Browse Topic: Artificial intelligence (AI)

Items (2,026)
Find
Applying Real World Traffic Data on Machine Learning Methods for Graph Based Cooperation Strategies of Automated Vehicles2025-01-0286To be published on 07/02/2025
With the increasing distribution of smart mobility systems, automated & connected vehicles are more and more interacting with each other and with smart infrastructure using V2X-communication. Hereby, the vehicles' position, driving dynamics data, or driving intention are exchanged. Previous research has explored graph-based cooperation strategies for automated vehicles in mixed traffic environments based on current V2X-communication standards. Thereby, the focus is set on cooperation optimization and maneuver negotiation. These strategies can be implemented through both centralized and decentralized computational approaches and are conflict-free by design. To enhance these previously established cooperation models, real-world traffic data is used to derive vehicle trajectories, providing a more accurate representation of actual traffic scenarios in order to enhance the practical application of the described methodology. Additionally, machine learning algorithms are employed to train
Flormann, Maximilian, Meyer, Felix, Henze, Roman
Technical Paper
Conception and Experience of Non-driving Related Activities in Semi-autonomous Vehicles2025-01-0260To be published on 07/02/2025
While semi-autonomous driving (SAE level 3 & 4) is already partially a reality, the driver still needs to take over driving upon notice. Hence, the cockpit cannot be designed freely to accommodate spaces for non-driving related activities. In the following use case, a mobile workplace is created by integrating a translucent acrylic glass pane into the cockpit and introducing joystick steering of the car. By using the technology Virtual Desktop 1, which is a software layer, any desktop application can be represented freely transformable on arbitrary physical and virtual surfaces. Thus, a complete Windows environment can be distributed across all curved and flat surfaces of an interior. The concept is further enhanced by a voice-driven generative AI which helps to summarize documents. A physical and a virtual demonstrator are created to experience and assess the mobile workspace, the well-being of the driver, external influences, and psychological aspects. The physical demonstrator is a
Beutenmüller, Frank, Reining, Nine, Rosenstiel, Reto, Schmidt, Maximilian, Layer, Selina, Bues, Matthias, Mendonca, Daisy
Technical Paper
Prediction of hazardous gaseous emissions from a gasoline engine during cold starts using machine learning methods2025-01-0321To be published on 07/02/2025
Internal combustion engines generate higher exhaust emissions of hazardous gases during the initial minutes after engine start. Experimental data from a state-of-the-art turbo-charged 3-cylinder, 999 cc gasoline engine are used to predict cold start emissions using two Machine Learning (ML) models: a Multilayer Perceptron (MLP) which is a fully connected neural network and an Encoder-Decoder Recurrent Neural Network (ED-RNN). Engine parameters and various temperatures are used as input for the models and NOx (Nitrogen Oxides), CO (Carbon monoxide) and unburned hydrocarbon (UHC) emissions are predicted. The dataset includes time series recordings from the Worldwide harmonized Light-duty vehicles Test Cycle (WLTC) and four Real Diving Emissions (RDE) cycles at ambient and initial engine temperatures ranging from -20 °C to +23 °C. In total, 21 cases are considered, consisting of eight different ambient temperatures and five distinct driving cycles. Each case consists of a sequence of 2500
Mangipudi, Manoj, Denev, Jordan A., Bockhorn, Henning, Trimis, Dimosthenis, Koch, Thomas, Debus, Charlotte, Götz, Markus, Zirwes, Thorsten, Hagen, Fabian P., Tofighian, Hesam, Wagner, Uwe, Braun, Samuel, Lanzer, Theodor, Knapp, Sebastian M.
Technical Paper
An LLM-Based Multi-Turn Task-Oriented Dialogue System for Machine Learning Algorithm Selection in Data Mining2025-01-0289To be published on 07/02/2025
For the systematic application of machine learning during data mining in product development processes, selecting a suitable algorithm is crucial for success. During an empirical study in the automotive industry, a team applying data mining to develop battery systems for battery electric vehicles was accompanied. Here, it could be observed that data mining tasks are often unique during product development processes and can differ in boundary conditions. Depending on these tasks, suitable machine learning algorithms must be selected. Because of the variety of machine learning paradigms, problems, and algorithms, it is often hard to select a suitable algorithm, especially for inexperienced data miners. This paper presents a large language model (LLM)-based, multi-turn, task-oriented dialogue system to support data miners in selecting machine learning algorithms that are suitable for their specific data mining tasks. This approach, called “Algorithm Selection Assistant” (ASA), enables
Hörtling, Stefan, Bause, Katharina, Albers, Albert
Technical Paper
Reducing Traffic Congestion with Reinforcement Learning-Driven Signal Control Systems2025-01-0283To be published on 07/02/2025
The road network is a critical component of modern urban mobility systems, with signalized traffic intersections playing a pivotal role. Traditionally, traffic light phase timings and duration at intersections are designed by transportation engineers using historical traffic data. Some modern intersections employ trigger-based mechanisms to improve traffic flow; however, these systems often lack global awareness of traffic conditions across multiple intersections within a network. With the increasing availability of traffic data and advancements in machine learning, traffic light systems can be enhanced by modeling them as agents operating in an environment. This paper proposes a Reinforcement Learning (RL)-based approach for multi-agent traffic light systems within a simulation environment. The simulation is calibrated using real-world traffic data, enabling RL agents to learn effective control strategies based on realistic scenarios. A key advantage of using a calibrated simulation
Kalra, Vikhyat, Tulpule, Punit, Giuliani, Pio
Technical Paper
Wheel Hop Detection in High-Torque Vehicles: A Machine Learning Approach with Focus on Data Coverage and Model Generalization2025-01-0271To be published on 07/02/2025
Wheel hop is a vibration phenomenon that occurs in cars with a high torque during high accelerations from low speed. Drivers perceive wheel hop as vertical oscillations in the vehicle frame. In severe cases, it can lead to component damage or deformation. Therefore, the affected vehicles must be safeguarded against these vibrations through safe component design and additional software-based functions. Conventional software-based solutions such as the traction control system (TCS) are not explicitly designed to prevent wheel hop. These systems often intervene too late and apply abrupt torque adjustments that compromise driving comfort. Based on observations on real measurements the TCS intervention in some cases leads to an amplification of the wheel hop effect. To address these challenges, we propose a novel approach based on machine learning (ML) for the early detection of wheel hop by leveraging real-time measurements from a rear-wheel drive vehicle. Unlike conventional methods, our
Chehoudi, Moatez, Moisidis, Ioannis, Sailer, Marc, Peters, Steven
Technical Paper
Leveraging LLM-based AI Agents for Boosting the Vehicle Testing Process2025-01-0300To be published on 07/02/2025
The validation process in research and development involves several complex stages, including test requests, planning, execution, and the analysis and evaluation of results. In the automotive domain, compliance with regulatory standards, such as those required for Euro 7 homologation, adds an additional layer of complexity. Implementing these regulations into operational validation workflows and ensuring their seamless integration with supporting tools remains a significant challenge. Recent advancements in Large Language Models (LLMs) have introduced innovative use cases across various domains. In particular, AI agents powered by LLMs demonstrate immense potential by autonomously performing complex tasks while utilizing user-defined tools. This capability extends far beyond traditional applications like knowledge management or text generation typically associated with LLMs. In this paper, we explore how a modern AI agent can be developed and integrated into existing IT tools for test
Unterschütz, Stefan, Hansen, Björn
Technical Paper
Methodology for Real-World Automated Function Development: From Virtual to On-Vehicle Implementation2025-01-0288To be published on 07/02/2025
The automotive industry is increasingly facing challenges stemming from growing system complexities, shortened development cycles, and the demand for rapid time-to-market transitions. Reinforcement learning (RL) has emerged as a promising approach to developing advanced control functions due to its adaptive and autonomous nature. The technique has already demonstrated its viability in virtualised X-in-the-Loop (XiL) environments. However, its application to real-world vehicle systems is inhibited by safety concerns, real-time constraints, and the integration into established software toolchains. This paper introduces a comprehensive methodology for developing practical control functions with RL: starting in a virtual environment, training then transitions to a Hardware-in-the-Loop (HiL) setup, and ultimately proceeds to a real vehicle. Utilising the open-source framework LExCI, the proposed approach facilitates seamless training across multiple development stages and showcases RL’s
Badalian, Kevin, Picerno, Mario, Lee, Sung-Yong, Schaub, Joschka, Andert, Jakob
Technical Paper
Development and Evaluation of a Combined Driveline Oscillation and Traction Controller Using Model Predictive Control and Reinforcement Learning: A Comparative Case Study2025-01-0291To be published on 07/02/2025
In electric vehicles, the control of driveline oscillations and tire trac- tion is critical for guaranteeing driver comfort and safety. Yet, achiev- ing sufficient driveline control performance remains challenging in the presence of rapidly varying road conditions. Two promising avenues for further improving driveline control are adaptive model predictive control (MPC) and model-based reinforcement learning (RL). We de- rive such controllers from the same non-linear vehicle model and vali- date them through pre-defined test scenarios. The MPC approach em- ploys input and output trajectory tracking with soft constraints to en- sure feasible control actions even in the presence of constraint viola- tions and is further supported by a Kalman filter for robust state estima- tion and prediction. In contrast, the RL controller leverages the model- based DreamerV3 algorithm to learn control policies autonomously, adapting to different road conditions without relying on external infor- mation
Uhl, Ramón Tamino, Schüle, Isabel, Ludmann, Laurin, Geist, A. René
Technical Paper
Leveraging AI for Automated Code Generation from Systems Engineering Specifications2025-01-0295To be published on 07/02/2025
The increasing complexity of modern vehicles and the automotive industry's shift towards Software Defined Vehicles (SDVs) require innovative solutions to streamline development processes. Traditional methods of software development often struggle to meet the demands for agility, scalability, and precision in this context. In response, this paper presents a noval approach utilizing Artificial Intelligence (AI), specifically Large Language Models (LLMs), to automate the generation of executable code directly from Systems Engineering (SE) specifications. This novel approach aims to transform how SE requirements are converted into implementation-ready code, reducing the inefficiencies and potential errors associated with manual translation. LLMs trained on domain-specific data are capable of interpreting complex requirements, managing dependencies, and generating consistent and accurate code. By integrating LLMs into the automotive software pipeline, companies can improve productivity
Padubrin, Marcel, Reuss, Hans-Christian, Brosi, Frank, Menz, Leonhard, Guerocak, Erol
Technical Paper
Research on Combustion Characteristics of Flash-Boiling Spray in Cylinders Based on Machine Learning Methods2025-01-0205To be published on 06/16/2025
With the increasing number of vehicles in operation, exhaust emissions from engines have exerted negative impacts on ecological environments, prompting researchers to actively pursue cleaner and more efficient in-cylinder combustion strategies. Flash-boiling spray technology, capable of generating superior fuel atomization under relatively low injection pressures, has emerged as a promising approach for achieving performance breakthroughs in gasoline direct injection (GDI) engines. While current research primarily focuses on morphological characterization and mechanistic analysis of flashboiling spray, there remains insufficient understanding of flame development characteristics under flash boiling spray conditions within engine cylinders. This study systematically investigates the combustion characteristics of TPRF and PRF fuels under both subcooled and flash-boiling spray conditions through the integration of image processing and machine learning methodologies. Experimental
Zhang, Weixuan, Shahbaz, Muhammad, Cui, Mingli, Li, Xuesong, Xu, Min
Technical Paper
Brake Noise CAE Enhancement with the Synergy between Design of Experiments and Machine Learning2025-01-012405/05/2025
The digitalization of industrial systems has led to increased data availability. Machine learning (ML) methodologies are now commonly used for data analysis in industrial contexts. Not all contexts have abundant data; sometimes data collection can be scarce or expensive. Design of Experiments (DOE) is a technique that provides an informative dataset for ML analysis when data are limited. It involves systematically designing experiments to collect relevant data points with regression models. Disc brake noise is a challenging problem in vehicle noise, vibration, and harshness (NVH). Different noise events occur under various operating conditions and across frequencies (1-16 kHz). To enhance computer-aided engineering (CAE) techniques for brake noise, ML is used to generate additional data. Sequential experimentation in DOE aligns well with ML’s ability to continuously learn and improve as more data become available. DOE is applied in CAE to collect data for training ML models. ML helps
Song, Gavin, Sridhar, Gurupriya, Vlademar, Michael, Venugopal, Narayana
Technical Paper
A Comparative Analysis of Fault Diagnosis by Vibration Signals for Critical Gear Components in Electric Vehicle Motor Testing Machines Using Machine Learning Algorithms2025-01-004005/05/2025
Electric vehicles (EVs) are shaping the future of mobility, with drive motors serving as a cornerstone of their efficiency and performance. Motor testing machines are essential for verifying the functionality of EV motors; however, flaws in testing equipment, such as gear-related issues, frequently cause operational challenges. This study focuses on improving motor testing processes by leveraging machine learning and vibration signal analysis for early detection of gear faults. Through statistical feature extraction and the application of classifiers like Wide Naive Bayes and Coarse Tree, the collected vibration signals were categorized as normal or faulty under both loaded (0.275 kW) and no-load conditions. A performance comparison demonstrated the superior accuracy of the wide neural networks algorithm, achieving 95.3%. This methodology provides an intelligent, preventive maintenance solution, significantly enhancing the reliability of motor testing benches.
S, Ravikumar, Sharik, N, Syed, Shaul, V, Muralidharan, D, Pradeep Kumar
Technical Paper
Leveraging Vision Language Models and Prompt Engineering for Bearing Defect Classification from Vibration Signals2025-01-012705/05/2025
Bearings are fundamental components in automotive systems, ensuring smooth operation, efficiency, and longevity. They are widely used in various automotive systems such as wheel hubs, transmissions, engines, steering systems etc. Early detection of bearing defects during End-of-Line (EOL) testing and operational phases is crucial for preventive maintenance, thereby preventing system malfunctions. In the era of Industry 4.0, vibrational, accelerometer, and other IoT sensors are actively engaged in capturing performance data and identifying defects. These sensors generate vast amounts of data, enabling the development of advanced data-driven applications and leveraging deep learning models. While deep learning approaches have shown promising results in bearing fault diagnosis, they often require extensive data, complex model architectures, and specialized hardware. This study proposes a novel method leveraging the capabilities of Vision Language Models (VLMs) and Large Language Models
Chandrasekaran, Balaji, Cury, Rudoniel
Technical Paper
High-Fidelity NVH Model Development for Electric Motors Using Deep Learning and Machine Learning Algorithms2025-01-012305/05/2025
Electric motor whine is a significant source of noise in electric vehicles (EVs). To improve the noise, vibration, and harshness (NVH) performance of electric propulsion systems, it is essential to develop a physics-based, high-fidelity stator model. In this study, a machine learning (ML) model is developed using an artificial neural network (ANN) method to accurately characterize the material properties of the copper winding, varnish, and orthotropic stator laminate structure. A design of experiments (DOE) approach using Latin hypercube sampling of parameters is implemented after evaluating alternative surrogate models. A finite element (FE) model is constructed using the nominal stator design parameters to train the ANN model using 121 DOE variables and 72,000 data points. The ML-trained ANN model is then verified to predict the driving point frequency response function (FRF) spectrum with reasonable accuracy. Subsequently, modal tests are conducted on the electric stator, and the
Rao, Bhyri Rajeswara, GSJ, Gautam, He, Song
Technical Paper
Active Noise Control with Head Tracking System2025-01-001205/05/2025
In active noise control, the control region size (same meaning as zone of control) decreases as the frequency increases, so that even a small moving of the passenger's head causes the ear position to go out of the control region. To increase the size of the control region, many speakers and microphones are generally required, but it is difficult to apply it in a vehicle cabin due to space and cost constraints. In this study, we propose moving zone of quiet active noise control technique. A 2D image-based head tracking system captured by a camera to generate the passenger's 0head coordinates in real time with deep learning algorithm. In the controller, the control position is moved to the ear position using a multi-point virtual microphone algorithm according to the generated ear position. After that, the multi-point adaptive filter training system applies the optimal control filter to the current position and maintains the control performance. Through this study, it is possible to
Oh, ChiSung, Kang, Jonggyu, Kim, Joong-Kwan
Technical Paper
Data-Driven Machine Learning Approach for Rattle Risk Mitigation2025-01-012605/05/2025
In the modern automotive industry, squeak and rattle issues are critical factors affecting vehicle perceived quality and customer satisfaction. Traditional approaches to predicting and mitigating these problems heavily rely on physical testing and simulation technologies, which can be time-consuming and resource-intensive, especially for larger models. In this study, a data-driven machine learning approach was proposed to mitigate rattle risks more efficiently. This study evaluated a floor console model using the traditional simulation-based E-line method to pinpoint high-risk areas. Data generation is performed by varying material properties, thickness, and flexible connection stiffness using the Hammersley sampling algorithm, creating a diverse and comprehensive dataset for generating a machine learning (ML) model. Utilizing the dataset, the top contributing variables were identified for training the ML models. Various machine-learning models were developed and evaluated, and the
Parmar, Azan, Rao, Sohan, Reddy, Hari Krishna
Technical Paper
Navigating Evolving Vehicle NVH Challenges with Application of Emerging Technologies in Artificial Intelligence and Machine Learning2025-01-013005/05/2025
The multifaceted, fast-paced evolution in the automotive industry includes noise and vibration (NVH) behavior of products for regulatory requirements and ever-increasing customer preferences and expectations for comfort. There is pressing need for automotive engineers to explore new and advanced technologies to achieve a ‘First Time Right’ product development approach for NVH design and deliver high-quality products in shorter timeframes. Artificial Intelligence (AI) and Machine Learning (ML) are trending transformative technologies reshaping numerous industries. AI enables machines to replicate human cognitive functions, such as reasoning and decision-making, while ML, a branch of AI, employs algorithms that allow systems to learn and improve from data over time. The purpose of the paper is to show an approach of using machine learning techniques to analyze the impact of variations in structural design parameters on vehicle NVH responses. The study begins by executing the Design of
Miskin, Atul R., Parmar, Azan, Raj, Sonia, Himakuntla, Uma Maheswar
Technical Paper
Mode Shape Identification Using Graph Neural Networks for Vehicle Structure Design2025-01-013105/05/2025
This paper introduces a novel, automated approach for identifying and classifying full vehicle mode shapes using Graph Neural Networks (GNNs), a deep learning model for graph-structured data. Mode shape identification and naming refers to classifying deformation patterns in structures vibrating at natural frequencies with systematic naming based on the movement or deformation type. Many times, these mode shapes are named based on the type of movement or deformation involved. The systematic naming of mode shapes and their frequencies is essential for understanding structural dynamics and “Modal Alignment” or “Modal Separation” charts used in Noise, Vibration and Harshness (NVH) analysis. Current methods are manual, time-consuming, and rely on expert judgment. The integration of GNNs into mode shape classification represents a significant advancement in vehicle modal identification and structure design. Results demonstrate that GNNs offer superior accuracy and efficiency compared to
Tohmuang, Sitthichart, Swayze, James L., Fard, Mohammad, Fayek, Haytham, Marzocca, Piergiovanni, Bhide, Sanjay, Huber, John
Technical Paper
Designing Rubber Flaps for Resonance Management: High-Frequency Tuning in Electric Vehicles Using AI/ML Approaches2025-01-012505/05/2025
High-frequency whine noise in electric vehicles (EVs) is a significant issue that impacts customer perception and alters their overall view of the vehicle. This undesirable acoustic environment arises from the interaction between motor polar resonance and the resonance of the engine mount rubber. To address this challenge, the proposal introduces an innovative approach to predicting and tuning the frequency response by precisely adjusting the shape of rubber flaps, specifically their length and width. The approach includes the cumulation of two solutions: a precise adjustment of rubber flap dimensions and the integration of ML. The ML model is trained on historical data, derived from a mixture of physical testing conducted over the years and CAE simulations, to predict the effects of different flap dimensions on frequency response, providing a data-driven basis for optimization. This predictive capability is further enhanced by a Python program that automates the optimization of flap
Hazra, Sandip, Khan, Arkadip
Technical Paper
Condition Monitoring of High-Speed Rotating Machinery Using Digital Twins & Machine Learning Techniques2025-01-012805/05/2025
In the era of Industry 4.0, the maintenance of factory equipment is evolving with new systems using predictive or prescriptive methods. These methods leverage condition monitoring through digital twins, Artificial Intelligence, and machine learning techniques to detect early signs of faults, types of faults, locations of faults, etc. Bearings and gears are among the most common components, and cracking, misalignment, rubbing, and bowing are the most common failure modes in high-speed rotating machinery. In the present work, an end-to-end automated machine learning-based condition monitoring algorithm is developed for predicting and classifying internal gear and bearing faults using external vibration sensors. A digital twin model of the entire rotating system, consisting of the gears, bearings, shafts, and housing, was developed as a co-simulation between MSC ADAMS (dynamic simulation tool) and MATLAB (Mathematical tool). The gear and bearing models were developed mathematically, while
Rastogi, Sarthak, Singhal, Srijan, Ahirrao, Sachin, Milind, T. R.
Technical Paper
Application of AI/ML in Hydrobush Tuning to Enhance Overall Value Proposition2025-01-013205/05/2025
In the highly competitive automotive industry, optimizing vehicle components for superior performance and customer satisfaction is paramount. Hydrobushes play an integral role within vehicle suspension systems by absorbing vibrations and improving ride comfort. However, the traditional methods for tuning these components are time-consuming and heavily reliant on extensive empirical testing. This paper explores the advancing field of artificial intelligence (AI) and machine learning (ML) in the hydrobush tuning process, utilizing algorithms such as random forest, artificial neural networks, and logistic regression to efficiently analyze large datasets, uncover patterns, and predict optimal configurations. The study focuses on comparing these three AI/ML-based approaches to assess their effectiveness in improving the tuning process. A case study is presented, evaluating their performance and validating the most effective method through physical application, highlighting the potential
Hazra, Sandip, Khan, Arkadip Amitava
Technical Paper
Leveraging User Feedback to Optimize the Tuning of EVSE Models2025-01-002005/05/2025
The implementation of active sound design models in vehicles requires precise tuning of synthetic sounds to harmonize with existing interior noise, driving conditions, and driver preferences. This tuning process is often time-consuming and intricate, especially facing various driving styles and preferences of target customers. Incorporating user feedback into the tuning process of Electric Vehicle Sound Enhancement (EVSE) offers a solution. A user-focused empirical test drive approach can be assessed, providing a comprehensive understanding of the EVSE characteristics and highlighting areas for improvement. Although effective, the process includes many manual tasks, such as transcribing driver comments, classifying feedback, and identifying clusters. By integrating driving simulator technology to the test drive assessment method and employing machine learning algorithms for evaluation, the EVSE workflow can be more seamlessly integrated. But do the simulated test drive results
Hank, Stefan, Kamp, Fabian, Gomes Lobato, Thiago Henrique
Technical Paper
Image Processing-Based Chip Detection by KIZKI Algorithm and Its Optimization2025-01-016805/02/2025
The process of producing aircraft parts involves the drilling of aluminum alloys. This creates a large amount of chips, which are removed using air, but sometimes they still remain within the holes. This is checked by inspectors through visual inspection. However, the quality of human inspection varies based on skill level and fatigue. Thus, image-based inspection should be used to stabilize and further improve inspection quality. This study aims to build a framework for chip detection based on image processing. Taking into account on-site implementation, the system must have low installation and running costs and be standalone. Therefore, we adopt the KIZKI algorithm, which satisfies these conditions. KIZKI means awareness in Japanese. This is a model of human peripheral vision and saccades. It does not require training like AI and can achieve high-speed and high-performance detection using a low-performance computer. In other words, there is no need for a computer with an expensive
Iinuma, Marin, Sato, Junya, Tsuji, Masahiko
Technical Paper
Approach to Optimize the Airplane Boarding Process by Using Image Recognition and Stochastic Simulation to Predict Overhead Bin Fill Levels2025-01-016405/02/2025
In single-aisle aircraft, the available storage space for carry-on baggage is inherently limited. When the aircraft is fully booked, it often results in insufficient overhead bin space, necessitating last-minute gate-checking of carry-on items. Such disruptions contribute to delays in the boarding process and reduce operational efficiency. A promising approach to mitigate this issue involves the integration of computer vision technologies with an appropriate data storage system and stochastic simulation to enable accurate and supportive predictions that enhance planning, reduce uncertainty, and improve the overall boarding process. In this work, the YOLOv8 image recognition algorithm is used to identify and classify each passenger’s carry-on baggage into predefined categories, such as handbags, backpacks, and suitcases. This classified data is then linked to passenger information stored in a NoSQL database MongoDB, which includes seat assignments and the number of carry-on items
Bergmann, Jacqueline, Hub, Maximilian
Technical Paper
Enhancing Efficiency in Manufacturing through Automated AI Agent-Based Bearing Defect Diagnostic System2025-01-015805/02/2025
Industrial bearings are critical components in aerospace, industrial, and automotive manufacturing, where their failures can result in costly downtime. Traditional fault diagnosis typically depends on time-consuming on-site inspections conducted by specialized field engineers. This study introduces an automated Artificial Intelligence virtual agent system that functions as a maintenance technician, empowering on-site personnel to perform preliminary diagnoses. By reducing the dependence on specialized engineers, this technology aims to minimize downtime. The Agentic Artificial Intelligence system leverages agents with the backbone of intelligence from Computer Vision and Large Language Models to guide the inspection process, answer queries from a comprehensive knowledge base, analyze defect images, and generate detailed reports with actionable recommendations. Multiple deep learning algorithms are provisioned as backend API tools to support the agentic workflow. This study details the
Chandrasekaran, Balaji
Technical Paper
Machine Learning-Based Vision Mapping and Planning for Flexibility in High-Mix Robotic Manufacturing2025-01-015705/02/2025
Industries that require high-accuracy automation in the creation of high-mix/low-volume parts, such as aerospace, often face cost constraints with traditional robotics and machine tools due to the need for many pre-programmed tool paths, dedicated part fixtures, and rigid production flow. This paper presents a new machine learning (ML) based vision mapping and planning technique, created to enhance flexibility and efficiency in robotic operations, while reducing overall costs. The system is capable of mapping discrete process targets in the robot work envelope that the ML algorithms have been trained to identify, without requiring knowledge of the overall assembly. Using a 2D camera, images are taken from multiple robot positions across the work area and are used in the ML algorithm to detect, identify, and predict the 6D pose of each target. The algorithm uses the poses and target identifications to automatically develop a part program with efficient tool paths, including
Langan, Daniel, Hall, Michael, Goldberg, Emily, Schrandt, Sasha
Technical Paper
Aerospace Developments: Vol. 3EPRCOMPV09202404/30/2025
Muelaner, Jody Emlyn, Moran, Matthew, Phillips, Paul
Reference
Traffic Flow Prediction Based on the Optuna–CNN–BiLSTM–GBRT Model12-09-01-000104/24/2025
Ma, Changxi, Jin, Renzhe
Journal Article
SAE TOMORROW TODAY: Advancing the Path to Full Autonomy1351104/21/2025
From advancements in ADAS (Advanced Driver Assistance Systems) to the importance of AI and redundancy, the path to autonomy is paved with delivering valuable intelligence that optimizes safety. Mobileye is a leader in autonomous driving and driver-assistance technologies, harnessing world-renowned expertise in computer vision, artificial intelligence, mapping, and data analysis. The company's diverse product portfolio ranges from driver assist systems to fully autonomous vehicles, all designed to enhance safety, comfort, and efficiency on the road. To learn more, we sat down with Nimrod Nehushtan, Executive Vice President of Business Development Strategy, to discuss the innovative AI technologies powering Mobileye's products, the importance of redundancy in ensuring reliability, and the company's strategic partnerships with leading OEMs. We'd love to hear from you. Share your comments, questions and ideas for future topics and guests to podcast@sae.org. Don't forget to take a moment to
Hineman, Marcie
Podcast
Statistical Analysis of Data Acquired from Propagating Flames in Gasoline Engines Using a Multiple Ion Probe2024-32-003504/18/2025
Multiple-ion-probe method consists of multiple ion probes placed on the combustion chamber wall, where each individual ion probe detects flame contact and records the time of contact. From the recorded data, it is also possible to indirectly visualize the inside of the combustion chamber, for example, as a motion animation of moving flame front. In this study, a thirty-two ion probes were used to record flames propagating in a two-stroke gasoline engine. The experiment recorded the combustion state in the engine for about 3 seconds under full load at about 6500 rpm, and about 300 cycles were recorded in one experiment. Twelve experiments were conducted under the same experimental conditions, and a total of 4,164 cycles of signal data were obtained in the twelve experiments. Two types of analysis were performed on this data: statistical analysis and machine learning analysis using a linear regression model. Statistical analysis calculated the average flame detection time and standard
Yatsufusa, Tomoaki, Okahira, Takehiro, Nagashige, Kohei
Technical Paper
Study on Motorcycle Rider Model Using Reinforcement Learning - Learning Examples Including Following Target Velocity and Basic Research on Rider Proficiency2024-32-002804/18/2025
In this study, an initial approach using deep reinforcement learning to replicate the complex behaviors of motorcycle riders was presented. Three learning examples were demonstrated: following a target velocity, maintaining stability at low speeds, and following a target trajectory. These examples serve as a starting point for further research. Additionally, the proficiency of the constructed models was examined using rider proficiency evaluation methods developed in previous studies. Initial results indicated that the models have the potential to mimic real rider behaviors; however, challenges such as differences between the model’s output and what humans can produce were also identified for future work.
Mitsuhashi, Yasuhiro, Momiyama, Yoshitaka, Yabe, Noboru
Technical Paper
Accelerating Battery Thermal Analysis by Integrating CFD Simulation and Machine Learning Techniques2024-32-012104/18/2025
The growing demand for sustainable transportation solutions and renewable energy storage systems has heightened the necessity for precise and effective prediction of battery thermal performance. However, achieving both precision and efficiency poses a challenge, necessitating exploration into diverse methodologies. The conventional use of Computational Fluid Dynamics (CFD) offers a comprehensive insight into thermal dynamics but prioritizes precision over efficiency. To enhance the efficiency of this traditional approach, numerous reduced-order modeling techniques have emerged, and the concept of Machine Learning (ML) presents a distinct avenue for enhancing simulation capabilities, particularly in the context of mobility solutions. This paper presents a novel approach to accelerate battery thermal analysis by integrating CFD and ML. The CFD simulations provide an intricate understanding of the thermal dynamics within batteries, encompassing fluid flow and temperature distributions
Devarajan, Gurudevan, Vaidyanathan, Ganesh, Bhave, Ajinkya, Ji, Lichao, Wang, Jiao, Zhou, Wei, He, Jiguang, Shi, Pengfei
Technical Paper
Development of a Robust Framework for Fault Detection and Quantification in Automotive Drum Brakes Using Vibration Measurements and ANN10-09-03-002404/15/2025
Passenger safety is of utmost importance in the automotive industry. Hence, the health of the components, especially the brake system, should be effectively monitored. On account of the significance of artificial intelligence in recent times, any brake fault resulting during operation can be accurately detected using a combination of advanced measurement techniques and machine learning algorithms. The current study focuses on developing and evaluating a robust framework to quantify and classify the faults of a general automotive drum brake. For this purpose, a new experiment for a drum brake, which can be operated under a controlled environment with known levels of faults, is developed. The experiment is instrumented to measure the fundamental dynamic signals (such as brake torque, the angular velocity of the brake drum, and brake shoe accelerations) during a braking event. The response signals from several experiments with various faults and operating conditions serve as the input
Yella, Akash, Bharinikala, Yuva Venkat Ajay, Sundar, Sriram
Journal Article
Learned Human Eye Gaze for Object Tracking: A Preliminary Study09-13-01-000404/15/2025
Visual object tracking technology is the core foundation of intelligent driving, video surveillance, human–computer interaction, and the like. Inspired by the mechanism of human eye gaze, a new correlation filter (CF) tracking algorithm, named human eye gaze (HEG) tracking algorithm, was proposed in this study. The HEG tracking algorithm expanded the tracking detection idea from the traditional detection-tracking to detection-judging-tracking by adding a judging module to check the initial and retrack the unreliable tracking result. In addition, the detection module was further integrated into the edge contour feature on the basis of the HOG (histogram of oriented gradients) extracting feature and the color histogram to reduce the sensitivity of the algorithm to factors such as deformation and illumination changes. The comparison conducted on the OTB-2015 dataset showed that the overall overlap precision, distance precision, and center location error of the HEG tracking algorithm were
Jiang, Yejie, Jiang, Binhui, Chou, Clifford C.
Journal Article
Predicting Real-time Crash Risks during Hurricane Evacuation Using Connected Vehicle Data09-13-01-000604/12/2025
Hurricane evacuations generate high traffic demand with increased crash risk. To mitigate such risk, transportation agencies can adopt high-resolution vehicle data to predict real-time crash risks. Previous crash risk prediction models mainly used limited infrastructure sensor data without covering many road segments. In this article, we present methods to determine potential crash risks during hurricane evacuation from an emerging alternative data source known as connected vehicle data that contain vehicle speed and acceleration information collected at a high frequency (mean = 14.32, standard deviation = 6.82 s). The dataset was extracted from a database of connected vehicle data for the evacuation period of Hurricane Ida on Interstate-10 in Louisiana. Five machine learning models were trained considering weather features and different traffic characteristics extracted from the connected vehicle data. The results indicate that the Gaussian process boosting and extreme gradient
Syed, Zaheen E Muktadi, Hasan, Samiul
Journal Article
Elongation Prediction of Die-Cast Aluminum Alloy Based on 3D Convolutional Neural Network Model05-18-04-003204/09/2025
This study aims to predict the impact of porosities on the variability of elongation in the casting Al-10Si-0.3Mg alloy using machine learning methods. Based on the dataset provided by finite element method (FEM) modeling, two machine learning algorithms including artificial neural network (ANN) and 3D convolutional neural network (3D CNN) were trained and compared to determine the optimal model. The results showed that the mean squared error (MSE) and determination coefficient (R2) of 3D CNN on the validation set were 0.01258/0.80, while those of ANN model were 0.28951/0.46. After obtaining the optimal prediction model, 3D CNN model was used to predict the elongation of experimental specimens. The elongation values obtained by experiments and FEM simulation were compared with that of 3D CNN model. The results showed that for samples with elongation smaller than 9.5%, both the prediction accuracy and efficiency of 3D CNN model surpassed those of FEM simulation.
Zhang, Jin-sheng, Zheng, Zhen, Zhao, Xing-zhi, Gong, Fu-jian, Huang, Guang-sheng, Xu, Xiao-min, Wang, Zhi-bai, Yang, Yutong
Journal Article
Empirical Analysis on Machine Vision Recognition of Green Bike Lanes for Vulnerable Road Users Safety2025-01-801704/01/2025
Deliberate modifications to infrastructure can significantly enhance machine vision recognition of road sections designed for Vulnerable Road Users, such as green bike lanes. This study evaluates how green bike lanes, compared to unpainted lanes, enhance machine vision recognition and vulnerable road users safety by keeping vehicles at a safe distance and preventing encroachment into designated bike lanes. Conducted at the American Center for Mobility, this study utilizes a vehicle equipped with a front-facing camera to assess green bike lane recognition capabilities across various environmental conditions including dry daytime, dry nighttime, rain, fog, and snow. Data collection involved gathering a comprehensive dataset under diverse conditions and generating masks for lane markings to perform comparative analysis for training Advanced Driver Assistance Systems. Quality measurement and statistical analysis are used to evaluate the effectiveness of machine vision recognition using
Ponnuru, Venkata Naga Rithika, Das, Sushanta, Grant, Joseph, Naber, Jeffrey, Bahramgiri, Mojtaba
Technical Paper
Role of AI in an Autonomous Vehicle Perception Systems: Use of Convolutional Neural Network Approach to Design the Vehicle Perception System2025-01-801204/01/2025
Artificial Intelligence has gained lot of traction and importance in the 21st century with use cases ranging from speech recognition, learning, planning, problem solving to search engines etc. Artificial Intelligence also has played a key role in the development of autonomous vehicles and robots ranging from perception, localization, decision to controls. Within the big AI umbrella there is machine learning which is all about using your computer to "learn" how to deal with problems without “programming". Deep learning is a branch of machine learning based on a set of algorithms that learn to represent the data directly from the input such as an image, text, Sound, etc. Within deep learning there are Convolutional Neural Networks and Recurrent Neural Networks (CNN/RNN). The study here used convolutional neural network approach to perform image/object recognition. Given that the objective of the autonomous or semi-autonomous vehicle is to promote safety and reduce number of accidents, it
Mansour, Iyad, Singh, Sanjay
Technical Paper
Combine HD Map Information to Enhance Road Topology Reasoning by Offline Supervised Training and on-Line Ensemble Learning2025-01-802104/01/2025
Topology reasoning plays a crucial role in understanding complex driving scenarios and facilitating downstream planning, yet the process of perception is inevitably affected by weather, traffic obstacles and worn lane markings on road surface. Combine pre-produced High-definition maps (HDMaps), and other type of map information to the perception network can effectively enhance perception robustness, but this on-line fused information often requires a real-time connection to website servers. We are exploring the possibility to compress the information of offline maps into a network model and integrate it with the existing perception model. We designed a topology prediction module based on graph attention neural network and an information fusion module based on ensemble learning. The module, which was pre-trained on offline high-precision map data, when used online, inputs the structured road element information output by the existing perception module to output the road topology, and
Kuang, Quanyu, Rui, Zhang, Zhang, Song, Yixuan, Gao
Technical Paper
Application of Machine Learning Model on Automotive Subframe Design2025-01-861904/01/2025
The automotive subframe, also referred to as a cradle, is a critical chassis structure that supports the engine/electric motor, transmission system, and suspension components. The design of a subframe requires specialized expertise and a thorough evaluation of performance, vehicle integration, mass, and manufacturability. Suspension attachments on the subframe are integral, linking the subframe to the wheels via suspension links, thus demanding high performance standards. The complexity of subframe design constraints presents considerable challenges in developing optimal concepts within compressed timelines. With the automotive industry shifting towards electric vehicles, development cycles have shortened significantly, necessitating the exploration of innovative methods to accelerate the design process. Consequently, AI-driven design tools have gained traction. This study introduces a novel AI model capable of swiftly redesigning subframe concepts based on user-defined raw concepts
Yang, Jiongzhi, Sarkaria, Bikramjit, Kumaraswamy, Prashanth, Kailkere Srinivas, Praveen
Technical Paper
Assessing the Energy Consumption Impact of Intelligent Driving Technologies on Electric Vehicles: A Comprehensive Review2025-01-859804/01/2025
With the continuous advancement of artificial intelligence technology, the automation level of electric vehicles (EVs) is rapidly increasing. Despite the improvements in travel efficiency, safety, and convenience brought about by automation, cutting-edge intelligent technologies also pose the potential of increased energy consumption, such as the computational power required by advanced algorithms and the energy usage of high-precision equipment, leading to higher overall energy consumption for connected or autonomous electric vehicles (CAEVs). To assess the impact of intelligent technologies on AEVs, this study innovatively provides a comprehensive evaluation of the impact of intelligent technologies on CAEV energy consumption from both positive and negative perspectives. After reviewing 59 relevant studies, the findings highlight energy savings achieved through Vehicle-to-Infrastructure and Vehicle-to-Vehicle cooperation as positive effects, while increased energy consumption from
Liu, Tianyi, Qi, Hao, Ou, Shiqi (Shawn)
Technical Paper
Artificial Intelligence Approach for Fuel Cell Model Parameter Calibration2025-01-863204/01/2025
Electrochemical model of a fuel cell involves several parameters which influence its polarization curve. For a numerical fuel cell model to match experimental polarization curve, it is critical to find the right values of these parameters. It is hard to find the values of all the parameters experimentally, and hence parameter calibration is required. A fully automated workflow for calibration of fuel cell model parameters in a three-dimensional Computational Fluid Dynamic (CFD) simulation is created. The CFD model captures detailed electrochemistry and water phase change. The CFD polarization curve is generated by sequentially running a series of simulations starting from low current densities to high current densities. Experimental polarization curve is used as the validation target. An objective function is defined as the L2 norm of the difference between the experimental and the CFD generated polarization curve measured at various current densities. For calibration, eight fuel cell
Champhekar, Omkar, Janakiraman, Arun, Gondipalle, Sreekanth, Ajotikar, Nikhil, Zehr, Randall
Technical Paper
Parameter Virtual Calibration Method for Vehicle Path Tracking Controller Based on the Deep Reinforcement Learning2025-01-831104/01/2025
Path tracking control, which is one of the most important foundations of autonomous driving, could help the vehicle to precisely and smoothly follow the preset path by actively adjusting the front wheel steering angle. Although there are a number of advanced control methods with simple structure and reliable robustness that could assist vehicles achieving path tracking, these controllers have many parameters to be calibrated, and there is a lack of guidance documents to help non-professional test site engineers quickly master calibration methods. Therefore, this paper proposes a parameter virtual calibration method based on the deep reinforcement learning, which provides an effective solution for parameter calibration of vehicle path tracking controller. Firstly, the vehicle trajectory tracking model is established through the kinematic relationship between the vehicle and the target path, combined with the Taylor series expansion linearization method. Next, a vehicle path tracking
Zhao, Jian, Guo, Chenghao, Zhao, HuiChao, Zhao, Yongqiang, Yu, Zhen, Zhu, Bing, Chen, Zhicheng
Technical Paper
Assessment of Injuries in Rear Impacts2025-01-872704/01/2025
Rear impacts make up a significant portion of crashes in the United States. To date, regulations on rear impacts have focused on fuel system integrity and seat performance, while most research has focused on seat performance in relation to occupants’ injuries, with some analyses of crash severity and seat belt effects. The performance of seats and seat belts may vary depending on the size of the occupant. Understanding how occupant characteristics, as well as crash scenarios, affect injury outcomes can show opportunities for further enhancements in rear impact occupant protection. This paper presents analyses using survey weighted logistic regression models to understand the factors affecting serious injury outcomes (i.e., MAIS 3+) in rear impacts, exploring the potential for improving occupant outcomes. Three separate models are evaluated, focusing on 1) overall injury level, 2) head, neck, and cervical-spine injuries, and 3) thorax, abdomen, thoracic- and lumbar-spine injuries for
Greib, Joshua, Jurkiw, Renee, Kryzaniwskyj, Tanja, Owen, Susan, Van Rooyen, Paul, Whelan, Stacey, Williamson, John
Technical Paper
Fully Automated Global Powertrain Optimization for Electric Truck Using Active Learning Algorithm and Backwards Simulation Approach2025-01-853604/01/2025
Electric trucks, due to their weight and payload, need a different layout than passenger electric vehicles (EVs). They require multiple motors or multi-speed transmissions, unlike passenger EVs that often use one motor or a single-speed transmission. This involves determining motor size, number of motors, gears, and gear ratios, complicated by the powertrain system’s nonlinearity. The paper proposes using a stochastic active learning approach (Bayesian optimization) to configure the motors and transmissions for optimal efficiency and performance. Backwards simulation is applied to determine the energy consumption and performance of the vehicle for a rapid simulation of different powertrain configurations. Bayesian optimization, was used to select the electric drive unit (EDU) design candidates for two driving scenarios, combined with a local optimization (dynamic programming) for torque split. By optimizing the electric motor and transmission gears, it is possible to reduce energy
Chen, Bicheng, Wellmann, Christoph, Xia, Feihong, Savelsberg, Rene, Andert, Jakob, Pischinger, Stefan
Technical Paper
Artificial Intelligence Technique Applied to Semi-Active Vibration Control in Vehicle Suspension System2025-01-827004/01/2025
This study investigates the influence of magnetorheological (MR) dampers in semi-active suspension systems (SASSs) on ride comfort, vehicle stability, and overall performance. Semi-active suspension systems achieve greater flexibility and efficacy by combining MR dampers with the advantages of active and passive suspension systems. The study aims to measure the benefits of MR dampers in improving ride comfort, vehicle stability, and overall system performance. The dynamic system model meets all required performance criteria. This study demonstrates that the proposed artificial intelligence approach, including a fuzzy neural networks proportional-integral-derivative (FNN-PID) controller, significantly enhances key performance criteria when tested under various road profiles. The control performance requirements in engineering systems are evaluated in the frequency and time domains. A quarter-car model with two degrees of freedom (2 DOF) was simulated using MATLAB/Simulink to assess the
M.Faragallah, Mohamed, Metered, Hassan, Abdelghany, M.A., Essam, Mahmoud A.
Technical Paper
Unleash the Power of Generative Design for Lightweighting and Reducing Carbon Footprints in Automotive Engineering2025-01-865704/01/2025
The automotive industry faces ongoing challenges in reducing vehicle mass and carbon emissions while ensuring structural integrity. Traditional design approaches often fail to address these issues comprehensively. This paper explores the application of generative design (GD) to optimize critical automotive components, specifically focusing on reducing mass and in turn carbon emissions. GD builds upon traditional topology optimization by employing iterative method using MELS approach to refine designs providing multiple alternative designs to choose from. MELS (Modified Extensible Lattice Sequence) specifically is used to equally spread-out points (designs) in a space by minimizing clumps and empty spaces. This property of MELS makes lattice sequences an excellent space filling DOE scheme. GD leverages the design of experiments (DOE) to vary key design variables systematically to generate and consider many potential design concepts for a given problem. It also uses artificial
Hosmath, Anjaney, Barai, Jay, Dhangar, Vinaykumar
Technical Paper
Identifying Negative Driver States that Share Commonalities for Interventions2025-01-867604/01/2025
Advancements in sensor technologies have led to increased interest in detecting and diagnosing “driver states”—collections of internal driver factors generally associated with negative driving performance, such as alcohol intoxication, cognitive load, stress, and fatigue. This is accomplished using imperfect behavioral and physiological indicators that are associated with those states. An example is the use of elevated heart rate variability, detected by a steering wheel sensor, as an indicator of frustration. Advances in sensor technologies, coupled with improvements in machine learning, have led to an increase in this research. However, a limitation is that it often excludes naturalistic driving environments, which may have conditions that affect detection. For example, reductions in visual scanning are often associated with cognitive load [1]; however, these reductions can also be related to novice driver inexperience [2] and alcohol intoxication [3]. Through our analysis of the
Seaman, Sean, Zhong, Peihan, Angell, Linda, Domeyer, Joshua, Lenneman, John
Technical Paper
Development of Deep Reinforcement Learning Traction Controllers for Front and Rear Wheel Drive Electrified Vehicles2025-01-880304/01/2025
Traction control plays a key role in improving vehicle safety, especially for driving scenarios involving low levels of tire-road friction. Over the past 30 years, academic and industrial research in traction controllers has mainly favored deterministic approaches. This paper introduces a traction control strategy based on a deep reinforcement learning agent tailored for straight-line acceleration maneuvers from standstill in low-friction conditions. The proposed agent is trained on two different electric vehicles, a front-wheel drive city car (from EU vehicle segment A), and a rear-wheel drive sedan (from EU vehicle segment D). The paper presents a deep reinforcement learning agent formulation suitable for training on different vehicles, assesses the performance of the resulting controllers in comparison with a benchmarking integral sliding mode controller, and evaluates their response to changes in vehicle mass, powertrain parameters and tire-road friction conditions. The assessment
Caponio, Carmine, Mihalkov, Mario, Hankovszki, Zoltan, Fuse, Hiroyuki, Ivanov, Valentin, Sorniotti, Aldo, Gruber, Patrick, Montanaro, Umberto
Technical Paper
Items per page:
1 – 50 of 2026