Papers
23
Total Citations
473
H-Index
11
About
Yunduan Cui is a robotics and artificial intelligence researcher whose work sits at the intersection of reinforcement learning, robot control, and human-robot interaction. His research focuses primarily on developing stable, sample-efficient reinforcement learning algorithms and applying them to challenging robotic manipulation tasks, including the notoriously complex domain of cloth handling. Cui's most influential contribution — a deep reinforcement learning framework with smooth policy updates for robotic cloth manipulation (179 citations) — demonstrated that high-dimensional visual inputs could be effectively leveraged for dexterous robot learning. He has consistently addressed core limitations of reinforcement learning, tackling instability and data inefficiency through innovations such as kernel dynamic policy programming, relative entropy regularization, and contact-safe model-based approaches. His 2023 work on continuous dynamic policy programming reflects his ongoing commitment to principled, theoretically grounded algorithm design, extended further into multi-agent systems in 2024. Beyond manipulation, Cui has explored pneumatic artificial muscle control, quaternion neural networks for inverse kinematics, and adaptive human-robot motor skill learning, showcasing remarkable breadth. With over 380 cumulative citations, his contributions are shaping how robots learn safely and efficiently from experience — a foundation increasingly critical to deploying intelligent robots in real-world environments.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8Uncertainty-Aware Contact-Safe Model-Based Reinforcement Learning18 citations · 2021
- 9
- 10Deep dynamic policy programming for robot control with raw images14 citations · 2017