SDIP Research Experiences for Undergraduates (REU)
Research Projects
Question and working hypothesis. Can a two-camera AI-vision system automatically monitor fish movement through an experimental open-channel flume while reducing manual video analysis? The working hypothesis is that calibrated, overlapping overhead views and automated detection and tracking can replace the existing three-camera arrangement while maintaining continuous coverage and reliable movement measurements under different flow, lighting, and water-surface conditions.
Student work and evaluation. The student will evaluate camera resolution, mounting height, orientation, field of view, illumination, and glare reduction. After constructing and calibrating the imaging setup, the student will collect representative videos and develop an object-detection and tracking pipeline, using a detector such as YOLO. Spatial registration will connect detections across camera views to reconstruct trajectories and estimate fish location, velocity, acceleration, and passage time. Evaluation against manually analyzed videos will examine detection accuracy, trajectory error, missed or interrupted tracks, movement-measurement error, and processing time. Tests under changing flow, lighting, glare, and surface reflections will assess robustness and identify conditions requiring further improvement.
Mentoring, integration, and deliverables. Dr. Abul Baki will serve as faculty mentor, guiding flume experiments, camera configuration, fish-monitoring requirements, and interpretation of movement. A graduate student may support sensing-system design, hardware integration, and AI implementation. Deliverables include a calibrated two-camera prototype and mounting arrangement, labeled videos, fish-detection and tracking software, an automated movement-analysis pipeline, comparative validation results, a system performance report, and a working laboratory demonstration. The student will also complete the REU research report and poster.
Question and working hypothesis. Can synchronized virtual reality (VR), inertial measurement units (IMUs), and electroencephalography (EEG) improve walking-balance assessment in young adults? The working hypothesis is that combining head and trunk kinematics with cortical activity will provide complementary information about sensory integration and postural control across eight Gait Sensory Interaction Test (GaitSIT) conditions involving four visual environments and firm or foam surfaces.
Student work and evaluation. The student will evaluate sensor placement, sampling frequency, headset compatibility, participant comfort, and motion-related EEG artifacts. A lumbar IMU and wearable EEG system will be integrated with GaitSIT using common timestamps and event markers. Processing will quantify mediolateral sway, velocity, acceleration, movement variability, and head-trunk coordination, alongside frontal and parietal EEG measures, including theta and alpha activity. Laboratory calibration and approved pilot testing will assess synchronization accuracy, signal quality, missing-data rates, repeatability, and sensitivity to sensory conditions. Analyses will consider walking speed where appropriate and examine whether multimodal recordings provide information beyond headset measurements alone.
Mentoring, integration, and deliverables. Dr. Kwadwo Appiah-Kubi will serve as faculty mentor, guiding sensory integration, VR-based balance assessment, EEG interpretation, and experimental evaluation. An engineering graduate student may assist with wearable-sensor integration, synchronization, signal processing, and data-pipeline development. Deliverables include an integrated VR-IMU-EEG prototype, synchronized acquisition software, sensor-placement and calibration procedures, an EEG preprocessing and artifact-reduction workflow, kinematic and cortical analyses, a pilot evaluation, and a laboratory demonstration. The student will complete the REU research report and poster.
Question and working hypothesis. Can a virtual reality (VR) rehabilitation system accurately measure upper-limb movement and hand-grip performance? The working hypothesis is that calibration and synchronization with reference instruments will establish reliable measurements of speed, range of motion, smoothness, trajectory deviation, and grip performance. Comparisons will identify measurement limitations and the system's ability to detect compensatory trunk movements.
Student work and evaluation. The student will examine the headset, controllers, and haptic glove and establish protocols for biceps curls and hand-squeezing exercises. Reference measurements will use optical motion capture or wearable inertial sensors, together with a force-sensing ball or hand dynamometer. The student will integrate acquisition systems, align timestamps, calibrate sensors, and develop algorithms for elbow motion, squeeze level, individual finger contribution, trunk compensation, and movement smoothness, including spectral arc length. Approved pilot trials will compare simultaneous recordings using measurement bias, limits of agreement, reference-measurement error, repeatability, and calibration drift. Results will identify measurements requiring further calibration or algorithm improvement before rehabilitation assessment.
Mentoring, integration, and deliverables. Dr. Shafique Chaudhry will serve as faculty mentor, guiding the VR platform, exercise requirements, haptic-glove integration, movement analysis, and interpretation of validation results. A graduate student may assist with sensor integration, synchronization, software development, and experimental testing. Deliverables include a synchronized measurement setup, calibration and testing procedures, paired VR and reference recordings, movement-analysis and trunk-compensation algorithms, quantitative validation results, and a laboratory demonstration. The student will complete a system validation report and the REU research report and poster.
Question and working hypothesis. Can a low-cost, noncontact imaging system capture high-quality iris images from infants and children with limited cooperation? The working hypothesis is that near-infrared illumination, optimized optics, automated acquisition, and real-time quality assessment will improve capture success while reducing acquisition time and repeated attempts. The study will address motion, gaze variation, eyelid occlusion, and illumination-related reflections.
Student work and evaluation. The student will evaluate cameras, lenses, working distances, mounting arrangements, illumination safety, and age-appropriate attention cues. An imaging prototype will integrate acquisition software with iris localization, focus assessment, motion-blur detection, and best-frame selection. Laboratory tests using suitable targets and adult volunteers will precede pediatric pilot testing, which will require research approvals and parental consent. Evaluation will measure capture success, acquisition time, attempts, sharpness, iris visibility, segmentation success, and image quality. Tests will vary working distance, head movement, gaze, and illumination. Where possible, comparisons with an existing device will examine whether quality improvements support segmentation and biometric matching.
Mentoring, integration, and deliverables. Prof. Stephanie Schuckers will serve as primary faculty mentor, guiding iris biometrics, image quality, and system evaluation. Dr. Masudul H. Imtiaz will serve as co-advisor for camera and illumination design, embedded integration, and image processing. One graduate student will assist with hardware, software, experimental setup, and testing. Deliverables include an imaging prototype, quality-assessment and frame-selection software, calibration and acquisition procedures, an approved image dataset, quantitative performance results, and a laboratory demonstration. The student will complete the REU research report and poster.
Question and working hypothesis. Can a flexible electrochemical smart bandage maintain reliable sensing and Bluetooth communication while reducing energy consumption? The working hypothesis is that a rechargeable power source and optimized energy management can extend operating lifetime without compromising measurements. The project will investigate whether supplementary energy harvesting provides benefits while maintaining flexibility and limiting electrical interference.
Student work and evaluation. The student will examine the existing EFR32xG-based flexible potentiostat and evaluate thin batteries, power-management circuits, flexible packaging, and photovoltaic, thermal, or motion-based harvesting options. Development will integrate a rechargeable battery, regulation and protection circuitry, and an interchangeable three-electrode interface. At least one harvesting approach will be evaluated. Firmware will optimize low-power modes, sampling, and scheduled Bluetooth transmission. Laboratory tests using controlled solutions or artificial sweat will compare chronoamperometry and cyclic voltammetry with a reference potentiostat. Evaluation will assess accuracy, repeatability, noise, drift, average and peak power, battery lifetime, temperature rise, and reliability during bending. Battery-only and harvesting-assisted configurations will be compared.
Mentoring, integration, and deliverables. Prof. Silvana Andreescu and Dr. Masudul H. Imtiaz will serve as faculty co-mentors, guiding electrochemical sensing and embedded-system integration, respectively. One graduate student will assist with hardware, firmware, measurements, and testing. Deliverables include a smart-bandage prototype, rechargeable power subsystem, harvesting feasibility evaluation, low-power firmware, Bluetooth acquisition interface, electrochemical validation results, energy-management comparison, and a laboratory demonstration. The student will complete a system performance report and the REU research report and poster. Testing will remain laboratory-based rather than clinical or diagnostic.
Question and working hypothesis. Can a single printed circuit board (PCB) measure individual airborne particle charge without collecting or contacting particles? A passing charged particle induces an electrode signal that depends on its height above the board. The proposed working hypothesis is that narrow-to-wide electrode pulse ratios can estimate particle height, enabling correction of height-dependent charge measurements.
Student work and evaluation. The student will simulate electrode widths, spacings, channel heights, and guard-trace layouts using COMSOL or a finite-element model. Results will establish a height lookup relation and guide electrode and low-noise amplifier selection. With laboratory personnel, the student will fabricate the PCB, build a channel with a grounded cover, and establish microcontroller- or data-acquisition-based recording. Calibration will use particles generated with a differential mobility analyzer; a fine nozzle may introduce particles at known heights. Python or MATLAB processing will detect paired pulses and estimate height, corrected charge, and velocity. Evaluation will examine noise floor, minimum detectable charge, agreement with a Faraday-cup reference, velocity accuracy, coincidence errors, and repeatability across boards.
Mentoring, integration, and deliverables. Dr. Suresh Dhaniyala will serve as faculty mentor, guiding aerosol generation, charging experiments, sensor physics, and interpretation within Clarkson's Center for Air and Aquatic Engineering and Science. A graduate student will support PCB layout, electronics testing, and data analysis as appropriate. Deliverables proposed here include simulation results, a sensor and channel prototype, a height-correction lookup relation, calibrated acquisition and analysis software, experimental datasets, and a performance report documenting sensitivity, accuracy, repeatability, and limitations.
Question and working hypothesis. Can strain measurements and a surrogate trained on finite-element simulations predict the tip deflection of a composite cantilever beam? The working hypothesis is that combining ANSYS simulations with experimental calibration will improve prediction accuracy within a defined elastic loading range. The digital twin will update deflection estimates as new strain measurements become available.
Student work and evaluation. The student will model beam geometry, laminate layup, material properties, and clamping conditions in ANSYS. Static or slowly varying loading simulations will generate surface strains and tip deflections, guide strain-gauge placement, and provide surrogate training data. The student will prepare the beam setup, install gauges, establish signal conditioning and acquisition, and develop a response-surface or Gaussian-process model in MATLAB or Python. Independent displacement measurements will provide validation references. Selected experiments will calibrate model parameters or correct predictions, while separate loading cases will remain reserved for validation. Evaluation will assess simulation-to-experiment strain agreement, root-mean-square and maximum deflection errors, calibration improvement, repeated-loading consistency, sensitivity to measurement noise, and prediction time.
Mentoring, integration, and deliverables. Dr. Mishal Thapa will serve as faculty mentor, guiding structural modeling, experimental planning, surrogate development, and interpretation through Clarkson's PREDICT Laboratory. Graduate student involvement is not specified. Deliverables include an instrumented cantilever demonstration, an ANSYS finite-element model, a calibrated surrogate, experimental and validation datasets, and a final report and presentation. The demonstration will connect incoming strain measurements to updated tip-deflection estimates, documenting prediction accuracy and limitations within the tested loading range.
Question and working hypothesis. Can a sensor-equipped robotic goose safely discourage geese from occupying public parks? The working hypothesis is that positioning, inertial, and proximity sensing will improve navigation reliability and support controlled, noncontact deterrence. The project will prioritize supervised operation, obstacle avoidance, and performance in uneven terrain, vegetation, and changing outdoor conditions.
Student work and evaluation. The student will integrate a Global Navigation Satellite System receiver, inertial measurement unit, and selected obstacle sensors, incorporating wheel encoders where available. Software will combine sensor measurements for localization and heading estimation, waypoint following, boundary monitoring, logging, battery monitoring, and remote stopping. Controlled laboratory and outdoor tests will measure position error, waypoint-tracking error, obstacle-detection reliability, route completion, operator interventions, communication reliability, power consumption, and operating lifetime. Subject to park approval and suitable conditions, supervised field trials may compare movement patterns using goose displacement, time to leave, and return behavior. Testing will avoid physical contact, nests, and flightless young.
Mentoring, integration, and deliverables. Dr. Ajay Sonar will serve as faculty mentor, guiding robotics, control, sensor integration, navigation, and experimental evaluation. Graduate student assistance, if available, will support hardware integration, programming, calibration, and testing. Deliverables include an instrumented robotic goose, localization and navigation software, obstacle detection, wireless monitoring and remote control, emergency-stop functionality, engineering performance results, and a working demonstration. Preliminary deterrence findings will depend on field authorization and conditions. The student will complete a system performance report and the REU research report and poster.
Question and working hypothesis. Can environmental and camera-specific metadata improve early wildfire smoke detection across camera networks while reducing false alarms? The working hypothesis is that combining visual features with weather, visibility, time of day, camera orientation, and terrain information will outperform an image-only model, particularly for distant smoke and confusing fog, clouds, dust, steam, and glare.
Student work and evaluation. The student will curate authorized images, video sequences, and available metadata and establish training, validation, and test partitions separated by camera and fire incident. After reproducing a compact image-only baseline, the student will investigate contextual fusion, feature conditioning, or confidence calibration. Tests on unseen cameras and events will compare smoke-event recall, precision, false alerts per camera-day, precision-recall performance, time-to-detection, and calibration error. Feature-removal experiments will identify useful contextual inputs, while subgroup analyses and bootstrap intervals will examine robustness and uncertainty. Evaluation will remain retrospective rather than operational wildfire warning.
Mentoring, integration, and deliverables. Prof. Stephen T. Casper, faculty mentor, and Peter C. Mehlitz, ODIN technical mentor, will jointly supervise the project. Casper will guide regional context, responsible AI, and practitioner needs; Mehlitz will guide data interfaces, software architecture, and open-source integration. Graduate student involvement is not specified. Deliverables include documented datasets and partitions, baseline and context-aware models, false-alarm and calibration analyses, an ODIN-compatible inference record or interface, reproducible code, and a demonstration. The student will complete a model performance report describing limitations, plus the REU research report and poster.
Question and working hypothesis. Can polymer-modified electrochemical sensors provide useful measurements of per- and polyfluoroalkyl substances (PFAS) and other emerging contaminants in water? The proposed working hypothesis is that polymer recognition properties, combined with contaminant-specific calibration, can support reliable environmental measurements. The project connects molecular recognition, sensor fabrication, and environmental testing through portable sensing platforms developed in Dr. Silvana Andreescu's research group.
Student work and evaluation. The student will design and fabricate sensors on printed circuit boards and modify electrochemical sensing elements for selected contaminants. Work will include preparing sensing platforms, calibrating responses for specific contaminant detection, and analyzing measurement data. Quality-control and quality-assurance procedures will support validation before applying sensors to water samples from nearby Superfund sites and other sources. Proposed evaluation will examine calibration consistency, repeatability of sensor responses, and the usefulness of measurements from environmental samples. The student will document fabrication steps, sensor modifications, calibration results, and quality checks so that performance can be interpreted alongside experimental conditions. The analysis will distinguish demonstrated detection capability from limitations requiring further development.
Mentoring, integration, and deliverables. Dr. Silvana Andreescu will serve as faculty mentor, guiding polymer recognition, electrochemical sensor modification, calibration, and environmental applications. Graduate student involvement is not specified. Proposed deliverables include modified PCB sensor prototypes, documented fabrication and calibration procedures, calibration datasets, quality-control and quality-assurance records, and analyses of environmental water samples. A final report and poster will summarize the sensing approach, experimental findings, and remaining limitations.
Question and working hypothesis. Can a compact ultrasonic sensing system detect and quantify changes in the internal structure and mechanical properties of granular composite materials without damaging the specimens? The working hypothesis is that ultrasonic wave
propagation characteristics, including wave velocity, attenuation, and frequency-dependent response, are sensitive to variations in material composition, particle bonding, porosity, and structural integrity. A properly designed ultrasonic sensor, combined with physics-based
modeling and signal analysis, can provide quantitative indicators of these material properties.
Student work and evaluation. During the design stage, the student will investigate ultrasonic transducer configurations, operating frequencies, sensor placement, coupling methods, and data acquisition requirements. The student will evaluate how sensor design and
measurement conditions influence sensitivity, repeatability, and signal quality. During the development stage, the student will assemble and calibrate an ultrasonic measurement system, manufacture or prepare representative granular composite specimens, and develop signal-processing methods to extract time-of-flight, wave velocity, attenuation, and spectral features. Computational modeling may be used to investigate wave propagation and interpret experimental observations. During implementation and testing, the student will conduct controlled experiments on specimens with different compositions or manufacturing conditions. Sensor performance will be evaluated using measurement accuracy, repeatability, sensitivity to material variations, signal-to-noise ratio, and agreement with independently measured material properties.
Mentoring, integration, and deliverables. Dr. Cetin Cetinkaya (Photo-acoustics Research Lab) will serve as faculty mentor and guide ultrasonic sensor design, experimental methodology, wave propagation analysis, and interpretation of material properties. The student will support sensor development, experimental measurements, signal processing, and integration with ongoing granular composite research, as appropriate.
Deliverables will include an ultrasonic sensor design, a calibrated experimental measurement system, a representative material dataset, signal-processing and analysis software, an assessment of sensor performance, and a working laboratory demonstration, along with the
common REU research report and poster.
