The capture of a target spacecraft by a chaser is an on-orbit docking operation that requires an accurate, reliable, and robust object recognition algorithm. Vision-based guided spacecraft relative motion during close-proximity maneuvers has been consecutively applied using dynamic modeling as a spacecraft on-orbit service system. This research constructs a vision-based pose estimation model that performs image processing via a deep convolutional neural network. The pose estimation model was constructed by repurposing a modified pretrained GoogLeNet model with the available Unreal Engine 4 rendered dataset of the Soyuz spacecraft. In the implementation, the convolutional neural network learns from the data samples to create correlations between the images and the spacecraft’s six degrees-of-freedom parameters. The experiment has compared an exponential-based loss function and a weighted Euclidean-based loss function. Using the weighted Euclidean-based loss function, the implemented pose estimation model achieved moderately high performance with a position accuracy of 92.53 percent and an error of 1.2 m. The in-attitude prediction accuracy can reach 87.93 percent, and the errors in the three Euler angles do not exceed 7.6 degrees. This research can contribute to spacecraft detection and tracking problems. Although the finished vision-based model is specific to the environment of synthetic dataset, the model could be trained further to address actual docking operations in the future.
In one, docking is defined as “when one incoming spacecraft rendezvous with another spacecraft and flies a controlled collision trajectory in such a manner to align and mesh the interface mechanisms”, and defined docking as an on-orbital service to connect two free-flying man-made space objects. The service should be supported by an accurate, reliable, and robust positioning and orientation (pose) estimation system. Therefore, pose estimation is an essential process in an on-orbit spacecraft docking operation. The position estimation can be obtained by the most well-known cooperative measurement, a Global Positioning System (GPS), while the spacecraft attitude can be measured by an installed Inertial Measurement Unit (IMU). However, these methods are not applicable to non-cooperative targets. Many studies and missions have been performed by focusing on mutually cooperative satellites. However, the demand for non-cooperative satellites may increase in the future. Therefore, determining the attitude of non-cooperative spacecrafts is a challenging technological research problem that can improve spacecraft docking operations. One traditional method, which is based on spacecraft control principles, is to estimate the position and attitude of a spacecraft using the equations of motion, which are a function of time. However, the prediction using a spacecraft equation of motion needs support from the sensor fusion to achieve the highest accuracy of the state estimation algorithm. For non-cooperative spacecraft, a vision-based pose estimator is currently developing for space application with a faster and more powerful computational resource.
คณะวิศวกรรมศาสตร์
The integration of intelligent robotic systems into human-centric environments, such as laboratories, hospitals, and educational institutions, has become increasingly important due to the growing demand for accessible and context-aware assistants. However, current solutions often lack scalability—for instance, relying on specialized personnel to repeatedly answer the same questions as administrators for specific departments—and adaptability to dynamic environments that require real-time situational responses. This study introduces a novel framework for an interactive robotic assistant (Beckerle et al. , 2017) designed to assist during laboratory tours and mitigate the challenges posed by limited human resources in providing comprehensive information to visitors. The proposed system operates through multiple modes, including standby mode and recognition mode, to ensure seamless interaction and adaptability in various contexts. In standby mode, the robot signals readiness with a smiling face animation while patrolling predefined paths or conserving energy when stationary. Advanced obstacle detection ensures safe navigation in dynamic environments. Recognition mode activates through gestures or wake words, using advanced computer vision and real-time speech recognition to identify users. Facial recognition further classifies individuals as known or unknown, providing personalized greetings or context-specific guidance to enhance user engagement. The proposed robot and its 3D design are shown in Figure 1. In interactive mode, the system integrates advanced technologies, including advanced speech recognition (ASR Whisper), natural language processing (NLP), and a large language model Ollama 3.2 (LLM Predictor, 2025), to provide a user-friendly, context-aware, and adaptable experience. Motivated by the need to engage students and promote interest in the RAI department, which receives over 1,000 visitors annually, it addresses accessibility gaps where human staff may be unavailable. With wake word detection, face and gesture recognition, and LiDAR-based obstacle detection, the robot ensures seamless communication in English, alongside safe and efficient navigation. The Retrieval-Augmented Generation (RAG) human interaction system communicates with the mobile robot, built on ROS1 Noetic, using the MQTT protocol over Ethernet. It publishes navigation goals to the move_base module in ROS, which autonomously handles navigation and obstacle avoidance. A diagram is explained in Figure 2. The framework includes a robust back-end architecture utilizing a combination of MongoDB for information storage and retrieval and a RAG mechanism (Thüs et al., 2024) to process program curriculum information in the form of PDFs. This ensures that the robot provides accurate and contextually relevant answers to user queries. Furthermore, the inclusion of smiling face animations and text-to-speech (TTS BotNoi) enhanced user engagement metrics were derived through a combination of observational studies and surveys, which highlighted significant improvements in user satisfaction and accessibility. This paper also discusses capability to operate in dynamic environments and human-centric spaces. For example, handling interruptions while navigating during a mission. The modular design allows for easy integration of additional features, such as gesture recognition and hardware upgrades, ensuring long-term scalability. However, limitations such as the need for high initial setup costs and dependency on specific hardware configurations are acknowledged. Future work will focus on enhancing the system’s adaptability to diverse languages, expanding its use cases, and exploring collaborative interactions between multiple robots. In conclusion, the proposed interactive robotic assistant represents a significant step forward in bridging the gap between human needs and technological advancements. By combining cutting-edge AI technologies with practical hardware solutions, this work offers a scalable, efficient, and user-friendly system that enhances accessibility and user engagement in human-centric spaces.
วิทยาเขตชุมพรเขตรอุดมศักดิ์
Bone tissue scaffolds are made from biomaterials that support rapid repair and healing. Scaffold fabricators have produced materials that are able to degrade a biosystem or human body excellently. Thus, this work aims to study the optimization of materials, shape, and the 3D printing process with FDM. Finite element analysis is also used to predict mechanical properties of the scaffold and find the optimal shape and pore size. However, the materials studied include PLA, PCL, and HA.
คณะวิทยาศาสตร์
With the development of space technology, wide-field sky surveys using telescopes have expanded the range of new data available for time-domain astronomical research. Traditional data analysis methods can no longer respond quickly and accurately enough to the growing volume of data. Thus, classifying time-series data, such as light curves, has become a significant challenge in the era of big data. In modern times, analyzing light curves has become essential for using machine learning techniques to handle and filter through massive amounts of data. Machine learning algorithms can be divided into two categories: shallow learning and deep learning. Numerous researchers have proposed and developed a variety of algorithms for light curve classification. In this study, we experimented with Support Vector Machine (SVM) and XGBoost, which are shallow machine learning algorithms, as well as 1D-CNN and Long Short-Term Memory (LSTM), which are deep learning algorithms, which are branches of deep machine learning, to classify variable stars. The training and testing data used in this study were from the Optical Gravitational Lensing Experiment-III (OGLE-III), consisting of variable star data from the Large Magellanic Cloud (LMC), categorized into five main classes: Classical Cepheids, δ Scutis, eclipsing binaries, RR Lyrae stars, and Long-period variables. The results demonstrate the performance analysis of each machine learning algorithm type applied to light curve data, while also highlighting the accuracy and statistical metrics of the algorithms used in the experiments.