Dongjie Yu 余冬杰

Dongjie Yu PhD student PANDA Lab School of Computing and Data Science (CDS) The University of Hong Kong (HKU) Mail box 132, Floor 3, Chow Yei Ching Bulding, The University of Hong Kong, Pokfulam Road, Central and Western District, Hong Kong 999077 Email: djyu [at] connect [dot] hku [dot] hk ORCID · Google Scholar · Github

Brief Introduction

I am a PhD student in Computer Science at CDS @ HKU, fortunately advised by Prof. Jia Pan. I am interested in robot learning approaches, especially imitation learning (IL) and reinforcement learning (RL), for manipulation tasks. I was an intern at Xiaomi Robotics working on a teleoperation system with ViTai tactile sensors in 2025. I also interned at Shanghai Qizhi Institute till later in 2026, working on RL post-training for imitation robot policies.

I received the M.S. and B.S. degree in Mechanical Engineering at the School of Vehicle and Mobility from Tsinghua University in 2023 and 2020, respectively. At THU, I was supervised by Prof. Shengbo Li. I also collaborated with Prof. Jianyu Chen, who leads the Intelligent System and Robotics Lab at IIIS. During my time at THU, I was interested in RL combined with optimal control, especially safe RL (where agents should not violate certain state constraints during and/or after the training process), and their applications to autonomous driving and robotics.

News

• 07/2026: Our RL-100 paper was accepted by Science Robotics. Check the code base here.

• 10/2025: A follow-up work called BiKC+ was accepted by IEEE T-ASE. Congrats to Dr. Xu and Dr. Chen!

• 08/2024: Check our WAFR 2024 paper BiKC: Keypose-Conditioned Consistency Policy for Bimanual Robotic Manipulation, introducing a hierarchical policy for bimanual robotic manipulation. Congrats to all co-authors!

• 09/2023: Check our paper Safe Model-Based Reinforcement Learning with an Uncertainty-Aware Reachability Certificate accepted by IEEE Transactions on Automation Science and Engineering. Congrats to all co-authors!

• 05/2022: Our paper Reachability Constrained Reinforcement Learning was accepted for short presentation at ICML 2022. Thank all co-authors!