My research connects robotic perception, physical interaction, and learning from sensor feedback. Nondestructive evaluation (NDE) is my primary research application, including autonomous, contact-aware ultrasonic inspection of reclaimed construction materials.
Robot Learning · Reinforcement Learning · Imitation Learning · Sim-to-Real
Developing learning-based approaches for robotic ultrasonic inspection using force-sensor feedback. Investigating force-guided contact-control policies and sim-to-real transfer using MuJoCo, NVIDIA Isaac Sim, and Isaac Lab.
Technologies
Robotic Perception · Multimodal AI · Computer Vision · Autonomous Inspection
Developed a vision-language-guided robotic inspection framework using Grounding DINO, SAM 2, CLIP, and Qwen2.5-VL-32B for material identification and contact-aware inspection planning.
Achieved 90.3% material-identification accuracy, 90.5% proper-contact selection, and an F1 score of 0.867 in the evaluated tasks.
Technologies
Imitation Learning · Reinforcement Learning · Sim-to-Real Transfer
Contact-Rich Manipulation · Force Feedback · Sensor-Guided Control
Object Detection · Instance Segmentation · 3D Reconstruction · Gaussian Splatting
Multimodal Reasoning · Scene Understanding · Contact-Aware Planning
NVIDIA Isaac Lab · Isaac Sim · MuJoCo · ROS 2
Nondestructive Evaluation · Ultrasonic Testing · Multimodal Sensing