Qiyang Yan

I'm on the job market. Please reach out through email if you have good opportunities!

I am currently working as a research intern at Peking University, JD and AgiBot, advised by Prof. Hao Dong.

Previously, I obtained my Master’s degree in 2024 and my Bachelor's degree in 2023 in Electrical and Electronic Engineering from Imperial College London, advised by Dr. Adam J. Spiers at the Manipulation and Touch Lab (MTL).

Outside of research, I’m passionate about climbing, with a best sport climbing grade of 7b (onsight) and indoor bouldering grade of V10 (7C+).

We are looking for interns to work on dexterous grasping and manipulation. Feel free to reach out by email if you’re interested!

Email  /  Linkedin  /  X  /  WeChat

News

[Aug, 2025] ClutterDexGrasp accepted to CoRL 2025 and selected as Oral (~5% of the accepted papers)!

[Jan, 2025] VF-IHM accepted to ICRA 2025!

[Jun, 2024] I completed my Master's at Imperial College London!

[Jun, 2023] I completed my Bachelor at Imperial College London!

Publications

Introducing the first zero-shot sim-to-real closed-loop system for target-oriented dexterous grasping in cluttered scenes, demonstrating robust performance across diverse objects and layouts.

Introducing an end-to-end data-efficient learning framework for variable friction hand, allowing gripper to learn to precisely manipulate arbitrary objects (the first time) for any target pose on real hardware within 2 hours, with error around 3mm and 3°.

Introducing a unified Real2Sim2Real framework: TwinRecon replicates visually and geometrically accurate scenes via 3D Gaussian Splatting and 6D pose estimation; TwinRigid jointly optimizes robot-object dynamics with limited human-in-the-loop data collection.

Research Projects

Responsible for dataset preparation for development of generalisable learning-based approaches to bridge the gap between various type of tactile sensors.

Developed the first learning-based framework for the variable-friction gripper to learn to manipulate irregular polygons on real robot, achieving 95% success rate with average errors around 6 mm and 6°.

Developed a vision-based closed-loop pick-manipulate-insert system with a variable-friction gripper and UR5e robotic arm using ROS. Developed a model-based IHM planner and UR5e arm trajectory planner, achieving a 92% success rate for this task for cube with positional accuracy around 3mm.

Experience

Academic Service

Conference Reviewer: CoRL'25, IROS'25