UMI-Bench 1.0: An Open and Reproducible Real-World Benchmark for Tabletop Robotic Manipulation with UMI Data
An open and reproducible real-world benchmark for tabletop robotic manipulation with UMI data; Xiaotong Li is the 7th author.
About
I am currently working as a VLA model researcher at Lumos Robotics, where I focus on robot manipulation and long-horizon tasks. My recent work includes depth-modality integration for VLA models, mask-guided object-centric manipulation, and improving non-Markovian long-horizon task execution through subtask decomposition and high-level memory modules. More broadly, I am interested in physical world models, VLA models, and robot learning that can understand and interact with the physical world.
I am seeking PhD opportunities in robotics, especially around physical world models, VLA models, and other related areas.
Education
2023/09-2025/10
MSc Robotics, Delft University of Technology
2019/09-2023/06
BEng Aircraft Design and Engineering, Nanjing University of Aeronautics and Astronautics
2022/07-2022/08
Robotics Summer School, Korea Advanced Institute of Science & Technology (KAIST)
Selected Projects
Learning Physics of Articulated Objects From RGBD Observation
Reconstructing a geometrically accurate, articulation-aware, and physically meaningful digital twin from RGB-D observations of articulated objects.
Building Explicit World Model for Open-world Object Manipulation
A manipulation framework that combines open-set perception, 3D digital twin reconstruction, and simulation-based strategy sampling without task-specific action demonstrations.
Model Predictive Control Approach for Multi-UAVs Planning and Motion Control
A model predictive control approach for trajectory tracking, collision avoidance, and planning of multiple UAVs in complex environments.
Two-stage Multi-UAV Planning Solution Combining A*, Model Predictive Control, and Artificial Potential Field
A two-stage planning approach for multi-UAV navigation using A*, MPC, and artificial potential fields.
Gradient Enhanced Neural Networks for Optimization of Mars Low Reynold Number Airfoil
A surrogate-modeling approach that uses gradient-enhanced neural networks to accelerate low-Reynolds-number airfoil optimization.
Dynamic Formation Transformation and Obstacle Avoidance of Unmanned Vehicles
A ROS-based unmanned vehicle project for formation transformation, mapping, navigation, and obstacle avoidance.
Working Experience and Internships
2025/12-now
Embodied AI Researcher, Lumos Robotics, Suzhou
2024/08-2024/12
Research Intern, TU Delft Autonomous Multi-Robots Lab, Delft
2023/07-2023/08
ROS Engineer, Bayer Crop Science, Shanghai
Awards
Competitions
- RoboMaster 2021 National, third prize
- RoboMaster 2021 Central Division, second prize
- RoboCup China Open 3D Detection, third prize
- RoboCup China Open Medical Robots Track, third prize
- Jiangsu Provincial Engineering Training Competition, second prize
- RoboMaster 2020 Online, second prize
Academic Honors
- First-Class Academic Scholarship, 3 times
- Second-Class Excellent Student Scholarship, 2 times
- Merit Student Honor, 2 times
Contact
Email: xiaotongli0105@gmail.nl · GitHub: Bojack-BJ