Research

1. Fault-Tolerant Workspace Analysis for Redundant Space Robots Experiencing Locked Joint Failures (NASA Kentucky EPSCoR RIDG award, PI: Biyun Xie, Funding Amount: $45,000, 8/1/2020 – 7/31/2021)
2. Human-Like Motion Planning of Collaborative Robots based on Human Arm Motion Analysis (SCEEE Development Fund, PI: Biyun Xie, Funding Amount: $32,500, 8/1/2020 – 7/31/2021)
3. Teaching Humanoid Robots by Demonstration with Preserved Dynamics and Adaptability Skills (UK College of Engineering Young Alumni Philanthropy Council Funding, PI: Biyun Xie, Funding Amount: $3,134, 1/1/2022-12/31/2022)
4. Autonomous Fault-Tolerant Operation of Redundant Robotic Arms (National Science Foundation, PI: Biyun Xie, Co-PI: Jiangbiao He, Funding Amount: $499,365, 9/1/2022-8/31/2025)
5. Developing a Demonstration-Based Motion Planner for Space Telerobots (NASA Kentucky EPSCoR RIDG award, PI: Biyun Xie, Funding Amount: $34,998, 4/1/2023 – 12/31/2024)
6. Automated Fabrication of Spherical Double-Helix Tensegrity (DHT) Robots (UK College of Engineering Mini-Grant Program, PI: Muhao Chen, Biyun Xie, Funding Amount: $10,000, 1/1/2025-5/15/2025)
7. CAREER: Safe and Reliable Human-Robot Shared Control for Robotic Telemanipulation in Complex and Extreme Environments (National Science Foundation, PI: Biyun Xie, Funding Amount: $602,698.00, 9/1/2025-8/31/2030)
8. CPS: Medium: An Autonomous Robotic System for Precision and High-Throughput Tomato Phenotyping in Large-Scale Greenhouses (National Science Foundation, PI: Biyun Xie, Co-PI:  Lauren Brzozowski, JiangBiao He, Hongkai Yu, Funding Amount: $1,179,584, 7/15/2025-7/14/2028)

1. Robot Motion Planning

Collision avoidance, optimal motion planning (optimizing path costs), constrained motion planning (satisfying specific constraints), and reactive and sensor-based motion planning are classic problems in robot motion planning. Numerous motion planning algorithms have been developed to tackle these essential problems, which can be categorized into four main groups:graph-based motion planners, sampling-based motion planners, optimization-based motion planners, and learning-based motion planners. Our lab focuses on developing new motion planning algorithms that leverage the strengths of each approach to effectively address these fundamental challenges in robot motion planning.

2. Fault-Tolerant Robots

For robotic systems operating in remote or hazardous environments, where routine maintenance is impractical or impossible, it is essential to proactively account for the likelihood of component failures. Such scenarios arise in high-stakes applications, including space exploration, nuclear waste remediation, and disaster response. Our research focuses on leveraging kinematic redundancy, i.e., robots with extra degrees of freedom, to design robust and fault-tolerant robotic systems. In contrast to conventional fault-tolerant control strategies that primarily emphasize post-failure recovery, our work adopts a fundamentally proactive approach. By integrating failure-awareness into both robot design and motion planning, our approach enables robots to maintain functionality and successfully complete tasks even after failures.

3. Human-Robot Shared Control

Human-robot shared control is a collaborative approach that blends human decision-making with robotic autonomy to enhance task performance and safety. Unlike fully autonomous systems or purely manual control, shared control allows both the human operator and the robot to contribute to the control process in real time. This dynamic partnership leverages the strengths of both parties, human intuition and adaptability combined with robotic precision and consistency. Shared control is particularly useful in complex or uncertain environments, such as assistive robotics, surgical systems, or remote exploration, where full autonomy may be risky or infeasible, yet purely manual operation is inefficient or cognitively demanding. Shared control also plays a crucial role in enabling data collection for modern robot learning frameworks. In particular, human-in-the-loop teleoperation provides high-quality demonstrations that are essential for training advanced models, including emerging Vision–Language–Action (VLA) systems. Our lab designs frameworks that tightly integrate autonomous motion planning with manual teleoperation, allowing robots to assist users by enforcing safety constraints, improving motion efficiency, and reducing operator burden, while still preserving human intent and control authority.

4. Robot Physical Intelligence

Robot physical intelligence aims to develop AI foundation models that enable robots to perceive, understand, and act effectively in complex, unstructured real-world environments. A central focus of our research is on Vision–Language–Action (VLA) models, integrated frameworks that unify visual perception, language understanding, reasoning, decision-making, and planning into a single end-to-end learning architecture. Building upon emerging VLA foundation models, our group develops novel, domain-adaptive VLA systems tailored to application-specific needs. To address the unique challenges, we design methods for data-efficient training, task-specific adaptation, and the integration of perception and action under real-world constraints.

5. Ethics in Human-Robot Interaction

Our lab collaborates with Dr. Jessica Barfield to study human-robot interaction, focusing on the ethical issues related to the design and use of social robots. Robot ethics, also known as roboethics, addresses the ethical concerns that arise from deploying robots in social settings. While much research in roboethics focuses on how humans are treated by robots, our work examines the ethical treatment of increasingly intelligent robots by humans. Specifically, we investigate whether the robot would be the subject of discrimination within society.