cv

Thank you for visiting.

Basics

Name Shengzhe GAN
Label Undergraduate Researcher / Student
Email shengzhegan04@gmail.com
Url https://github.com/Lgx521
Summary Undergrad student in Information Engineering at SUSTech, enrolled in the Wo'loong Program.

Education

  • 2023.09 - 2027.06

    Shenzhen, China

    Bachelor's Degree
    Southern University of Science and Technology
    Information Engineering (Electronic and Electrical Engineering)
    • Robotics Motion and Control
    • Robotics Perception and Intelligence
    • Digital Signal Processing
    • Signals and Systems
    • Analog/Digital Circuits
    • Probability and Statistics
    • Linear Algebra

Research Experience

  • 2025.09 - 2025.12
    Remote Research Intern
    Visual Computing Group, Harvard University
    Topic: AREA3D Framework Development. Advisor: Dr. Fangneng Zhan.
    • Developed an active reconstruction agent unifying feed-forward geometric confidence with VLM-based semantic guidance.
    • Reproduced and optimized complex baselines including FisherRF and Air-Embodied.
    • Conducted large-scale 3D Gaussian Splatting experiments on Replica and OmniObject3D datasets, achieving SOTA PSNR/SSIM.
    • Implemented 'Dual-Field' uncertainty fusion using Monte Carlo ray sampling and frustum-based decay for NBV selection.
  • 2025.07 - 2025.09

    Notre Dame, IN, USA

    On-Site Research Assistant
    ROAR Lab, University of Notre Dame
    Topic: Active Perception. Advisor: Prof. Mengxue Hou.
    • Architected a complete grasping system implementing YOLO, SAM, and GraspNet within a ROS2 framework on Kinova Gen3 6DoF Arm.
    • Architected a complete grasping system implementing YOLO, SAM, and GraspNet within a ROS2 framework on a Kinova Gen3 6DoF arm.
    • Enhanced 3D Gaussian Splatting (3DGS) by embedding a 'graspness' metric into primitives to guide next-best-view selection.
    • Prototyped models with PointNet++ and Point Transformers to learn continuous Signed Distance Fields (SDFs).
  • 2025.03 - Present

    Shenzhen, China

    Undergraduate Researcher
    Robotics and Computer Vision Lab, SUSTech
    Topic: Reflection Awareness in SLAM system. Advisor: Prof. Hong Zhang.
    • Augmented Fast-LIO with a vision-based module to detect mirrors and filter reflective LiDAR points via 3D-to-2D projection.
    • Developing a real-time ROS node that uses vision-based masks to filter reflective artifacts from LiDAR scans.
    • Benchmarked the enhanced SLAM system against the baseline, validating improved trajectory accuracy (APE/RPE) and map quality.

Skills

Robotics & AI
ROS humble/ROS noetic
3D Gaussian Splatting (3DGS)
SLAM (Fast-LIO, Hector SLAM)
PyTorch
YOLO / SAM / GraspNet
PointNet++ / Point Transformers
Simulation & Tools
Habitat-Sim
CoppeliaSim
Kinova Gen3 Arm
ESP32

Projects

  • 2025.09 - 2026.01
    Vision-Language Grasping System
    Course project for Advanced Electronic Science Experiments II. Zero-Shot Robotic Manipulation via Large Vision-Language Models (VLM).
    • Zero-shot perception with YOLO-World + SAM (IoU > 0.9).
    • 6-DOF grasp planning via GraspNet-1Billion aligned to ROS2 TF.
    • Collision-free planning with Alpha Shape obstacle modeling + MoveIt2.
  • 2025.11 - 2025.12
    Full-Stack Network Simulation System
    Course project for Data Communication and Networks. End-to-End Protocol Stack Simulation from Physical Signals to Application Logic.
    • Cross-layer engine bridging physical signal processing with Layers 2–5.
    • BPSK/QPSK modulation, 2×2 MIMO with Zero-Forcing, Rayleigh + AWGN.
    • Hamming (7,4) FEC, Stop-and-Wait ARQ, and multi-hop routing.
  • 2025.04 - 2025.06
    Chinese Chess HCI Robot
    Course project for Robotics Motion and Control.
    • Engineered an end-to-end chess robot in ROS, integrating a KINOVA Gen3 Lite arm.
    • Deployed a custom-trained YOLOv8s model for real-time piece recognition.
    • Implemented custom optimization-based inverse kinematics (IK) solver.
  • 2025.03 - 2025.06
    Autonomatic SLAM Mobile Robot
    Course project for Digital Circuits Experiment.
    • Developed a wireless TCP/IP bridge using an ESP32 and virtual serial ports.
    • Deployed Hector SLAM for real-time 2D mapping, fusing LiDAR with odometry.
    • Engineered a data pipeline with custom ROS nodes to parse sensor streams.
  • 2024.04 - 2024.05
    Advanced 2048 Game System with AI Integration
    Java Application Project. Feature-rich 2048 game with Monte Carlo AI solver and secure user management.
    • Modular 2048 game engine supporting dual modes (Powers of 2 & 3) via extensible OOP.
    • Monte Carlo AI agent (1k iterations/step) for optimal move prediction and auto-play.
    • Secure user system with Base64 encryption and MD5 checksum for save-file integrity.
    • Rich UI with dynamic themes, music, Hammer props, and infinite Undo history.

Awards

Languages

Chinese
Native speaker
English
Intermediate, TOEFL iBT 96

Hobbies

Photography
Enjoy capturing moments and landscapes
Running
Enjoy long-distance running