Previous REU Projects UAH

Multisensor Fire detection algorithms have been a pivotal focus point in research. More importantly, implementing the algorithms into autonomous vehicles such as robots and unmanned aerial vehicles (UAVs) has elevated research. This paper presents a fire detection method based on image processing. The basic idea of the proposed project is to gather a BGR (blue, green, and red) model and use image processing to convert the model to an HSV (hue, saturation, and value) color space, which can extract the fire based on colors. The extracted fire pixels will be verified to see if it’s a real fire by the absolute difference of pixels in each frame. If the pixels, are growing from frame to frame a fire is detected. Once, the fire is detected a contour box is placed around it and the distance of the fire is placed on the frame. This fire detection algorithm is implemented into the ROS (Robotic Operating System) Environment. The results show the fire detection accuracy rate and images of the LIMO detecting the fire. Further research can be done to improve the accuracy of the distance calculation, and a notification system can be implemented to inform people nearby that a fire has been detected.

This report explores the research conducted during the summer of 2024 as part of an NSF-REU internship at the University of Alabama Huntsville Department of Electrical and Computer Engineering in the Nonlinear and Complex Systems Laboratory. Research, simulation, and construction of a physical 1st-order chaotic oscillator established a basis of knowledge in nonlinear dynamics. Simulation and test results of the chaotic oscillator aligned with outcomes of similar studies, demonstrating successful replication of previous work. Resistive synchronization of chaotic oscillators yielded intriguing results, as varying clock inputs to each oscillator still produced synchronized waveforms. Modifying the circuit design also led to the observation of unexpected waveforms, leading to a promising potential for future research. The chaotic potential of buck-boost converters was explored, with simulation results aligning with published studies. Unlike the intentions of the referenced papers, our research aims to study and potentially harness these chaotic attributes rather than avoid them. Although not a comprehensive study of buck-boost converters, this work has inspired future research work that will continue beyond the completion of the REU program.

This paper proposes a novel intelligent human activity recognition (HAR) framework based on a new design of Federated Split Learning (FSL) with Differential Privacy (DP) over edge networks. Our FSL-DP framework leverages both accelerometer and gyroscope data, achieving significant improvements in HAR accuracy. The evaluation includes a detailed comparison between traditional Federated Learning (FL) and our FSL framework, showing that the FSL framework outperforms FL models in both accuracy and loss metrics. Additionally, we examine the privacy-performance trade-off under different data settings in the DP mechanism, highlighting the balance between privacy guarantees and model accuracy. The results also indicate that our FSL framework achieves faster communication times per training round compared to traditional FL, further emphasizing its efficiency and effectiveness. This work provides valuable insight and a novel framework which was tested on a real-life dataset.

Lithium-ion Battery Testbed 

Student: Joshua F Barns, Alabama A&M University