Shuijing Liu
The University of Texas at Austin
shuijing.liu@utexas.edu
Bio
Shuijing Liu is a postdoctoral scholar in Computer Science Department at The University of Texas at Austin. She received her Ph.D. in Electrical and Computer Engineering Department at University of Illinois Urbana-Champaign in 2024. Her interest is at the intersection of human-centered robotics and machine learning. Specifically, Shuijing works on learning interaction models for robot navigation in challenging human environments. Besides, she is also interested in human-robot interaction through language. She has presented her work at premier robotics conferences such as ICRA, IROS, and CoRL. Prior to her Ph.D., she earned a Bachelor’s degree in Computer Engineering at University of Illinois Urbana-Champaign.
Areas of Research
- Robotics
Learning Structured Interaction Models for Robot Navigation in Human Environments
Robots are becoming increasingly prevalent in our daily lives. However, these autonomous agents work best in isolation, such as restricted areas in warehouses. Enabling these robots to coexist with humans remains an unsolved challenge, because subtle and dynamic interactions among different agents are difficult to infer. To pave the way for a broader deployment of robots in human spaces, I leverage the underlying structures in interactive scenarios to improve robot decision-making. For robots to navigate among humans, I developed an intention-aware reinforcement learning framework and an attention-based graph network, allowing the robot to learn a safe and socially aware navigation policy. For robots to fulfill human language commands such as “Bring me some water,” I developed visual-audio representations that allow robots to associate language with the visual world. In addition, I integrated the learned representations into end-to-end robotic systems, which demonstrate good generalization to novel environments and high user satisfaction in user study. My work highlights the synergy between human intention prediction and robot path planning. My future objectives are: 1) Developing a unified interaction model for various interaction modalities, such as motion, language, and gestures; 2) Enabling lifelong robot learning from non-experts in everyday settings.