Haoyang(Henry) Li
About Me
Hi!👋 My name is Henry Li, and I am a Research Scientist in the Department of Computer Science and Engineering at UC San Diego, where I am advised by Prof. Hao Su. I obtained my bachelor’s degree in Intelligent Manufacturing from Tongji University. I also had the opportunity to help with robot development and embodied AI research at startup Hillbot. During my undergraduate years, I also led an entrepreneurial initiative focused on healthcare robotics (see here) 🏄🏻♂️.
🏢 Office: Franklin Antonio Hall 3302 OR (Atkinson Hall 5202), 9500 Gilman Drive, La Jolla, CA 92093
Research Interests
- Robotic Manipulation: Mobile Manipulation, Reinforcement Learning, Imitation Learning
- Sim2Real: Real2Sim2Real Pipeline, Robotic Data Scaling, 3D Vision
- AI for Design: Mechanical Structure Optimization, Personalized Surgery
News
- [Jun. 2024] Made the first real-world robot demo of Hillbot.
- [Nov. 2023] Early participant in embodied intelligence startups Hillbot AI.
- [Aug. 2023] Joined UC San Diego as a Research Scientist.
- [May. 2023] Awarded with Outstanding Graduate of Shanghai.
- [Mar. 2022] First Prize in Chinese College Students Entrepreneurship Competition.
Publications
-
CVPR
IEEE Conference on Computer Vision and Pattern Recognition, 2025. -
RAL
IEEE Robotics and Automation Letters, 2023.PDF Project Page Shanghai Outstanding Graduate Thesis -
ASME
ASME Student Mechanism and Robot Design Competition (Finalist), 2022.PDF Project Page 2021 MIT Technology Review
Projects
VR-Enhanced Dual-Arm Dexterous Teleoperation
Innovative teleoperation system using Meta Quest 3 and Rokoko Gloves for precise robotic manipulation. Over 2,000 robot arm manipulation demonstrations were collected. Electrotactile stimulation will improve tactile feedback in the project's advanced VR technology and dexterous robotic hands.
Sim2Real Efficient Policy Training
Innovative research in robotic manipulation through advanced simulation techniques, bridging the gap between simulated environments and real-world applications. The project focuses on developing robust control strategies that can effectively transfer from simulation to physical robots.
Next-Generation Home Cleaning Robot
Next-generation robotic vacuum with flexible home use and lightweight robotic arm. It uses advanced navigation and manipulation to complete tasks that other cleaning robots cannot. System uses rule-based GraspNet for reliable object handling and ROS2 Navigation2 stack for effective navigation from garbage collection to board cleaning and box manipulation.
First Real-World Robot Demo of Hillbot
Grasp Pose Generation
This project developed an AI-driven grasping and design pipeline for robotic manipulation. It generated parallel-jaw grasp poses, validated them in simulation, and created a grasp motion planner, collecting over 200 demonstrations. Additionally, the pipeline optimized the humanoid robot's mechanical design, enhancing its grasping capabilities.
De-Icing Intelligent Robot
An intelligent robotic system designed for autonomous de-icing operations, combining advanced control systems with specialized hardware for efficient ice removal in challenging conditions.
Machining and Fabrication
Proficient in precision fabrication, utilizing laser cutting for aluminum and stainless steel sheet panels, 3D printing in PLA and soft TPU materials, and aluminum profile assembly. Experienced in creating support structures, including platforms for bimanual robotic arms and various robust frameworks for advanced robotics applications.
Services