Research

An on-device robotics framework that fine-tunes LLMs for household planning while grounding decisions in robot capabilities, memory, perception, and speech.

A controllable dataset generation engine that turns household robot tasks into scalable symbolic instruction-action pairs for instruction-grounded learning.

A competition-validated open-hardware mobile manipulator for RoboCup@Home that combines a modular ROS stack with on-device RDMM planning on embedded hardware.

A compact, compliant, and affordable humanoid mobile manipulation robot designed for safe household interaction and versatile everyday task execution.

A robotic painting system that turns text-to-image inpainting into physical canvas actions by switching between marker and eraser tools.

An interactive robotic art platform where multiple users collaboratively build a physical canvas through natural-language prompts, diffusion-based inpainting, and robotic drawing and erasing.

A RO-MAN 2023 paper on a robotic portrait drawing system designed to balance rendering speed and output quality on physical canvas.

A portrait-drawing robotic arm system that combines machine-learning-based sketch generation with physical canvas execution to produce recognizable portraits in under 80 seconds.