CV
Huashuo Lei (雷化硕)
MPhil Student in Robotics and Autonomous Systems
Summary
MPhil student working on embodied AI, robot learning, vision-language-action models, and long-horizon robotic manipulation.
Education
- MPhil in Robotics and Autonomous SystemsPresentThe Hong Kong University of Science and Technology (Guangzhou)Courses: Supervisor: Prof. Haoang Li
- B.S. in Robotics Engineering2025Harbin Institute of Technology
Work Experience
- Algorithm Intern深圳超维动力科技有限公司Contributed to embodied AI foundation model design and training-data development.
- Unified action spaces and velocity representations across multiple training-data sources.
- Expanded background-texture diversity to improve generalization in visually complex environments.
- Algorithm Engineer InternJan 2023 - Apr 2023Yijiahe Technology Co., Ltd.Worked on vision-guided control, hand-eye calibration, humanoid manipulator simulation, real-time trajectory planning, and assembly-line manipulation.
- Built pipelines with OpenCV, Eigen, PCL, MoveIt, and ROS.
Skills
Programming and machine learning
- Python
- C++
- PyTorch
- TensorFlow
- JAX
- CUDA
- Bash
Robotics and simulation
- ROS/ROS2
- MoveIt
- Isaac Sim
- MuJoCo
- Gazebo
- PyBullet
- OpenCV
- Point-cloud processing
Publications
- RoboMemArena: A Keyframe-Annotated Benchmark for Memory-Dependent Robotic Manipulation2026
- Static-Dynamic Disentanglement for Efficient Multi-Frame Vision-Language-Action Models2026
- EnvTrap: Revealing the Environment-Only Attack Surface in Embodied AI via Consequence-Blind Action Execution2026NeurIPS 2026, PosterSecurity analysis of environment-only attack surfaces in embodied AI.
- HCSG: Human-Centric Semantic-Geometric Reasoning for Vision-Language Navigation2026Preprint, 2026Human-centric semantic-geometric reasoning for vision-language navigation.
- Rethinking the Practicality of Vision-Language-Action Model: A Comprehensive Benchmark and An Improved Baseline2026
- High-Accuracy Early Recognition of Upper-Limb Motions for Exoskeleton-Assisted Mirror Rehabilitation2025IEEE Robotics and Automation Letters, 2025Early recognition of upper-limb motions for rehabilitation assistance.
- MonoAttack: A Strong Attack Framework with Depth-Migration and Attribute-Tampering for Monocular 3D Object Detection2025IJCNN 2025
- Manipulating the Bounding Box: Multimodal Controlled Backdoor Attacks on 3D Visual Grounding Models2025IJCNN 2025
- Exploring Disentangled Appearance-Motion Contexts for Temporal Activity Localization2025IJCNN 2025
Portfolio
- RoboMemArena2026
- WAM and robot learning projects2025Robot learningModel pre-training, robot data processing, evaluation, and embodied manipulation experiments.