Cheol-Hong Min

This paper presents the development of an autonomous indoor service robot that can navigate and interact with an unfamiliar environment. Equipped with a depth camera, IMU, drive motors, and a gripper arm, it uses object, text, and barcode recognition to identify and interact with specific items. The system is built on a Robotis TurtleBot3 with an OpenMANIPULATOR-X arm, running ROS, and an Intel RealSense D435i camera. The platform demonstrates autonomous recognition and task execution indoors, with potential for expansion to diverse applications. In addition, the robot estimates depth information and size information utilizing the depth camera information. Such information is important not only in object manipulation but also in assessing spatial characteristics for intelligent obstacle avoidance.

This paper presents a solution for capturing physical activities in a hybrid work environment and controlling posture using a single multimodal sensor. The sensor system is attached to the subject's Upper Body (chest). The sensor collects accelerometer data during eight different activities and processes it using MATLAB for data analysis. In MATLAB, the data is fed to a machine learning algorithm for classification applications. Additionally, real-life data were collected from the office environment and the home environment. By doing so, we were able to condense the data into the type of physical activity, capture highly desirable points of interest, and obtain accuracy from the MATLAB classification learner. Using this knowledge, we were able to test how accurately our simulation can predict outcomes and assess its capabilities in a real-world scenario. The proposed system can be used to detect various motions, including good sitting posture, twisting and turning, and walking, all from a single sensor.

In this paper, we describe the development of an autonomous robot capable of interacting with an unfamiliar, unmapped environment. Utilizing multiple integrated sensors and actuators, including a depth camera, IMU, and drive motors. This indoor service robot can learn and interact with its surroundings. The robot can be instructed to interact with specific objects in its surroundings and perform a given task by leveraging an object and text recognition algorithm. The autonomous robot consists of a Robotis TurtleBot3 robot, an OpenMANIPULATOR-X gripper arm integrated with Robot Operating System (ROS), and an Intel RealSense D435i stereo camera for object and text recognition. This work aims to demonstrate the ability of the robot to operate in an indoor environment to recognize numbers, read names, and identify objects, for indoor robot applications. The exhibited platform enables further expansion across various applications and environments, allowing it to complete tasks autonomously.

An MBientLab MetaMotionR device was used to collect accelerometer and gyroscope data when worn on the left and right wrist while a subject curled. This data was analyzed, and features were generated. Feature files were annotated with appropriate classes and used as training data in a decision tree algorithm within MATLAB. The feature generation process was iterated, showing improvements in each iteration. The following conclusions were drawn from the outcomes of various models that were created: acceleration data holds more importance than gyroscope data, the variance feature holds more importance during the sweeping motion than other evaluated features, the mean feature holds more importance during the shooting motion than other evaluated features, sensor location can be predicted by the dataset and does not greatly affect the outcome of predicted classes, the side that the subject sweeps on can be predicted and does not greatly affect the outcome of predicted classes, and the direction the subject curls the stone can be predicted by the dataset.

The proposed multilayer superlattice-based amorphous Se structure by Esaki exhibits fascinating features, including quantum size effects, negative differential resistance, and sequential resonant tunneling. This study utilizes data from the transport characteristics of the structure measured using variations of current from the applied voltage (IV) when placed under dark conditions and when illuminated with a superlattice structure deposited on the backside of n-type Si. Computational values and functional fitting curves are plotted using machine learning algorithms, such as Neural Net Fitting and Regression Learner, to optimize the multinano interfaces. Unlike other methodologies used in analyzing nonsuperlattice Se and As 2 Se 3 structure, the methods employed in this study in modeling the Se and As 2 Se 3 superlattice will not only provide significant advancement by observing vital predictions on the responsibility and sensitivity of the photodetectors but also refurbish nano-electronics device fabrication with efficient tools for researchers in the domains of data analysis for semiconductor material and characterizations, as well as paving the way for future research.

The performance of fifth-generation (5G) coding techniques, such as low-density parity-check (LDPC) codes, is significantly affected by the error floor, which occurs in high signal-to-noise ratio (SNR) regions. Federated transfer learning (FTL), with its ability to learn from distributed data, can effectively improve error correction in low-SNR environments and extract crucial decoding features for high-SNR conditions. This paper introduces a two-stage decoding method that integrates layered LDPC decoding with FTL to mitigate the error floor and enhance system performance. Experimental results demonstrate that this approach significantly reduces error rates and improves decoding efficiency.

This paper explores the integration of Large Language Models (LLMs) and secure Gen-AI technologies within engineering design and manufacturing, with a focus on improving inventory management, component selection, and recommendation workflows. The system is intended for deployment and evaluation in a real-world industrial environment. It utilizes vector embeddings, vector databases, and Approximate Nearest Neighbor (ANN) search algorithms to implement Retrieval-Augmented Generation (RAG), enabling context-aware searches for inventory items and addressing the limitations of traditional text-based methods. Built on an LLM framework enhanced by RAG, the system performs similarity-based retrieval and part recommendations while preserving data privacy through selective obfuscation using the ROT13 algorithm. In collaboration with an industry sponsor, real-world testing demonstrated strong results: 88.4% for Answer Relevance, 92.1% for Faithfulness, 80.2% for Context Recall, and 83.1% for Context Precision. These results demonstrate the system’s ability to deliver accurate and relevant responses while retrieving meaningful context and minimizing irrelevant information. Overall, the approach presents a practical and privacy-aware solution for manufacturing, bridging the gap between traditional inventory tools and modern AI capabilities and enabling more intelligent workflows in design and production processes.

Monitoring environmental conditions in a testing laboratory can be a nuanced activity. Depending on the test conditions required by the IEC or ISO standard to which a particular device is being tested, as well as the specific t ests in question, the need for robust environmental sensing and data storage is mandatory. This paper will review one method that is being used at some unnamed IEC / ISO 17025 certified test lab and propose a possible solution for exploration utilizing Microelectromechanical system (MEMS) based environmental sensors and Low Power Wide Area Networks (LPWAN) for Internet of Things (IoT) devices using the LoRaWAN open standard protocols. Tests confirmed t he s uccess o f o ur design with high LoRaWAN reliability.

This paper presents an automated system for semiconductor chip recognition, classification, and pick-and-place operations using a custom object detection model, optical character recognition (OCR), and robotic kinematics. The Hailo AI processor, integrated with a Raspberry Pi, performs recognition, while classification is handled by EasyOCR. A dynamic detection pipeline accommodates varying chip attributes, including pin count, size, and labeling. The process includes image preprocessing, bounding box extraction, and robotic actuation. Experimental testing yielded a 97.9% classification accuracy, demonstrating the system's adaptability, speed, and precision for real-time sorting in smart manufacturing environments.

The TurtleBot is a customizable robotics platform that can use various sensors to map its location, determine a path to take, and recognize objects. This is achieved by using a relatively low-cost pre-built robot framework with several commercially available components. The Raspberry Pi main controller runs robot movement software and outputs lidar data while the host PC runs computationally intensive exploration and object recognition algorithms. This study developed a search and explore algorithm based on the Breadth First Search algorithm for the TurtleBot to explore an area and search for an object using a camera. A framework was also outlined to explore possible ways to incorporate the Intel RealSense D435i depth camera into the current TurtleBot setup to be able to recognize a selected object in real time and use the depth information to assist a grabbing arm in grabbing its target.