Robotics Commons^™

Contextual Augmentation In Artificial Intelligence, Emmanuel Joshua Balogun

College of Graduate Studies: Theses & Dissertations

Contextual understanding is a significant challenge of Large Language Models (LLMs), which are typically trained on general-purpose datasets. Due to this, LLMs fail to capture nuanced or domain-specific information and may struggle to interpret user queries accurately. Consequently, prompt engineering can become complex in automating, and LLMs are prone to “hallucinating”—generating random or irrelevant texts—when they lack sufficient context. This undermines their ability to provide focused, accurate responses. Accordingly, this thesis seeks to enhance the contextual understanding capabilities of Artificial Intelligence systems to facilitate more precise and relevant answer generation. Study A looks into a new approach to combating misinformation …

Self-Supervised Contrastive Learning Using Eeg Signals For Mental Stress Assessment, Ugochukwu Uraechu

College of Graduate Studies: Theses & Dissertations

A study is presented to investigate self-supervised contrastive learning (SSCL) models using physiological data obtained from non-invasive wearable sensors for mental stress assessment. The present work involved acquisition of electroencephalography (EEG) signals using wearable sensors, signal preprocessing, data augmentation, and investigation of self-supervised contrastive learning (SSCL) algorithms for multi-class mental stress assessment. Seven volunteers participated in this study executing various mental tasks while wearing an OpenBCI head cap to acquire EEG signals. The acquired EEG signals were preprocessed and utilized for data augmentation in time and frequency domains with different SSCL models. Optimal data augmentation combinations and SSCL models were …

Implementation Of Static Rfid Landmarks In Slam For Planogram Compliance, Brennan L. Drake

Honors College Theses

Autonomous robotic systems are becoming increasingly prevalent in everyday life and exhibit robust solutions in a wide range of applications. They face many obstacles with the foremost of which being SLAM, or Simultaneous Localization and Mapping, that encompasses both creation of the map of an unknown environment and localization of the robot in said environment. In this experiment, researchers propose the use of RFID tags in a semi-dynamic commercial environment to provide concrete landmarks for localization and mapping in pursuit of increased locational certainty. With this obtained, the ultimate goal of the research is to construct a robotics platform for …

A Camera-Only Based Approach To Traffic Parameter Estimation Using Mobile Observer Methods, Temitope D. Jegede

College of Graduate Studies: Theses & Dissertations

As vehicles become more modern, a large majority of vehicles on the road will have the required sensors to smoothly interact with other vehicles and infrastructure on the road. There will be many benefits of this new connectivity between vehicles on the road but one of the most profound improvements will be in the area of road accident prevention. Vehicles will be able to share information vital to road safety to oncoming vehicles and vehicles that are occluded so they do not have a direct line of sight to see a pedestrian or another vehicle on the road.

Another advantage …

Design And Fabrication Of A Force-Displacement Control Mechanism For Bone-Surgical Tool Testing, Kenneth Nwagu

College of Graduate Studies: Theses & Dissertations

This project focuses on the design and fabrication of an experimental setup for orthopedic-tool testing, tailored for a surgical instrumentation company. The multifaceted project encompasses a literature review, conceptual design, prototyping, and rigorous testing, resulting in a versatile control system capable of assessing various orthopedic tools, including bone drills, saws, burrs, and power handpieces.

Orthopedic surgical procedures (which include cutting and/or drilling into bone) often need to be performed on bones for faster recovery. The drilling and cutting process can cause an increase in temperature at the cutting site which can cause bone necrosis. The tools also need to be …

Unobtrusive Assessment Of Student Engagement Levels In Online Classroom Environment Using Emotion Analysis, Sasirekha Anbusegaran

College of Graduate Studies: Theses & Dissertations

Measuring student engagement has emerged as a significant factor in the process of learning and a good indicator of the knowledge retention capacity of the student. As synchronous online classes have become more prevalent in recent years, gauging a student's attention level is more critical in validating the progress of every student in an online classroom environment. This paper details the study on profiling the student attentiveness to different gradients of engagement level using multiple machine learning models. Results from the high accuracy model and the confidence score obtained from the cloud-based computer vision platform - Amazon Rekognition were then …

Quantitative Performance Assessment Of Lidar-Based Vehicle Contour Estimation Algorithms For Integrated Vehicle Safety Applications, David M. Mothershed

College of Graduate Studies: Theses & Dissertations

Many nations and organizations are committing to achieving the goal of `Vision Zero' and eliminate road traffic related deaths around the world. Industry continues to develop integrated safety systems to make vehicles safer, smarter and more capable in safety critical scenarios. Passive safety systems are now focusing on pre-crash deployment of restraint systems to better protect vehicle passengers. Current commonly used bounding box methods for shape estimation of crash partners lack the fidelity required for edge case collision detection and advanced crash modeling. This research presents a novel algorithm for robust and accurate contour estimation of opposing vehicles. The presented …

End-To-End Learning: Using Neural Networks For Vehicle Control And Obstacle Avoidance, Keith Russell

Honors College Theses

End to End learning is a method of deep learning which has been used to great effect to solve complex problems which would normally be performed by humans. Within this thesis, a neural network was created to mimic the steering patterns of humans in highway driving situations. A Turtlebot was used in place of a car and was tested within a laboratory on a closed loop track to drive within the lanes created for it. The network architecture was based on that of Nvidia’s model which was used for predicting steering angles of a vehicle. The network was successfully trained …

Hydrogen Fuel Cell Gasket Handling And Sorting With Machine Vision Integrated Dual Arm Robot, Devin C. Fowler

College of Graduate Studies: Theses & Dissertations

Recently demonstrated robotic assembling technologies for fuel cell stacks used fuel cell components manually pre-arranged in stacks (presenters), all oriented in the same position. Identifying the original orientation of fuel cell components and loading them in stacks for a subsequent automated assembly process is a difficult, repetitive work cycle which if done manually, deceives the advantages offered by automated fabrication technologies of fuel cell components and by robotic assembly processes. We present an innovative robotic technology which enables the integration of automated fabrication processes of fuel cell components with robotic assembly of fuel cell stacks into a fully automated fuel …

Ego-Localization Navigation For Intelligent Vehicles Using 360° Lidar Sensor For Point Cloud Mapping, Tyler Naes

College of Graduate Studies: Theses & Dissertations

With its prospects of reducing vehicular accidents and traffic in highly populated urban areas by taking the human error out of driving, the future in automobiles is leaning towards autonomous navigation using intelligent vehicles. Autonomous navigation via Light Detection And Ranging (LIDAR) provides very accurate localization within a predefined, a priori, point cloud environment that is not possible with Global Positioning System (GPS) and video camera technology. Vehicles may be able to follow paths in the point cloud environment if the baseline paths it must follow are known in that environment by referencing objects detected in the point cloud …

Mechanical Design And Optimization Of An Interactive Animatronic Bald Eagle, Eric Burns

Phi Kappa Phi Research Symposium (Archived)

Animatronics is a specialized sub-category of mechatronics, a fusion of mechanical and electrical engineering. The field has grown from small, individual projects into a major industry. As animatronics progress, mechanical engineers are pushed to design internal structures which occupy ever-decreasing spaces and to ensure designs can undergo maintenance and modifications smoothly. This research investigates methods of reducing space required for mechanisms and several other beneficial methods of development as well as varying satisfactions for audiences when exposed to actor-controlled systems rather than pre-scripted functions. The mechanical systems are designed using CAD software available at Georgia Southern. On-campus, resources are used …

The Design Of A Mobile, Animatronic Bald Eagle, Eric Burns

Phi Kappa Phi Research Symposium (Archived)

Modern animatronics are used as a tool to help engage audiences in realistic and entertaining experiences. Many animatronic figures are dynamic in movement, but static in mobility. The research presented covers the beginnings of mobile animatronic research. An animatronic bald eagle, Opportunity, was designed for that purpose. Animations for Opportunity are generated by integration of Arduino, Pololu Mini Maestro, and Hitec Servo Motors. Arduino receives signals from flex sensors which cause voltage variations when “flexed”. The 1:1 scale, animatronic eagle can be perched on the arm of an “actor” who simulates an eagle’s caretaker. The flex sensors are located within …