Automated driving (AD) promises major benefits in road safety, traffic efficiency, and environmental impact, yet full autonomy remains challenging due to highly dynamic, uncertain traffic environments and the need to interact safely with unpredictable human drivers. This work proposes a unified Nonlinear Model Predictive Control (NMPC) framework that functions both as a low-level controller and a high-level decision-making system, enabling automated vehicles to plan smooth and safe maneuvers over a predictive horizon. Unlike many existing NMPC implementations that are constrained to narrowly defined tasks and require scenario-specific parameter tuning, the proposed approach integrates nonlinear vehicle dynamics, multi-vehicle interactions, and safety constraints within a single formulation designed to generalize across complex driving situations. The framework is validated through simulations of challenging real-world maneuvers, including merging in extra-urban and urban settings (with compliance to stop-and-go and traffic regulations) and navigating multi-vehicle roundabouts with dense interactions. Results indicate collision-free operation, smooth trajectories without abrupt maneuvers, and robust adaptability to varying traffic densities and speed profiles.

Despite the rapid evolution of automated driving, public information about deployability in production vehicles remains limited. We present a Field Programmable Gate Array (FPGA) prototyping study of an AI module for estimating time-to-lane-change (TTLC), focusing on key deployment metrics such as latency, memory footprint, and power/energy consumption. FPGAs enable custom acceleration while providing flexibility to explore multiple design trade-offs. Particularly, we investigate different quantization strategies, given their potential to substantially reduce model size while preserving predictive performance. This paper makes four main contributions that advance the state-of-the-art through an FPGA-based implementation. We present a hardware-aware FPGA design methodology that balances memory efficiency and task accuracy. We provide a comprehensive analysis of the trade-offs achievable by varying AI model type, operating frequency, and quantization precision. We find a set of Pareto-optimal solutions along the power-latency axes, identifying a clear elbow at a co-processor frequency of 25/50 MHz depending on model and quantization type. We implement a Transformer architecture that preserves state-of-the-art TTLC predictive performance while meeting the 10 Hz real-time constraint and reducing total energy consumption by 83% compared to a high-end microcontroller.

Advanced driving simulations are increasingly used in automated driving research, yet freely available data and tools remain limited. We present a new open source framework for synthetic data generation for lane change (LC) intention recognition in highways. Built on the CARLA simulator, it advances the state-of-the-art by providing a 50-driver dataset, a large-scale 3D map, and code for reproducibility and new data creation. The 60 km highway map includes varying curvature radii and straight segments. The codebase supports simulation enhancements (traffic management, vehicle cockpit, engine noise) and Machine Learning (ML) model training and evaluation, including CARLA log post-processing into time series. The dataset contains over 3,400 annotated LC maneuvers with synchronized ego dynamics, road geometry, and traffic context. From an automotive industry perspective, we also assess leading-edge ML models on STM32 microcontrollers using deployability metrics. Unlike prior infrastructure based works, we estimate time-to-LC from ego-centric data. Results show that a Transformer model yields the lowest regression error, while XGBoost offers the best trade-offs on extremely resource-constrained devices. The entire framework is publicly released to support advancement in automated driving research.

Driver distraction is one of the main causes of traffic accidents. While there are different types of distraction (manual, visual, cognitive), cognitive distraction is particularly challenging, being only partially related to visual features detectable through cameras or an eye tracker system. Moreover, since cognitive distraction is not a point in time phenomenon, spotting this kind of distraction requires the processing of a certain time interval, which poses a further challenge for real-time performance. After a data collection campaign with N =42 subjects undertaking a twenty-question task (TQT) in a driving simulator, we developed a driver cognitive distraction detection system, with the goal of filling in some key gaps we identified in the literature towards real world deployment. First, we assessed the effectiveness of state-of-the-art time series-oriented deep learning models in learning features from 60 Hz raw input signals, thus implementing an end-to-end machine learning approach, without manual feature engineering. We demonstrated that such models are able to classify time-windows as small as 0.5 seconds, and are also more robust to sensor failures. Second, also with the support of AI explainability, we showed that processing vehicular data is fundamental to ensure performance, while physiological signals provide a less important, but still useful, contribution. Third, through a between- and within-subject design comparison, we showed that eye-tracker and, particularly, physiological signals are much more prone to inter-individual variability, thus overfitting. This is fundamental to consider for commercial deployment, as it would require fine-tuning the system with data from the actual end-user. Fourth, we quantitatively measured the effect of such variability for all types of signals, demonstrating its huge relevance, and shown that deep learning models dedicated to time series processing are better able to generalize across users than the more commonly employed shallow machine learning models. Finally, with a focus on in-vehicle deployability, which is of significant industrial interest, we measured also such metrics as model size, inference time, and energy consumption, showing feasibility on two embedded platforms, which is a key advancement towards on-board deployment of robust, real-time cognitive distraction detection systems.

Dataset availability is a fundamental requirement for advancing Automated Driving Functions (ADFs) through ever more reliable machine learning systems. The huge costs related to the development of real-world datasets have spurred the utilization of simulations environments to collect data from virtual reality scenarios. Such synthetic data can be used to pre-train neural models, that may in turn be fine-tuned on a smaller amount of real-world data. We present the development of a lightweight simulation environment, specifically targeting data collection for Driver Maneuver Intention Recognition (DMIR) systems. We created a realistic highway driving scenario to capture Lane Change (LC) intentions. The simulation setup offers a driving experience through a comprehensive system that closely replicates real-world conditions, including a wide screen, steering wheel, and pedals for the vehicle control. Procedural traffic generation was implemented to maintain computational efficiency while ensuring realism. Preliminary tests with four participants demonstrated the simulator's effectiveness, yielding an average of six LCs per minute. User feedback highlighted areas for improvement, such as perceived road width and traffic realism, indicating potential enhancements like additional cameras for a broader field of view. Our findings underscore the viability of simulation environments in extensively and cost-effectively generating data usable to advance AD technologies.

Driver distraction recognition is gaining increasing interest in improving traffic safety, as well as in automated driving. This article reports the experience we have gained developing a driver distraction detection (DDD) system within the Hi-Drive research project on driving automation. Targeting on-board deployability, we have faced several leading-edge research issues that have not been addressed together in published research works. We propose a compact sensory configuration and a limited computational resource system architecture, also exploiting careful manual and automated labeling, trying to find a tradeoff among conflicting needs in terms of accuracy, privacy preservation, energy efficiency, and costs. Our system detects two levels of insufficient attention, which are keys not only for designing a proper driver warning and information management strategy but also for better managing the transition among different automation levels. Our experiments confirmed on real-world data, and in the three-class task, the importance of distinguishing users among training, validation, and testing to prevent overestimating model performance by overfitting individual participant patterns present in all three subsets. We analyzed the complexity of the three-class problem, which is also related to the relatively low representation of the intermediate distraction class in the dataset. We showed that the size of the classifiable time window is a critical performance factor and found that a 5 s length seems to achieve the best tradeoff between latency, time resolution, and the need for capturing sufficient temporal information to detect distraction. Another empirical finding comes from a SHAP values-based explainability analysis and concerns the importance of vehicular signals for the detection task, particularly in the three-class problem. This is significant, as such signals are inexpensively available in vehicles, and their processing does not add further privacy concerns. Finally, assessing performance on three state-of-the-art embedded platforms, we observed that the developed deep learning models are able to effectively run on limited-resource, on-board deployable devices, meeting real-time performance requirements, also on a mainstream, low-cost microcontroller. We argue that these findings open significant perspectives toward an effective and efficient field deployment of DDD electronic systems.

This paper presents a novel tool for generating driving scenario datasets, that are a key asset to advance research and development in automated driving and driver assistance systems. The tool relies on the MATLAB. Automated Driving Toolbox and focuses on the overtaking maneuver. It uses simulated vehicular data, without relying on camera-equipped real-world vehicles, thus providing a low-cost solution, while allowing to abstract the main action features, that are very important for the pre-training of machine learning models. The tool has been designed to target customization (in terms, e.g., of road curvature radii), in order to allow meeting specific requirements, while its interoperability (e.g., multiple-format export) supports integration with other development environments. A preliminary analysis of the first scenarios generated with the tool confirms the validity of the system under development.

Artificial Intelligence (AI) offers the potential to transform our lives in radical ways. In particular, when AI is combined with the rapid development of mobile communication and advanced sensors, this allows autonomous driving (AD) to make a great progress. In fact, Autonomous Vehicles (AVs) can mitigate some shortcomings of manual driving, but at the same time the underlying technology is not yet mature enough to be widely applied in all scenarios and for all types of vehicles. In this context, the traditional SAE-levels of automation (J3016B: Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles—SAE International. Available online: https://www.sae.org/standards/content/j3016_201806/ ) can lead to uncertain and ambiguous situations, so yielding to a great risk in the control of the vehicle. In this context, the human drivers should be supported to take the right decision, especially on those edge-cases where automation can fail. A decision-making system is well designed if it can augment human cognition and emphasize human judgement and intuition. It is worth to noting here that such systems should not be considered as teammates or collaborators, because humans are responsible for the final decision and actions, but the technology can assist them, reducing workload, raising performances and ensuring safety. The main objective of this paper is to present an intelligent decision support system (IDSS), in order to provide the optimal decision, about which is the best action to perform, by using an explainable and safe paradigm, based on AI techniques.

The development of increasingly automated vehicles (AVs) is likely to lead to new challenges around how they will interact with other road users. In the future, it is envisaged that AVs, manually driven vehicles, and vulnerable road users such as cyclists and pedestrians will need to share the road environment and interact with one another. This paper presents a test track study, funded by the H2020 interACT project, investigating pedestrians’ reactions towards an AV’s movement patterns and external Human Machine Interfaces (eHMIs). Twenty participants, standing on the side of a test-track road and facing an approaching AV, were asked to raise their arm to indicate: (1) when they could perceive the AV’s eHMI, which consisted of either a Full Light Band (FLB) or a Partial Light Band (PLB); (2) when they perceived the deceleration of the AV (with eHMI vs. no eHMI); and (3) when they felt safe to cross the road in front of the approaching AV (with eHMI vs. no eHMI). Statistical analyses revealed no effects of the presence of an eHMI on the pedestrians’ crossing decision or deceleration perception, but significant differences were found regarding the visibility of the FLB and PLB designs. The PLB design could be perceived at further distances than the FLB design. Both eHMI solutions were generally well-received, and participants provided high ratings of acceptance, perceived safety, and confidence around the AV.

The Shared Control (SC) cooperation scheme, where the driver and automated driving system control the vehicle together, has been gaining attention through the years as a promising option to improve road safety. As a result, advanced interaction methods can be investigated to enhance user experience, acceptance, and trust. Under this perspective, not only the development of algorithms and system applications are needed, but it is also essential to evaluate the system with real drivers, assess its impact on road safety, and understand how drivers accept and are willing to use this technology. In this sense, the contribution of this work is to conduct an experimental study to evaluate if a previously developed shared control system can improve overtaking performance on roads with oncoming traffic. The evaluation is performed in a Driver-in-the-Loop (DiL) simulator with 13 real drivers. The system based on SC is compared against a vehicle with conventional SAE-L2 functionalities. The evaluation includes both objective and subjective assessments. Results show that SC proved to be the best solution for assisting the driver during overtaking in terms of safety and acceptance. The SC's longer and smoother control transitions provide benefits to cooperative driving. The System Usability Scale (SUS) and the System Acceptance Scale (SAS) questionnaire show that the SC system was perceived as better in terms of usability, usefulness, and satisfaction.

Automated and highly automated vehicles still need to interact with the driver at different cognitive levels. Those who are SAE level 1 or 2 consider the human in the loop all the time and require strong participation of the driver at the control level. Yet, to increase safety, trust, and driver comfort with this mode of automation, systems with a strong cooperative component are needed. In this sense, this paper introduces the design of a vehicle controller based on shared control, together with an arbitration system, and the design of a visual human-machine interface (HMI) to foster the mutual understanding between driver and automation while sharing driving task. The driver-automation cooperation is achieved through incremental support, in a continuum spectrum from manual to full automation. Additionally, the design of an HMI to support the driver in a takeover maneuver is presented, considering that this functionality is a key component of vehicle SAE levels 3 and 4. The work presented in this paper represents the latest advances in the integration being carried out within the framework of the PRYSTINE project.

Driving is a complex activity which requires constant care and attention. Intelligent Advance Driver Assistance Systems (ADAS) can improve vehicle control performance and, thus, drivers and passengers safety. In particular, identification and prediction of driving intention can provide prompt information to drivers and vehicles in their vicinity that are fundamental for avoiding collisions. In this paper, we propose a lane change prediction model based on machine learning able to distinguish between left and right lane changes, a distinction that becomes particularly important when driving in a highway. Models have been trained and validated using a real dataset gathered online by using a high-tech demonstrator vehicle provided by Centro Ricerche Fiat (i.e., Fiat Research Center). Data, which refer to real driving conditions on a highway, have been collected by monitoring different drivers showing different behaviors. We address the problem of unbalanced data, typical of real data sets, and propose two prediction models based on Support Vector Machines and Random Forests. The results of our computational experiments show the validity of the approach with respect to state of the art models, both in terms of prediction accuracy and prediction time.

Autonomous vehicles (AVs) must share space with pedestrians, both in carriageway cases such as cars at pedestrian crossings and off-carriageway cases such as delivery vehicles navigating through crowds on pedestrianized high-streets. Unlike static obstacles, pedestrians are active agents with complex, interactive motions. Planning AV actions in the presence of pedestrians thus requires modelling of their probable future behavior as well as detecting and tracking them. This narrative review article is Part II of a pair, together surveying the current technology stack involved in this process, organising recent research into a hierarchical taxonomy ranging from low-level image detection to high-level psychological models, from the perspective of an AV designer. This self-contained Part II covers the higher levels of this stack, consisting of models of pedestrian behavior, from prediction of individual pedestrians' likely destinations and paths, to game-theoretic models of interactions between pedestrians and autonomous vehicles. This survey clearly shows that, although there are good models for optimal walking behavior, high-level psychological and social modelling of pedestrian behavior still remains an open research question that requires many conceptual issues to be clarified. Early work has been done on descriptive and qualitative models of behavior, but much work is still needed to translate them into quantitative algorithms for practical AV control.

The “classical” SAE LoA for automated driving can present several drawbacks, and the SAE-L2 and SAE-L3, in particular, can lead to the so-called “irony of automation”, where the driver is substituted by the artificial system, but is still regarded as a “supervisor” or as a “fallback mechanism”. To overcome this problem, while taking advantage of the latest technology, we regard both human and machine as members of a unique team that share the driving task. Depending on the available resources (in terms of driver’s status, system state, and environment conditions) and considering that they are very dynamic, an adaptive assignment of authority for each member of the team is needed. This is achieved by designing a technology enabler, constituted by the intelligent and adaptive co-pilot. It comprises (1) a lateral shared controller based on NMPC, which applies the authority, (2) an arbitration module based on FIS, which calculates the authority, and (3) a visual HMI, as an enabler of trust in automation decisions and actions. The benefits of such a system are shown in this paper through a comparison of the shared control driving mode, with manual driving (as a baseline) and lane-keeping and lane-centering (as two commercial ADAS). Tests are performed in a use case where support for a distracted driver is given. Quantitative and qualitative results confirm the hypothesis that shared control offers the best balance between performance, safety, and comfort during the driving task.

In this paper, we explore how MDP can be used as the framework to design and develop an Intelligent Decision Support System/Recommender System, in order to extend human perception and overcome human senses limitations (because covered by the ADS), by augmenting human cognition, emphasizing human judgement and intuition, as well as supporting him/her to take the proper decision in the right terms and time. Moreover, we develop Human-Machine Interaction (HMI) strategies able to make “transparent” the decision-making/recommendation process. This is strongly needed, since the adoption of partial automated systems is not only connected to the effectiveness of the decision and control processes, but also relies on how these processes are communicated and “explained” to the human driver, in order to achieve his/her trust.

Interest in autonomous vehicles (AVs) has significantly increased in recent years, but despite the huge research efforts carried out in the field of intelligent transportation systems (ITSs), several technological challenges must still be addressed before AVs can be extensively deployed in any environment. In this context, one of the key technological enablers is represented by the motion-planning and control system, with the aim of guaranteeing the occupants comfort and safety. In this paper, a trajectory-planning and control algorithm is developed based on a Model Predictive Control (MPC) approach that is able to work in different road scenarios (such as urban areas and motorways). This MPC is designed considering imitation-learning from a specific dataset (from real-world overtaking maneuver data), with the aim of getting human-like behavior. The algorithm is used to generate optimal trajectories and control the vehicle dynamics. Simulations and Hardware-In-the-Loop tests are carried out to demonstrate the effectiveness and computation efficiency of the proposed approach.

Autonomous driving is disrupting the automotive industry as we know it today. For this, fail-operational behavior is essential in the sense, plan, and act stages of the automation chain in order to handle safety-critical situations on its own, which currently is not reached with state-of-the-art approaches.The European ECSEL research project PRYSTINE realizes Fail-operational Urban Surround perceptION (FUSION) based on robust Radar and LiDAR sensor fusion and control functions in order to enable safe automated driving in urban and rural environments. This paper showcases some of the key exploitable results (e.g., novel Radar sensors, innovative embedded control and E/E architectures, pioneering sensor fusion approaches, AI-controlled vehicle demonstrators) achieved until its final year 3.

For automated and partially automated cars, there are new crucial questions to answer: “When should the driver or the automated system take control of the vehicle?" ; and also: “Can both control the vehicle together at the same time, or can this create potential conflicts?" . These are non-trivial issues because they depend on different conditions, such as the environment, driver’s state, vehicle capabilities, and fault tolerance, among others. This paper will describe a human-machine cooperation approach for collaborative driving maneuvers, developed in the EU funded project PRYSTINE. In particular, this study presents the work-in-progress and will focus attention on the proposed architecture design and the corresponding use case for testing.

The advancement of SAE Level 3 automated driving systems requires best practices to guide the development process. In the past, the Code of Practice for the Design and Evaluation of ADAS served this role for SAE Level 1 and 2 systems. The challenges of Level 3 automation make it necessary to create a new Code of Practice for automated driving (CoP-AD) as part of the public-funded European project L3Pilot. It provides the developer with a comprehensive guideline on how to design and test automated driving functions, with a focus on highway driving and parking. A variety of areas such as Functional Safety, Cybersecurity, Ethics, and finally the Human–Vehicle Integration are part of it. This paper focuses on the latter, the Human Factors aspects addressed in the CoP-AD. The process of gathering the topics for this category is outlined in the body of the paper. Thorough literature reviews and workshops were part of it. A summary is given on the draft content of the CoP-AD Human–Vehicle Integration topics. This includes general Human Factors related guidelines as well as Mode Awareness, Trust, and Misuse. Driver Monitoring is highlighted as well, together with the topic of Controllability and the execution of Customer Clinics. Furthermore, the Training and Variability of Users is included. Finally, the application of the CoP-AD in the development process for Human-Vehicle Integration is illustrated.

Driving automation is radically changing the role of the driver. The proliferation of driving assistance systems is increasingly transforming driver’s tasks from vehicle control operations to supervising activities. However, the process of turning the driver into a passenger is far from being accomplished. This paper describes an innovative interaction approach developed in the framework of EU funded project AutoMate. The overarching aim of AutoMate is to build a “TeamMate System”, in which the human and the automation cooperate with each other to achieve a safe, pleasant and efficient driving. Through an effective interaction, and by sharing perception, decision and action, they can negotiate specific behaviors and maneuvers in order to build a team based on trust. In order to measure the effectiveness of this concept, a driving simulator experiment has been conducted: findings suggest that the concept of Human-Machine Team can increase the trust in automation and improve the efficiency in specific conditions.