tightly coupled development processes and extensive regression-testing demands. As agile and over-the-air update strategies increase the number of system versions, executing full regression test suites becomes infeasible due to resource and time constraints. Existing test case prioritization techniques, developed largely within the software domain, do not account for cross-domain dependencies and therefore overlook faults that arise only through interactions between heterogeneous engineering artifacts. This work presents a change-aware, cross-domain test case prioritization method that builds on Multi-Domain Dependency Graphs (MDDGs) as a unified representation of CPS. By propagating modifications via graph-based relations across all engineering domains, change impact is estimated and used to derive execution orders that reflect the system’s structural and semantic dependencies. The approach is instantiated via a set of graph-based metrics that score test cases along paths connecting changed artifacts and test-relevant elements. A modular implementation with interfaces to existing engineering tools enables integration into modelbased development environments without requiring changes to underlying domain-specific models. The method is evaluated in representative scenarios from an embedded vehicle platform. Across different datasets, the approach achieves Average Percentage of Faults Detected (APFD) values between 60% and 90%, exceeding the 50% expected from random test ordering, indicating improved fault detection relative to random test execution while preserving the full test suite and thus avoiding coverage losses associated with minimization techniques.

The automotive software architecture is struggling with production stability and efficiency, as frequent and regular software updates incur side effects on software-related processes in the end assembly. To close the gap, this article refines and formalizes the concept of production-oriented software partitioning (POSP), a novel architecture for in-car software which physically decouples software used within production from production-irrelevant software. Experimental evaluation using a developed prototype confirms the feasibility of the proposed architecture. The prototype further showcases the potential of POSP, achieving a 16% reduction in software size for production use. Moreover, the article contributes a generic model for analyzing automotive software systems at the software component level, a formal workflow for refactoring legacy systems toward POSP, and a redundancy-based method for partitioning software systems following the generic model. These contributions are demonstrated on an industry-grade AUTOSAR Classic platform, highlighting the potential of production-oriented software partitioning (POSP) for short-term industrial adoption and enhanced stability and efficiency in large-scale automotive production.

Data-driven systems depend on task-relevant data, yet data collection pipelines remain passive and indiscriminate. Continuous logging of multimodal sensor streams incurs high storage costs and captures irrelevant data. This paper proposes a declarative framework for intent-driven, on-device data collection that enables selective collection of multimodal sensor data based on high-level user requests. The framework combines natural language interaction with a formally specified domain-specific language (DSL). Large language models translate user-defined requirements into verifiable and composable DSL programs that define conditional triggers across heterogeneous sensors, including cameras, LiDAR, and system telemetry. Empirical evaluation on vehicular and robotic perception tasks shows that the DSL-based approach achieves higher generation consistency and lower execution latency than unconstrained code generation while maintaining comparable detection performance. The structured abstraction supports modular trigger composition and concurrent deployment on resource-constrained edge platforms. This approach replaces passive logging with a verifiable, intent-driven mechanism for multimodal data collection in real-time systems.

The ongoing transformation of the automotive industry is primarily driven by vehicle electrification and automation. In this context, system safety and reliability remain critical requirements in the development of modern vehicles and their subsystems. However, conventional diagnostic approaches exhibit significant limitations in addressing the increasing complexity of automated electric vehicles. Electrified powertrains introduce novel fault conditions, while highly automated driving systems reduce the possibility of human supervision of safety-critical components. This paper proposes a predictive diagnostic framework for battery systems in automated electric vehicles. The proposed methodology employs a two-stage diagnostic architecture based on Support Vector Machines (SVM) and one-dimensional Convolutional Neural Networks (1D CNNs) for real-time detection and prediction of electrical faults. The framework was experimentally validated on a fully automated electric vehicle prototype. The obtained results demonstrate high performance of proposed concept, achieving fault classification accuracies of 96.43% for the SVM model and 99.85% for the 1D CNN model. Furthermore, the proposed approach enables the prediction of specific battery-related faults up to 49 s prior to their actual occurrence.

Humans perceive and reason about their surroundings in four dimensions-three spatial and one temporal axis-by building persistent, structured internal representations that encode semantic meaning, spatial layout, and temporal dynamics. These multimodal memories enable them to recall past events, infer unobserved states, and integrate new information into context-dependent reasoning. Inspired by this capability, we introduce R4, a training-free framework for retrieval-augmented reasoning in 4D spatio-temporal space that equips vision-language models (VLMs) with structured, lifelong memory. R4 continuously constructs a 4D knowledge database by anchoring object-level semantic descriptions in metric space and time, yielding a persistent world model that can be shared across agents. At inference, natural language queries are decomposed into semantic, spatial, and temporal keys to retrieve relevant observations, which are integrated into the VLM's reasoning through an iterative retrieval-reasoning loop. Unlike classical retrieval-augmented generation methods, retrieval in R4 operates directly in structured 4D space, enabling episodic and collaborative reasoning without training. Experiments on embodied question answering and navigation benchmarks demonstrate that R4 substantially improves retrieval and reasoning over spatio-temporal information compared to baselines, advancing a new paradigm for embodied 4D reasoning in dynamic environments.

Modern AI systems are increasingly constrained not by model capacity but by the quality and diversity of their data. Despite growing emphasis on data-centric AI, most datasets are still gathered in an open-loop manner which accumulates redundant samples without feedback from the current coverage. This results in inefficient storage, costly labeling, and limited generalization. To address this, this paper introduces FCDC, a paradigm that formulates data collection as a closed-loop control problem. FCDC continuously approximates the state of the collected data distribution using an online probabilistic model and adaptively regulates sample retention using based on feedback signals such as likelihood and Mahalanobis distance. Through this feedback mechanism, the system dynamically balances exploration and exploitation, maintains dataset diversity, and prevents redundancy from accumulating over time. Besides showcasing the controllability of FCDC on a synthetic dataset, experiments on a real data stream show that FCDC produces more balanced datasets by 25.9 while reducing data storage by 39.8. These results demonstrate that data collection itself can be actively controlled, transforming collection from a passive pipeline stage into a self-regulating, feedback-driven process at the core of data-centric AI.

Agile methods have shaped the development of enterprise software systems during the last two decades. However, many modern cyber-physical systems (CPS) are still developed in as yet waterfall-like processes. The consequence is that CPS development misses out on such advantages of agile methods as handling changing requirements providing fast updates, or dealing with fast feedback on product quality. This is especially problematic today, when the software in CPS systems is more networked than ever, requiring updates to keep pace in an ever evolving network-connected technical environment, as well as patching too often software-induced cyber-security vulnerabilities. In sum, modern CPS must be developed so as to meet the need for updates at intervals of rapidly accelerating frequency. In this paper, we discuss the lack of systematic cross-model consistency management as one of the reasons why established agile methods are not used in CPS development. We present a road map that leads to systematic consistency management, laying the foundations of novel agile methods in CPS development. We discuss solutions in the context of model-driven automotive systems engineering. This domain especially can serve as a litmus test of agility in CPS development, because automotive systems engineering stands to benefit substantially from agile methods to address such pressing issues as strong assurance of dependability and configurability while also offering the flexibility of software over-the-air updates.

With the increased use of information technology (IT), cars are transforming into software-defined vehicles. Software with machine learning algorithms is realizing components dedicated to automated driving. However, the established processes for development and testing cannot keep pace with the innovations in vehicle software architecture. An increasing number of IT experts are involved in system design and development to cope with the rising proportion of software in vehicles. The development team is growing and spans the mechanical, electronic, software, and machine learning domains, resulting in a heterogeneous team structure. These heterogeneous teams must overcome challenges related to collaboration during development and testing. A process is needed to test the vehicle software architecture, which is currently changing to service orientation. The testing process must be re-usable for software updates after release to maintain the vehicle software throughout the entire product life cycle. In this paper, we propose an approach where we use existing concepts from the IT domain for collaboration and continuous software testing and, where appropriate, tailor them to the automotive industry's needs, considering existing organizational structures and established standards. The approach presented is abstracted to a set of technologies that create a solution for the existing problems being as independent of the realization as possible.

The switch to service-oriented architectures in the automotive sector offers the opportunity to address key industry challenges, such as the reusability of hardware and software components, and the more efficient integration and maintainability of digital functions. The service-oriented architecture also provides the basis for the implementation of cloud-based vehicle software components. Offloading to the cloud is an upcoming option for compute-intensive functions to satisfy high resource demands. Advanced control mechanisms, such as model predictive control represent a particular component that requires high computing resources. Against this background, the overarching aim of this study is to investigate the control-over-the-air approach using the example of heating,ventilation, and air-conditioning control of battery electric buses. For this purpose, a hardware setup was used to assess the suitability of cloud-based control and investigate the need for a fallback in the vehicle and potential energy savings through offloading.

Multimodal large language models (MLLMs) hold the potential to enhance autonomous driving by combining domain-independent world knowledge with context-specific language guidance. Their integration into autonomous driving systems shows promising results in isolated proof-of-concept applications, while their performance is evaluated on selective singular aspects of perception, reasoning, or planning. To leverage their full potential a systematic framework for evaluating MLLMs in the context of autonomous driving is required. This paper proposes a holistic framework for a capability-driven evaluation of MLLMs in autonomous driving. The framework structures scenario understanding along the four core capability dimensions semantic, spatial, temporal, and physical. They are derived from the general requirements of autonomous driving systems, human driver cognition, and language-based reasoning. It further organises the domain into context layers, processing modalities, and downstream tasks such as language-based interaction and decision-making. To illustrate the framework's applicability, two exemplary traffic scenarios are analysed, grounding the proposed dimensions in realistic driving situations. The framework provides a foundation for the structured evaluation of MLLMs' potential for scenario understanding in autonomous driving.

Cyber-Physical Systems (CPS) are currently the predominantly used kind of system across industries. They distinguish themselves through a strong dependency between the engineering domains. The difficulty arising from these dependencies is handled by using Model Based Systems Engineering (MBSE) and clearly defined organizational processes. The importance of both is underlined by how certification processes are handled. For the CPS most people come in contact with, the car, the required certification process not only requires documentation of test results but also that certain development steps, such as the creation of a safety concept, have been taken in the vehicle development. After the certification of a CPS, the development lifecycle, however, does not end. Among others, two recurring causes can regularly be seen in vehicle development that create the need for re-certification. Firstly, components of the vehicle might become unavailable due to supply chain issues. Secondly, already deployed vehicles will receive software updates, with a re-certification required according to UN ECE R156. In this paper, we propose a method addressing the issue of recertification by combining MBSE with knowledge about organizational processes. More concretely, we suggest tracking the system engineering process across all involved domains within an activity diagram. This activity diagram is checked for consistency whenever a change occurs. The change can be directly traced by following the consistency specification and the activity diagram to the specific steps required for re-certification. This enables us to accurately assess the impact of our changes on safety artifacts, as well as allowing the automated selection of required actions in the re-certification process.