Mingrui YU (于铭瑞)

Postdoctoral Researcher in Robotics

MSC Lab, UC Berkeley

Email: mingrui@berkeley.edu / mingruiyu98@gmail.com

About

I am a postdoctoral researcher in the MSC Lab at UC Berkeley, working with Prof. Masayoshi Tomizuka. I received my Ph.D. from the Department of Automation at Tsinghua University in June 2026, advised by Prof. Xiang Li.

Research Interest: My research focuses on physics-grounded robotic dexterity, with the goal of enabling reliable and generalizable manipulation across diverse objects, contact-rich interactions, and robot embodiments. I study how physical knowledge can make robot learning more reliable, efficient, and transferable. My work is guided by three principles: exploiting physical structure to make complex manipulation tractable, adapting to physical uncertainty through feedback and interaction, and scaling physical knowledge across tasks and embodiments. Methodologically, I combine physical modeling, optimization and control, and learning rather than treating dexterity as purely a data-scaling problem. Looking forward, I am particularly interested in using trustworthy physics simulation and human experience to generate scalable physical experience, and in transforming this experience into closed-loop, embodiment-aware, and adaptive manipulation skills.

Aspiration: My long-term goal is to build a physical foundation for general-purpose robot dexterity. I believe the next scaling frontier for embodied AI is physical: robots must not only understand what the world looks like or what tasks mean, but also learn how contact, force, motion, and embodiment shape what is physically possible. I hope to pursue this direction throughout my career, bridging rigorous robotics and scalable learning toward increasingly capable physical intelligence.

Recent News

07/2026 I joined the MSC Lab at UC Berkeley as a postdoctoral researcher, working with Prof. Masayoshi Tomizuka.

06/2026 I received my Ph.D. from the Department of Automation at Tsinghua University, was honored with the Outstanding Graduate of Beijing, Outstanding Graduate of Tsinghua University, and Outstanding Doctoral Dissertation of Tsinghua University awards, and gave a speech as the graduate student representative at the Department of Automation Graduation Ceremony.

06/2026 Our ICRA 2026 paper, "CoorGrasp: Coordinated Contact Control for Adaptive Dexterous Grasping Under Uncertainty," received the Best Poster Award at the ICRA 2026 Workshop on Dexterity with Multifingered Hands.

05/2026 I successfully defended my Ph.D. dissertation! 🎉

04/2026 Two papers accepted by RSS 2026 (contributed), titled "GS-Playground: A High-Throughput Photorealistic Simulator for Vision-Informed Robot Learning" and "Causal World Modeling for Robot Control".

03/2026 One paper accepted by RA-L, titled Kinematics-Aware Diffusion Policy with Consistent 3D Observation and Action Space for Whole-Arm Robotic Manipulation.

02/2026 One paper accepted by ICRA 2026 (Oral), titled "CoorGrasp: Coordinated Contact Control for Adaptive Dexterous Grasping Under Uncertainty".

01/2026 Two papers accepted by RA-P, titled "Analyzing Key Objectives in Human-to-Robot Retargeting for Dexterous Manipulation" and "Robust In-Hand Reorientation with Hierarchical RL-Based Motion Primitives and Model-Based Regrasping".

10/2025 Celebrating ten years of growing together! Yixi Li and I held our Wedding Ceremony in Qingdao, the hometown where our story first began. 🎉🎉🎉

10/2025 We successfully organized the IROS 2025 Workshop on Robotic Manipulation of Deformable Objects (ROMADO).

09/2025 We successfully organized the CoRL 2025 Workshop on Beyond Rigid Worlds: Representing and Interacting with Non-Rigid Objects.

03/2025 One paper accepted by RA-L, titled "Robotic In-Hand Manipulation for Large-Range and Precise Object Movement: The RGMC Champion Solution".

07/2024 One paper accepted by IJRR, titled "Generalizable Whole-Body Global Manipulation of Deformable Linear Objects by Dual-Arm Robot in 3-D Constrained Environments".

Show all ↓

Publications

HUGS: Guiding Unified Dexterous Grasp Synthesis Across Modes and Scales via Learned Human Priors

M. Yu*, Y. Jiang*, Y. Jia, K. Lv, X. Yan, L. Huang, Y. Ren, and X. Li

arXiv preprint arXiv:2607.04554, 2026

About

Recent News

Publications

HUGS: Guiding Unified Dexterous Grasp Synthesis Across Modes and Scales via Learned Human Priors

Towards Human-level Dexterous Teleoperation

UniLab: A Heterogeneous Architecture for Robot RL Beyond GPU-Dominant Paradigms

CoorGrasp: Coordinated Contact Control for Adaptive Dexterous Grasping Under Uncertainty

Arm-Aware Guided Dexterous Grasp Generation with Arm-Agnostic Grasp Models

Robust Model-Based In-Hand Manipulation with Integrated Real-Time Motion-Contact Planning and Tracking

UniStateDLO: Unified Generative State Estimation and Tracking of Deformable Linear Objects Under Occlusion for Constrained Manipulation

EmbodiSteer: Steering Embodiment-Agnostic Visuomotor Policies with Joint-Space Guidance for Zero-Shot Cross-Embodiment Deployment

UniLab: A Heterogeneous Architecture for Robot RL Beyond GPU-Dominant Paradigms

GS-Playground: A High-Throughput Photorealistic Simulator for Vision-Informed Robot Learning

Causal World Modeling for Robot Control

Kinematics-Aware Diffusion Policy with Consistent 3D Observation and Action Space for Whole-Arm Robotic Manipulation

Analyzing Key Objectives in Human-to-Robot Retargeting for Dexterous Manipulation

Robotic In-Hand Manipulation for Large-Range and Precise Object Movement: The RGMC Champion Solution

Generalizable Whole-Body Global Manipulation of Deformable Linear Objects by Dual-Arm Robot in 3-D Constrained Environments

Robust In-Hand Reorientation with Hierarchical RL-Based Motion Primitives and Model-Based Regrasping

In-Hand Following of Deformable Linear Objects Using Dexterous Fingers with Tactile Sensing

Contact-Implicit Model Predictive Control for Dexterous In-Hand Manipulation: A Long-Horizon and Robust Approach

Global Model Learning for Large Deformation Control of Elastic Deformable Linear Objects: An Efficient and Adaptive Approach

A Coarse-to-Fine Framework for Dual-Arm Manipulation of Deformable Linear Objects with Whole-Body Obstacle Avoidance

Learning to Estimate 3-D States of Deformable Linear Objects from Single-Frame Occluded Point Clouds

Shape Control of Deformable Linear Objects with Offline and Online Learning of Local Linear Deformation Models

Education

Research Experience

MSC Lab, UC Berkeley

Robbyant, Ant Group

Center of Embodied AI and Robotics, BIGAI

Robotics X, Tencent

MSC Lab, UC Berkeley

Professional Activities

Awards, Honors & Grants