SYSun Yi

Sun Yi

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Sun Yi has a total of 34 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Robotics in Ionizing Radiation Environments: Vulnerabilities, Hardening, and Future Directions

    by Francis A. K. Djirackor, Josephine T. Djirackor, Amanuel Abrdo Tereda, Sun Yi, Younho Seong

    This review provides a comprehensive assessment of current advances in deploying robotic systems in ionizing‐radiation environments. It examines major radiation types and their effects on electronic, sensing, actuation, power, and communication subsystems, with particular attention to total ionizing dose, displacement damage, and single‐event effects. The review synthesizes reported component‐level vulnerabilities, radiation‐hardening strategies, shielding approaches, fault‐tolerant architectures, and emerging intelligent methods for monitoring and adaptation. Experimentally reported component‐level radiation effects are compared where data are available, while author‐developed models and illustrative simulations are explicitly distinguished from experimentally validated results. Key research gaps include standardized testing protocols, limited radiation‐robotics datasets, mixed‐field qualification, and the need for system‐level validation linking component degradation to mission performance. This work aims to help engineers and researchers in developing the next generation of radiation‐tolerant robotic systems.

  • Open Access

    Topology-Dependent Performance of Free-Space Photonic Quantum Networks Under Noise

    by Stefalo Acha, Sun Yi

    Photonic quantum communication enables secure and high-fidelity information transfer beyond classical limits, with direct relevance to emerging quantum networks operating in free-space environments. While physical-layer models of depolarizing noise, Gamma–Gamma turbulence statistics, entanglement swapping, and decoy-state QKD security bounds are individually well established, prior work typically treats these components in isolation or under fixed network assumptions. In this work, we develop a unified topology-aware analytical framework that simultaneously integrates free-space optical link budgets, turbulence-induced visibility degradation, depolarizing qubit noise, multi-hop entanglement cascade dynamics, teleportation fidelity thresholds, CHSH nonlocality certification, and asymptotic decoy-state secret key rate bounds across star, mesh, and ring graph structures. Rather than introducing new physical channel models, we demonstrate that identical physical links exhibit fundamentally different end-to-end performance once embedded within different network topologies. Mesh architectures minimize visibility cascade through hop-count reduction but incur quadratic hardware scaling. Star topologies minimize link count but concentrate noise and synchronization overhead at the hub. Ring configurations offer linear hardware scaling with multiplicative fidelity degradation. The results establish topology as a first-order design parameter in near-term free-space quantum networks operating without full quantum repeater infrastructures. While motivated by distributed multi-agent architectures, the framework applies broadly to terrestrial, airborne, and satellite-assisted photonic quantum communication systems.

  • Trajectory and Flow Optimization for Multi-Part, Multi-Location Pick-and-Place Tasks Using Nonlinear Model Predictive Control

    by Amanuel Abrdo Tereda, Sun Yi, Jagannathan Sankar, Yimesker Yihun, Richard Holdbrook

    This paper presents a comprehensive trajectory and flow optimization framework in multi-part, multi-location pick-and-place operations using Nonlinear Model Predictive Control (NLMPC). The proposed system enables a single robotic manipulator to execute multiple sequential pick-and-place actions across spatially distributed locations within a single operational cycle, significantly improving throughput and flexibility in industrial automation tasks. A central contribution of this work is the introduction of terminal cost penalization in the NLMPC formulation, targeting joint velocity and acceleration at the end of the trajectory to enable smooth and precise motion termination. Additionally, Euclidean distance constraints are incorporated to enhance the final pose accuracy of the end effector. The system is validated through extensive simulation experiments using a KINOVA Gen3 robotic arm in both obstacle-free and obstacle-present environments, where object and obstacle positions are predefined. Results show that penalizing the cost function improves end-effector precision, reducing Euclidean distance error by 35.9% in the obstacle-free case and 10.6% in the obstacle-present scenario. The NLMPC framework also maintains real-time feasibility, with an average computation time of 0.045 seconds per control update, well below the 0.55-second control loop interval. These findings confirm the practical viability of the proposed approach for high-performance, constraint-aware robotic control. Supplementary materials, including simulation videos and open-source MATLAB code, are provided to support reproducibility and future research. Note to Practitioners—Modern industrial automation systems demand flexible and precise robotic solutions capable of handling diverse objects in dynamic workspaces. This paper presents an NLMPC-based trajectory planning strategy tailored for multi-part, multi-location pick-and-place operations. The proposed approach supports flexible, precise, and collision-free robotic motion, making it well-suited for advanced material handling applications. Incorporating penalties on terminal velocity and acceleration leads to smoother stops, increased placement precision, and reduced mechanical wear, while applying Euclidean distance constraints ensures accurate final positioning of the end-effector. The method is computationally efficient and runs in real time on a standard CPU. Although the simulations assume known object locations, the framework is extensible to real environments with integrated perception. Open-source code and videos are provided to support adoption and replication.


  • Supervised Classification Model for Neural Correspondence to Environmental Uncertainty in Multiple Cue Judgment System with Decision Support

    by Yoo-Sang Chang, Younho Seong, Sun Yi

    Human agents play an essential role in the multi-agent system (MAS) for appropriate action with information from autonomous robot agents with cutting-edge artificial intelligence (AI) technologies. Human agents can make decisions for proper action by evaluating system performance states, such as valid or invalid system performance caused by system errors under environmental uncertainties. Thus, the decision support system (DSS) equipped with certain levels of AI can be useful in MAS through data analysis on the information from agents to assist human agents. However, humans can make inappropriate decisions because of an unreliable decision-making environment under environmental uncertainty. This study aims to capture human neurological responses and then classify the cognitive states interacting with unreliable system environments, using concepts of expected and unexpected uncertainties in the multiple cue judgment system with decision support. We captured the brain’s response through EEG and then the cognitive state classification model was developed with supervised machine learning algorithms based on data captured. This study shows that K-Nearest Neighbors (KNN) shows 70 % classification performance above as the best and generally acceptable performance, and we could find that baseline classes, to remove temporal drift in EEG data preprocessing, hindered classification performances.

  • Open Access

    Cooperative Intelligent Control of Multi-Agent Systems (MAS) Through Communication, Trust, and Reliability

    by Stefalo Acha, Sun Yi

    The field of Multi-Agent Systems (MAS) has achieved significant advancements in modern research and development. This study focuses on enhancing trust evaluation, communication efficiency, and adaptive navigation in scenarios where agents have limited prior knowledge. Key contributions include the development of a high-intelligence MAS system that integrates key input data, such as real-time parameters regarding agents’ distances from one another, their distances to target locations, weather conditions, visibility, machine learning capabilities, positions relative to safe or unsafe environments for trust evaluation, delays in communication, and potential cyber threats. These factors trigger a dynamic topology-switching mechanism to secure agents or minimize communication delays in high-security operations. The MAS implements these strategies based on an adaptive communication model, enabling agents to execute various steps during data pooling effectively. Agents utilize real-time data to coordinate flock movements, ensuring dynamic and robust control through data pooling. For example, in a topology requiring a lead agent, the lead agent provides navigation instructions based on pooled data collected during mission execution. This data may involve repositioning proper area coverage, considering agents’ visibility, distance, or environmental disturbances. Four main topologies are implemented in this experiment: directed mesh with two lead agents (type A), directed mesh with one lead agent (type B), star topology (type C), and ring topology (type D). Type B and C topologies are well-suited for communication without delays or disturbances but perform poorly when the system experiences delays (e.g., noise disturbances exceeding a threshold frequency of 5 Hz). In contrast, type A and D topologies are more effective in handling communication delays. By implementing a topology-switching mechanism, this research streamlines the application of two or more topologies in real-life scenarios. It utilizes type B or C topologies in undisturbed conditions and switches to type A or D when perturbations occur. This optimization minimizes communication delays during mission execution and flight time. The research demonstrates significant improvements in trust evaluation, communication efficiency, and overall MAS performance, with implications across various domains, including image and video mining. In these areas, the integration of domain-specific agents enhances processes such as preprocessing, feature extraction, and interpretation. Results show improved accuracy and reliability in data analysis and decision-making across diverse applications, particularly in scenarios involving complex spatial objects and varying environmental conditions.


  • Open Access

    A Comprehensive Assessment of Decision Making, Communication and Data Extraction in Multi-Agent System

    by S. N. N. Acha, S. Yi, S. M. Shenouda

    This research focuses on assessing the application of agent-based communication and self-navigation to achieve tasks such as image fusion. In this multi-agent system experiment, agents are tasked with a common goal: accessing a shared environment with no more than a 5% overlap between each other. The resulting data—whether images, video, or other types of recordings—are fused or merged to create a comprehensive view of the accessed area. Various communication patterns (topologies) are employed within the multi-agent system to minimize communication delays. Agents utilize Mesh, Star, or Ring topologies as needed, and based on the topology applied at any given instance, a lead agent assists with baseline orientation, referencing the positions and angles of other agents for efficient navigation and data collection. Through this collective effort, the system aims to create a comprehensive representation of the object or dataset by combining multiple images. This synthesis enhances the system's capacity for thorough analysis and informed decision-making. At its core, the primary objective of this endeavor is to address existing research gaps concerning communication self-navigation and image (data) fusion within multi-agent systems. The focus lies in developing a methodology that enables agents to autonomously navigate and communicate, guided by their specific interests in target objects. This autonomy empowers agents to efficiently gather data and collaborate towards identifying and analyzing objects of interest. Central to achieving this objective is the optimization of communication protocols within the multi-agent system. This optimization involves transitioning between different topology systems, each meticulously designed to handle delays caused by specific perturbations affecting the system. By adapting the communication topology to the prevailing conditions, the research aims to enhance the efficiency and reliability of data exchange among agents. Moreover, the research endeavors to strengthen coordination within the multi-agent system, particularly in scenarios where communication is challenged or contested. This is achieved through the implementation of refined data pooling strategies, wherein agents collaborate to aggregate and analyze data effectively. By optimizing data pooling methodologies, the research aims to improve the accuracy and reliability of decision-making processes within the system. An essential aspect is conducting a comprehensive literature review to identify potential enhancements in agents' communication strategies based on the importance of the data being exchanged. By leveraging insights from existing research, the development of robust communication protocols tailored to the specific needs and objectives of the multi-agent system is presented.


  • Open Access

    Control of Leading-Edge Shock of Train Using Deep Neural Network to Prevent Unstart

    by Stefalo Acha, Sun Yi, Frederick Ferguson

    The primary aim of this research is to create a comprehensive neural network model that can effectively regulate the position of the leading-edge shock in a scramjet by manipulating the required backpressure, thereby achieving, and maintaining hypersonic speeds. By utilizing computational fluid dynamic data, a dynamic model is constructed using a neural network-based approach to control the positions of the leading-edge shock train. The scramjet isolator, which is a duct where pressure increases from the inlet to the combustor via a series of shock waves, necessitates precise control of the leading-edge shock locations during scramjet operation. The model employed in this research project is a neural network adaptive controller implemented in MATLAB/Simulink software, which accounts for the nonlinear characteristics of the plant and predicts its future behavior. To enhance control performance, a robust controller is employed, integrating a learning rule that reduces the error percentage throughout the system's lifespan. The neural network is trained using flight behavior datasets, enabling it to learn from a set of training patterns. Plant identification is achieved through a neural network, capturing the system dynamics, and enabling the neural network to function as a controller. Additionally, the controller's performance is validated through simulations and optimization analyses. This research presents an adaptable, robust, and effective control system that provides added reliability and reduces disturbances.


  • Open Access

    Network-Based Intrusion Detection for Industrial and Robotics Systems: A Comprehensive Survey

    by Richard Holdbrook, Olusola Odeyomi, Sun Yi, Kaushik Roy

    In the face of rapidly evolving cyber threats, network-based intrusion detection systems (NIDS) have become critical to the security of industrial and robotic systems. This survey explores the specialized requirements, advancements, and challenges unique to deploying NIDS within these environments, where traditional intrusion detection systems (IDS) often fall short. This paper discusses NIDS methodologies, including machine learning, deep learning, and hybrid systems, which aim to improve detection accuracy, adaptability, and real-time response. Additionally, this paper addresses the complexity of industrial settings, limitations in current datasets, and the cybersecurity needs of cyber–physical Systems (CPS) and Industrial Control Systems (ICS). The survey provides a comprehensive overview of modern approaches and their suitability for industrial applications by reviewing relevant datasets, emerging technologies, and sector-specific challenges. This underscores the importance of innovative solutions, such as federated learning, blockchain, and digital twins, to enhance the security and resilience of NIDS in safeguarding industrial and robotic systems.