Enhanced fuzzy-MFC-based traction control system for electric vehicles. - In: IEEE Xplore digital library, ISSN 2473-2001, (2023), insges. 6 S.
Modern vehicles require the installation of motion control systems to ensure driving safety. In electric vehicles, these systems are convenient to be developed and applied due to the better response of the electric motor compared to the internal combustion engine. Therefore, the development of traction control systems for electric vehicles is of great interest to many researchers. In this study, a wheel slip control algorithm for electric vehicles is proposed by considering the vehicle as an equivalent inertial system. Based on the monotonicity of the algorithm, a fuzzy controller is also incorporated in the study so that the wheel slip control can adapt to the actual road conditions. Its performance is verified by comparative simulations with baseline anti-slip methods for different road conditions and vehicle velocities.
Applicability study of model-free reinforcement learning towards an automated design space exploration framework. - In: 2023 IEEE Symposium Series on Computational Intelligence (SSCI), (2023), S. 525-532
Design space exploration is a crucial aspect of engineering and optimization, focused on identifying optimal design configurations for complex systems with a high degree of freedom in the actor set. It involves systematic exploration while considering various constraints and requirements. One of the key challenges in design space exploration is the need for a control strategy tailored to the particular design. In this context, reinforcement learning has emerged as a promising solution approach for automatically inferring control strategies, thereby enabling efficient comparison of different designs. However, learning the optimal policy is computationally intensive, as the agent determines the optimal policy through trial and error. The focus of this study is on learning a single strategy for a given design and scenario, enabling the evaluation of numerous architectures within a limited time frame. The study also highlights the importance of plant modeling considering different modeling approaches to effectively capture the system complexity on the example of vehicle dynamics. In addition, a careful selection of an appropriate hyperparameter set for the reinforcement learning algorithm is emphasized to improve the overall performance and optimization process.
Extending teleoperated driving using a shared X-in-the-loop environment. - In: Engineering for a changing world, (2023), 3.1.092, S. 1-14
The strong progress in modern vehicle system technology requires new methodological approaches for the development and validation of new vehicle systems. In particular, due to increasing automation, classical development methods and testing scenarios need to be evolved. Consequently, the publication focuses on an extension of teleoperated driving by the X-in-the-loop (XIL) approach. Within this framework, the classical concept based on VPN-LTE networking is analyzed and discussed at first. With this implementation, the remote control of a real vehicle is presented based on the use of a dynamic driving simulator. Especially for the development and validation of such concepts, an extension with the XIL methodology can improve this process. For this reason, the architecture of teleoperated driving is subsequently extended by networking with additional system components. The feasibility, the functionalities as well as the challenges that arise with such an extension based on the XIL methodology are shown. Within the scope of this study, the achieved transmission times for the control variables and for the video data stream are demonstrated. Based on different driving maneuvers, the achievable repeatability is discussed.
Self-adapting motion cueing algorithm based on a kinematics reference model. - In: Engineering for a changing world, (2023), 3.1.085, S. 1-9
Due to a number of advantages over traditional development methods, the importance of dynamic driving simulators in automotive research and development has grown continuously in recent years. Motion simulation via motion cueing algorithms contributes significantly to the driving experience and provides the driver with valuable information about the current driving dynamics. The adaptation and tuning process of these algorithms can be difficult and timeconsuming tasks. It needs to be repeated after changes to the vehicle or driving scenario. This paper discusses and presents an adaptive or rather self-adapting motion cueing algorithm (MCA) concept. The approach is based on the integration of a kinematic reference model to dynamically and adaptively adjust the motion behavior dynamically and adaptively. This concept allows to reduce the parameter tuning effort drastically in long term, since the algorithm can adapt itself to different conditions such as vehicle type, driving situation, or driver behavior. In the following, the proposed algorithm structure is explained and illustrated. The advantages of the proposed MCA are demonstrated by an experimental comparison with a classical algorithm. Thereby it is shown how a self-adaptation of the algorithm can proceed and how to avoid violation of workspace boundaries.
Letter from the special issue editors : special issue on modern vehicle dynamics control systems. - In: SAE International journal of vehicle dynamics, stability, and NVH, ISSN 2380-2170, Bd. 7 (2023), 3, S. 265-267
Estimation of the tire-road interaction forces by using Pacejka’s formulas with combined slips and camber angles. - In: SAE technical papers, ISSN 2688-3627, (2023), SAE technical paper 2023-01-0646
The growing market demand for highly automated and autonomous vehicles and the need to equip vehicles with ever higher standards of comfort, safety and performance requires knowledge of physical quantities that are often difficult or expensive to measure directly. The absence of direct sensors, the difficulty of implementation, and their cost have led researchers to identify alternative solutions that allow estimating the physical quantity of interest by aggregating other available information. The interaction forces between tire and road are among the most significant. Given that the dynamics of a vehicle are strongly linked to the forces exchanged between the tire and the road, their knowledge is fundamental in the development of control systems aimed at improving performance in terms of handling, road holding or comfort. This paper presents a new technique for the estimation of tire-road interaction forces based on the integration of models and measures. A Central Difference Kalman filter was applied to a Double Track Model. The non-linear Kalman filter allowed us to handle the non-linearity of the system. The tire-road interaction was modelled through Pacejka's magic formulas that into account the combined longitudinal and lateral slips and the camber angle. This version made it possible to carry out complex and realistic manoeuvres. The realized estimator also considers the influence of lateral and longitudinal load transfers and aerodynamic forces in the three spatial directions. The Camber angle used in this observer was estimated through neural networks. The measures used are longitudinal velocity, yaw rate, longitudinal slip and wheel steering angles.
Methodologic assessment of brake-by-wire system modelling with regard to accuracy, model complexity and optimization efforts. - In: SAE technical papers, ISSN 2688-3627, (2023), SAE technical paper 2023-01-0666
Brake-by-wire systems are an innovative and important component of modern high-performance and also electrified vehicles. Due to their decoupled architecture, they enable driver-independent vehicle dynamics control (e.g., brake torque blending) and easy integration of assistance functionalities (e.g. Emergency Brake Assist (EBA)). On the other hand, the development of these functions can cause high costs and development effort, and testing can be critical in case of improper gain tuning. Therefore, already in the concept phase, a large part of the testing is shifted to virtual environments and simulations that allow safe and reproducible experiments without damage. Therefore, suitable and reliable models are needed to represent reality as accurately as possible. This paper deals with the modelling of a purely electrohydraulic brake-by-wire system and a hybrid system with electrohydraulic brakes on the front axle and electromechanical brakes on the rear axle. For comparison, both an experimental approach based on a second-order transfer function and an analytical model are used. These approaches are then evaluated in terms of their accuracy and reliability using real measurements in different dynamic test setups. Finally, it is shown how accurate the approaches are and what advantages can be achieved by using the different methods for system modelling.
Road parameter estimation with drone-vehicle communication. - In: SAE technical papers, ISSN 2688-3627, (2023), SAE technical paper 2023-01-0664
The presented study is dedicated to the technology supporting vehicle state estimation and motion control with a concept drone, which helps the vehicle in sensing the surroundings and driving conditions. This concept allows also extending the functionality of the sensors mounted on the vehicle by replacing or including additional parameter observation channels. The paper discusses the feasibility of such a drone-vehicle interaction as well as demonstrates several design configurations. In this regard, the paper presents a general description of the proposed drone system that assists the vehicle and describes an experiment in measuring the profile of the road with a range sensor. The results obtained in the experiment are described in terms of the accuracy to be achieved using the drone and are compared with other studies, which use the methods of estimation from the sensors mounted on the vehicle. The proposed measurement concept can be applied to a large number of vehicle systems such as adaptive cruise control, active or semi-active suspension, and wheel slip control. The road profile is captured in real-time by a drone, and the telemetry data is processed by the host computer.
Deep learning for the estimation of the longitudinal slip ratio. - In: 2023 IEEE International Workshop on Metrology for Automotive, (2023), S. 193-198
In a road vehicle, the interaction forces between tire and road are strongly influenced by the longitudinal slip ratio. This kinematic quantity, therefore, represents one of the most important in the study of vehicle dynamics. The real-time knowledge of this quantity can allow the estimation of the interaction forces and the development of control systems to improve safety and handling. In particular, Anti-lock Braking Systems (ABS) and Traction Control Systems (TCS). Direct measurements of this quantity would require the insertion of sensors inside the tire, with consequent manufacturing complexity and increased costs. This paper proposes an estimate of the longitudinal slip ratio based on other easily measurable or estimable quantities. This estimator makes use of Neural Networks and is based on preliminary physical considerations.
Nutzung verschiedener Methoden des maschinellen Lernens zur Regelung von Fahrzeugsystemen. - Ilmenau : Universitätsbibliothek, 2023. - 1 Online-Ressource (XX, 219 Blätter)
Technische Universität Ilmenau, Dissertation 2023
Die Steuerung fahrzeugtechnischer und nicht-fahrzeugtechnischer Systeme wird mit zunehmender Vernetzung und Digitalisierung in ihrer Komplexität weiter zunehmen. Die an vielen Stellen manuelle Parametrierung dieser Systeme stößt dabei zunehmend bezüglich Zeit und Kosten an ihre Grenzen. Seit 2010, als der Grafikkartenhersteller NVIDIA mit dem CUDA Toolkit eine Möglichkeit erschuf, Matrixrechenopperationen schnell und effizient auf Graphic Processing Units (GPUs) durchführen zu können, erhielt das maschinelle Lernen (konkreter - Deep Learnings) neue Möglichkeiten sich zu entwickeln. Seitdem hat sich die Nutzung künstlicher neuronaler Netze als Funktionsapproximator für verschiedene Bereiche der Wissenschaft und Technik enorm weiterentwickelt. Dabei wird es möglich real gemessene Daten zu verwenden, um Systemabbildungen (Simulationen) zu erzeugen, ohne den Umweg über eine analytische und/oder numerische Methode gehen zu müssen (welche nach Modellbildung immer mit realen Messwerten validiert werden muss). Ziel ist es ein System ohne Zuhilfenahme menschlicher Arbeitskraft vollautomatisch Regeln zu können. So soll es möglich sein, neu entwickelte und produzierte Maschinen schnell und effizient mithilfe von Aktuatoren und elektronischen Steuerelementen nach spezifischen Vorgaben einsatzfähig zu bekommen. Die vorliegende Arbeit beschreibt eine Methodik, durch überwachtes Lernen (Supervised Learning) datenbasierte Modelle, sogenannte digitale Zwillinge zu entwickeln. Mithilfe künstlicher neuronaler Netze werden stationäre und instationäre Modelle entwickelt und es werden statistischen Methoden zur Datenaufbereitung erläutert. Digitale Zwillinge sind nötig, da reale Umgebungen wie Prüfstände oder Maschinen teuer im Betrieb und langsam bei der Zustandsraumdurchschreitung sind. Nach erfolgreicher Generierung eines digitalen Zwillings wird bestärkendes Lernen (Reinforcement Learning) genutzt, autonom komplexe und mehrdimensionale Zustandsräume zu untersuchen und selbstständig Regelstrategien zu entwickeln. In Summe werden also digitale Zwillinge des Supervised Learnings auf Basis real gemessener Daten als Simulationsumgebungen genutzt, um mithilfe der Algorithmen des Reinforcement Learnings, selbstständig arbeitende Agenten zu trainieren. Drei Systeme werden dabei abgebildet. Ein Rohrsystem, ein semiaktiver Stoßdämpfer und ein Abgasturbolader. Am Beispiel des Abgasturboladers wird gezeigt, wie gut Reinforcement Learning im Vergleich zum klassischen PID Regler performt, dass die Methodik prinzipiell funktioniert, es aber noch weiterer Entwicklung bedarf, um tatsächlich direkt Verwendbare Regelalgorithmen zu erhalten, die denen klassischer Vorgehensweisen von heute ebenbürtig (oder gar überlegen) sind.