About Me
Hello. I am now living in Guangdong, ShenZhen, pursuing my Master's degree in Control Science and Engineering in Harbin Institute of Technology(ShenZhen). Also,I received my bachelor's degree from HITSZ. Now I am a member of nROS-lab, who is doing research and projects on Vision-Language-Action Model and 3D Scene Online Understanding.
Education and Internship
Internship
LimX Dynamics - Embodied AI and Manipulation Group Nov 2024 - Apr 2025
Master's
Harbin Institute of Technology (Shenzhnen) - Control Science and Engineering @nROS-lab Sep 2024 - Present
Bachelor's
Harbin Institute of Technology (Shenzhnen) - Automation @nROS-lab Sep 2020 - Jun 2024
Projects
Apple Vision Pro Teleoperation
We implemented the Demo of TeleVison. We use the realman RM65 arm and the Inspire dexterous hand to build the experiment platform.The teleoperation system can be used to collect dual arm mobile platform joints and sensors data for imitation learning and embodied AI. Hand retargeting is solved using dexipilot.The message communication is implemented by gRPC. The camera streaming relys on the webrtc.
LimX VGM
LimX VGM: A Video Generation Model for Embodied Manipulation. By fine-tuning existing video generation models with human manipulation data, it can achieve task understanding, decomposition, object manipulation trajectory generation, and robot execution with only scene images and task instructions as prompts. The entire process uses zero real robot sample data, achieving generalization across multiple platforms. This is the first time in China that human manipulation data has been directly applied to robot manipulation.
VR Immersive Visualization
For complex and dangerous post-disaster scenarios, sometimes the autonomous intelligence of existing robots cannot be fully relied upon, requiring human intervention. We use R3LIVE as the odometry to provide RGB colored point clouds for VR users, enabling real-time visualization of the robot's environment. I use OpenXR for VR development programming to achieve high-performance rendering.
Camera Surround View Stitching
The first "SparkLink Cup" University Application Challenge is a themed application challenge for university developers, jointly organized by the International SparkLink Alliance and multiple universities. The competition topic is the development of a camera surround view stitching system, requiring the completion of unmanned vehicle bird's-eye view fusion and stitching on a specified platform. The requirements include no obvious traces at stitching points, balanced global image brightness, no obvious distortion, no obvious delay, and smooth and stable operation. We use C++ for program development and have optimized multi-threading and CUDA acceleration, ultimately achieving a processing speed of 28ms per frame. Team 1504Ders from Harbin Institute of Technology (Shenzhen) won the second place in the first "SparkLink Cup" University Application Challenge.
Tetris Robot
We created a robot that plays Tetris. Using a robotic arm and an RGB camera, it can pick up Tetris pieces and place them on the board according to specified rules.