Applicants will apply through NSF’s ETAP system. Participants will be selected based on their transcripts, personal statements, and recommendations.

Dr. Diane Peters, the PI for this project, is a Professor of Mechanical Engineering. She has a long history of mentoring undergraduate students both while working in industry and later when she went into teaching engineering. She is the faculty advisor for many student organizations on campus. She has published papers with undergraduate student authors on many different topics. She has a strong interest in control systems and their applications, including how they can be used to develop autonomous vehicles.

Dr. Rui Zhu, the co-P.I. for this proposal, is an Assistant Professor of Computer Science. His research aims at the security, privacy, and performance in wireless network systems and applied machine learning algorithms. His research interests span a broad range of problems, including PHY-layer security in wireless networks, millimeter-wave communication security, optimization of machine learning models, autonomous mobile robots (AMRs), and vehicular networks. His experience in guiding and facilitating undergraduate research in REU programs in the past greatly contributed to the success and growth of the students.

Dr. Mehrdad Zadeh is a Professor of Electrical and Computer Engineering. Dr. Zadeh is an advisor for multiple student teams in the area of autonomy. He is interested largely in the areas of Cyber Physical systems, including VR/AR and Automated Driving. Dr. Zadeh has advised multiple students at both the graduate and undergraduate levels in research on autonomous driving, and has developed several new courses on relevant areas.

Dr. Javad Baqersad is an Associate Professor of Mechanical Engineering at Kettering University. He has been working with undergraduate students as part of capstone projects as well as research theses. Some of these capstone projects include designing and building autonomous RC cars; others involve design and construction of a drone for structural health monitoring. Yet others include the KU-iTire project and control of highly flexible rotating structures. 

Dr. Chinwe Tait is an Assistant Professor of Electrical Engineering in the Department of Electrical and Computer Engineering (ECE). Dr. Tait spent five years working on cutting-edge technology as a Sensor Engineer in the automotive industry, and brings that industry perspective to her research and mentoring of students. Dr. Tait is the first in her family to obtain a graduate-level degree, and has mentored numerous underrepresented minority (URM) students ranging from high school age to university level.

Dr. Lisa Gandy is an Associate Professor in the Department of Computer Science at Kettering University. Prior to joining Kettering University, Dr. Gandy was an assistant and then associate professor at Central Michigan University and served as chair of the department. Dr. Gandy’s research interests focus on natural language processing and text mining. Dr. Gandy has mentored the Women in Technology student organization at CMU for 10 years and is currently mentoring the student chapter of ACM at Kettering University.

Dr. Giuseppe Turini is a Professor in the Department of Computer Science at Kettering University. Prior to joining Kettering, Dr. Turini worked as a researcher in computer-assisted surgery at the EndoCAS Research Center of the University of Pisa (Italy), and in computational geometry at the Visual Computing Lab of the National Research Council (Pisa, Italy). His research activities include: computer graphics and computer vision, VR-AR, HMI-HCI, serious game development, and interactive physics simulation, applied to computer-assisted surgery and medicine.

Dr. Foroogh Rouhollahi is an Assistant Professor of Chemical Engineering at Kettering University. Since 2018, she has focused on battery materials and systems, with expertise in developing nanomaterials for battery applications, as well as in battery testing and monitoring. She has also contributed to digital twin development for engineering systems and served as co-PI on an NSF MRI proposal to acquire a Scanning Laser Doppler Vibrometer (SLDV). She has been leading efforts to apply SLDV for innovative battery monitoring solutions directly tied to this program.

State of Health and State of Charge Monitoring of Battery Systems Using Laser Doppler Vibrometry (Mentors: Dr. Foroogh Rouhollahi, Dr. Javad Baqersad, Dr. Diane Peters)

Rechargeable batteries with advanced electrodes have been developed to meet the increasing demand for electric vehicles (EVs), autonomous vehicles, and portable electronic devices. Due to their high energy density, lithium-ion batteries (LIBs) have become the dominant technology for energy storage. Continuous monitoring of batteries is crucial to ensure safety, reliability, and optimal performance, particularly for autonomous vehicles where unexpected failures cannot be tolerated. Researchers at Kettering University are developing an integrated system for real-time health monitoring of energy-storage devices. We create a digital twin of a battery by integrating physics-based modeling with sensor-acquired data to predict both the state of charge (SOC) and state of health (SOH). A 3D scanning laser Doppler vibrometer (SLDV) system will be used to collect ultrasonic guided-wave data from battery cells, providing high-resolution insights into internal structural and mechanical changes. These data will enable KU researchers to develop, train, and validate advanced algorithms for next-generation battery monitoring systems that support safe and reliable operation of EVs and autonomous vehicles.

Smart School Bus with Heated Seats (Mentor: Dr. Zadeh)  

This project addresses the need for improved efficiency, comfort, and sustainability in school buses. Current school buses use propane heaters, which are inefficient, requiring approximately 10 kW of power. This project aims to replace this system with a more efficient electric heated seat solution. The central research question is how to create a highly efficient, automated heating system that only uses energy when needed, thereby reducing waste and improving overall energy management. The project also explores the integration of solar power and the ability for the bus to exchange energy with the power grid, adding layers of sustainability and functionality. Students will develop a smart school bus equipped with a selective heated seat system that uses AI-powered image processing to detect passengers. The system will use a camera with object detection software like YOLOv5 or OpenCV to determine if a seat is occupied. A microcontroller, such as a Raspberry Pi, will then process this information and control the heater for that specific seat, turning it on when a passenger is detected and off when the seat is empty. 

Intelligent Tire (KU-iTire) for Improved Ride and Handling of Autonomous Vehicles (Mentors: Dr. Javad Baqersad, Dr. Mehrdad Zadeh)

With the rapid advancement of autonomous vehicles and driver-assistance systems, there is a growing need for intelligent tire technologies that can accurately assess vehicle and road conditions in real time. Current vehicle control algorithms estimate tire–road interaction parameters indirectly, making them susceptible to sudden or unmodeled changes in road surface conditions. This project proposes KU-iTire, an intelligent tire model that leverages physics-informed machine learning to predict braking and traction requirements using strain and acceleration data collected from sensors embedded within the tire. As part of this effort, we will integrate open-source computer-vision techniques to capture the vibration and deformation behavior of a rotating tire. High-speed cameras will record tire deformation under various operating conditions, and the resulting video data will be processed using optical-flow and related Python-based algorithms to extract full-field deformation information. These measurements will be incorporated into the development and validation of the KU-iTire model. Students with strong programming skills are encouraged to apply, particularly those with backgrounds in vibration analysis, signal processing, or computer vision.

Investigation of Metal-oxide Nanosensors for Early Detection of Electric Vehicle Battery Failure (Mentor: Dr. C. Tait)

As vehicle electrification is becoming more prevalent, safety in regards to Li-ion batteries is becoming more of a concern. When a defect develops in the battery cell (or outside of the battery itself), flammable gas generation, extensive heating, and, eventually, a fire and explosion may arise. Recent research points to gas sensors as a possible early detector of battery failure by sensing gas emissions developed during initial leakage. In this project, the student will explore pre-fabricated nanosensors through a structured four-step process. They will use advanced equipment like the Environmental Scanning Electron Microscope (ESEM) and X-ray Photoelectron Spectroscopy (XPS) to confirm sensor composition, measure sensor responses to gas emissions, and convert this data into gas concentration metrics. The findings will help improve early detection methods for battery failure, enhancing the safety of electric vehicle batteries.

In Year 2 of this long-term project, the REU student will help to implement nanosensor data with a simulated EV BMS (battery management system) for early detection of battery failure. Under the guidance of the Faculty Mentor, the student will 1) integrate sensor response data with the simulated BMS via MATLAB (and/or other software), 2) gather performance data in a control case (BMS only) and test case (BMS with sensor data) for comparison, 3) draw conclusions, lessons learned, and future work based on the results, 4) further optimize the integration of sensor data with the simulated BMS, and 5) potentially co-author a conference paper on the findings (alongside the Faculty Mentor).

Control-Based Detection of Driver Incapacitation (Mentor: Dr. Diane Peters)

A variety of different medical conditions and issues can result in a driver becoming incapacitated, including uncontrolled seizures, diabetes, heart attacks, and strokes. There are a number of approaches to detecting these emergencies in order to take action to ensure safety and respond appropriately. While many of these approaches are focused on monitoring the driver appearance, another approach would be to monitor driving behavior, and compare the driver’s actions to those of a hypothetical autonomous vehicle control system to determine when the driver may be behaving “irrationally”. In this project, a student will use both simulation and physical experiments to determine whether a system based on comparing a virtual driver to a real driver can effectively determine when a driver might be behaving in a way that indicates a problem is present. 

Display Design for Drivers with Limited Peripheral Vision (Mentor: Dr. Diane Peters)

There are a variety of cases where drivers could have limited peripheral vision, including those who have had certain types of surgery to stop epileptic seizures. This currently results in those drivers losing driving privileges, due to the inability to see conditions and hazards outside of a narrow range immediately in front of them. Systems with full or partial autonomy could restore the ability of someone with this situation to drive safely. As a step towards this, this project will focus on the design of a display system that would provide a driver with visual information about events occurring outside of their range of vision. Once the display is designed, it will be tested to determine how well drivers can perceive information from it and respond appropriately.

In conducting this project, the REU student will be exposed to concepts from cognitive psychology as well as engineering design. Depending on the outcome of the work, the student may have the opportunity to file an invention disclosure as well as produce scholarly publication(s).

Building Trust in Autonomous Vehicles Using Object Explanations with Multimodal Cues (Voice, Captions, and Symbols) (Mentors: Dr. Lisa Gandy, Dr. Giuseppe Turini)

A major hurdle to widespread adoption of autonomous vehicles (AVs) is ensuring that users trust the system’s decisions and actions. Prior work shows that trust improves when the system provides meaningful explanations of what it perceives and why it responds in particular way. Explanations can be delivered through multiple modalities, and accessible, human-interpretable presentations have been shown to strengthen user trust. With modern vision-language models, AVs can now generate real-time descriptions of their surroundings in ways that are natural and intuitive for users. In this project, students will design and implement a Unity based user interface that presents real-time, multimodal explanations of what the vehicle is perceiving.   The system will communicate object explanations through a combination of synthesized voice, on-screen text captions, and symbolic visual indicators.

The explanation component will be powered by ChatGPT-4.0 with vision, which can interpret visual input frames and generate concise, context-aware descriptions of the environment. ChatGPT-4.0’s multimodal capabilities allow it to analyze images, identify relevant objects or interactions, and produce natural-language summaries suitable for both text and speech outputs.

If time allows, students will conduct a small user study using the simulator. Participants will interact with the multimodal explanation system during simulated driving scenarios. Measures of user trust and perceived transparency will be collected after each session to determine the effectiveness of multimodal explanations in fostering trust.

Secure and Intelligent V2X Communication Systems (Mentor: Dr. Rui Zhu)

Vehicle-to-Everything (V2X) communication plays a critical role in modern intelligent transportation systems by enabling vehicles to exchange real-time information that improves safety, efficiency, and situational awareness. However, the openness of wireless channels and the safety-critical nature of V2X data introduce substantial security risks, including falsified messages, spoofed trajectories, adversarial attacks on machine learning models, and limitations of resource-constrained edge devices. To address these challenges, this REU research theme explores cutting-edge approaches that combine artificial intelligence, cyber-physical simulation, embedded systems, and wireless security to build more robust and trustworthy V2X communication infrastructures. Students will work across three interconnected projects that span digital twin development, AI security, and edge deployment, gaining hands-on experience while contributing meaningful advancements to the security of next-generation autonomous and connected vehicle systems. Under this broad umbrella, Dr. Zhu has three available projects that students could work on.

Project 1: Digital Twin for V2X Misbehavior Simulation and Detection

This project develops a V2X Digital Twin, a simulation-driven virtual environment that models road geometry, vehicle movement, wireless communication, and malicious behavior injections. Students will build a SUMO- or OMNeT++-based digital twin to simulate realistic Basic Safety Message (BSM) exchanges among vehicles. Multiple misbehavior scenarios, such as position spoofing, replay attacks, speed tampering, and sudden trajectory manipulation, will be injected into the simulation. The resulting data will be used to analyze traffic flow disruption, collision probability, and the effectiveness of machine-learning-based misbehavior detectors.

Deliverables include a reusable simulation framework, a labeled dataset of misbehavior events, visual analytics dashboards, and comparative evaluation of detection algorithms. This project is highly hands-on and enables students to explore cybersecurity, wireless communication, and traffic simulation in an integrated environment.

Project 2: Adversarial Machine Learning Attacks Against V2X Misbehavior Detectors

Machine-learning models increasingly serve as the core technology behind V2X security, enabling detection of falsified position, speed, or trajectory data. However, these models are vulnerable to adversarial attacks, carefully crafted inputs that deceive the detector while appearing benign. This project studies the adversarial robustness of V2X misbehavior detectors and develops methods to construct and defend against adversarial Basic Safety Messages. Students will design both white-box and black-box adversarial attacks (e.g., FGSM, PGD, SPSA) to minimally perturb features such as position, heading, acceleration, and timestamp to evade detection. The project investigates the physical plausibility of such attacks and evaluates their impact on model confidence, detection accuracy, and safety-critical decision making. Students will also explore lightweight defense mechanisms such as adversarial training, feature clipping, smoothing, and confidence-based filtering.

This project provides experience in AI security, adversarial ML, and transportation cybersecurity. Students will produce adversarial datasets, robustness analysis, and visualization of adversarial trajectories.

Project 3: Real-Time Edge Deployment of V2X Misbehavior Detection

In V2X systems, real-time misbehavior detection must be performed close to the road, on roadside units (RSUs) and edge devices with limited computational power. This project focuses on deploying lightweight machine learning models on edge devices (e.g., Raspberry Pi 5) to perform real-time security analysis of received Basic Safety Messages. Students will port trained misbehavior detection models (MLP, LSTM, quantized models, or CfC variants) to Raspberry Pi via ONNX Runtime or PyTorch Lite. They will measure inference latency, CPU/GPU load, memory consumption, and energy usage under various conditions. Students will explore model optimization techniques such as quantization, pruning, and distillation to meet strict real-time latency constraints.

The project concludes with building a real-time V2X security pipeline on the Pi, capable of receiving live BSM streams over Wi-Fi or LoRa and issuing warnings when malicious behavior is detected. Students gain hands-on experience in edge AI, IoT security, and embedded machine learning.