Grounded-SAM + VGN ROS2 Grasping Pipeline

• Built a ROS2-based perception-to-grasp pipeline integrating Grounded-SAM, RGB-D perception, semantic point cloud generation, TSDF construction, and VGN-based grasp candidate prediction.
• Designed the pipeline toward scene-aware robotic grasping and motion planning.

VGN Data Generation & Training Experiments

• Customized VGN data generation pipeline for TSDF-based robotic grasp prediction.
• Generated packed and piled synthetic grasping scenes and conducted VGN training experiments on the KHU Seraph GPU server.

4-DOF Robot Arm

• Designed the 4-DOF robotic arm in Fusion 360 and fabricated it using 3D printing.
• Implemented URDF modeling, RViz2 visualization, and inverse-kinematics-based end-effector control.

Smart Pet Feeder

• Built an ESP32-based IoT pet feeder with MQTT communication, RFID-based pet identification, load-cell weight sensing, and scheduled feeding control.
• Implemented the MQTT communication layer between the ESP32 device and the mobile app, including JSON command handling, schedule updates, and feeding status publishing.
• Awarded in a university software project competition.

AI Rookie Competition, 2026

• Advanced to the final round of AI Rookie Competition, a national undergraduate AI competition hosted by the Ministry of Science and ICT and organized by NIPA and TTA.
• Participating in a project-based AI challenge with technical education and research infrastructure support, including domestic AI model tracks connected with KT, LG AI Research, NC AI, SKT, and Upstage.

Robotics / Vision

ROS2 · Grounded-SAM · YOLO · Point Cloud Projection · Multi-view Fusion · Qwen

Robot Learning / Grasping

VGN · TSDF Representation · Synthetic Data Generation · Grasp Dataset Construction · Model Fine-tuning

Robotics / Hardware

Fusion 360 · URDF Modeling · Dynamixel · Inverse Kinematics

Programming / System

Python · C++ · Linux · CUDA Environment Setup · GPU Server Usage