Papers
11
Total Citations
255
H-Index
7
About
Shiyu Jin is a leading researcher in robotic manipulation, with a primary focus on the challenging domain of deformable object manipulation, particularly cables and belts. His work addresses the core difficulty of modeling and controlling objects with high degrees of freedom and unpredictable deformation. Jin’s major contributions include pioneering a hybrid offline-online learning framework using Graph Neural Networks to predict cable dynamics, achieving robust manipulation with over 60 citations. He has also developed novel frameworks like SPR-RWLS for real-time cable tracking and local deformation approximation, and a trajectory optimization formulation for assembling belt drive units. Beyond deformable objects, Jin has advanced robotic assembly by learning insertion primitives with discrete-continuous hybrid action spaces, and has pushed the boundaries of sample efficiency in reinforcement learning by integrating Large Language Models through his RLingua framework. His recent work on RT-Grasp explores reasoning-based grasping via multi-modal LLMs. With over 250 total citations and a consistent stream of high-impact publications from 2019 to 2024, Jin’s research is instrumental in enabling robots to perform complex, real-world tasks that require both precise control and adaptive learning.
Research Focus
Key Achievements
Top Papers
- 1
- 2Robotic Cable Routing with Spatial Representation46 citations · 2022
- 3Robust Deformation Model Approximation for Robotic Cable Manipulation41 citations · 2019
- 4
- 5
- 6
- 7
- 8
- 9Reasoning Grasping via Multimodal Large Language Model4 citations · 2024
- 10VIHE: Virtual In-Hand Eye Transformer for 3D Robotic Manipulation2 citations · 2024