Humanoid Robot

My humanoid robotics research focuses on coordinated whole-body control that combines locomotion and manipulation. I am developing learning-based and model-based methods for generating feasible robot motions, adapting human demonstrations to humanoid platforms, and executing complex tasks in simulation and on real robots.

My Role Description

1. Humanoid Whole-Body Teleoperation Pipeline

I designed and implemented an NVIDIA Isaac Sim-based teleoperation environment for whole-body humanoid motion control and demonstration data collection. I also developed a human-to-robot motion-retargeting pipeline that converts human motion into whole-body joint references for real-time humanoid teleoperation. The integrated workflow supports demonstration data collection on both simulated and real humanoid robots. Through this project, I gained experience in motion retargeting, whole-body control, NVIDIA Isaac Sim, Isaac Lab, and real-robot deployment.

2. IROS 2026 Humanoid IKEA Assembly Challenge

I am developing a precise base-positioning method that aligns a humanoid robot with manipulation targets using task-specific reference poses. I am also training and evaluating Vision-Language-Action policies for learning humanoid manipulation skills from demonstration data. The final system integrates reinforcement learning-based locomotion, precise positioning, and learned manipulation policies for autonomous furniture assembly. Through this project, I am developing experience in reinforcement learning, humanoid locomotion, precise positioning, Vision-Language-Action models, and manipulation policy learning.