Blog posts

Humanoid Retargeting

1 minute read

Published: May 04, 2025

Intro

The large boost of humanoid robots this year largely thanks to the learning-based motion control and learning from human motion datasets. All the dancing, KongFu and athletic movements are based on this kind of retargeting from human datasets -> IL(Imitation Learning) based RL(Reinforcement Learning). After trying different retargeting and learning method and seeing different humanoid doing cool movements on stage, I think it is good for me to summarize it. This blog try to to do such thing.

Thoughts on Dexterous Manipulation Tasks

4 minute read

Published: May 04, 2025

🖐️ Two Types of Dexterous Manipulation Tasks: Functional Grasping vs. In-Hand Manipulation

Learning Stability: Reflections on Physical Constraints in Dexterous Manipulation

1 minute read

Published: May 02, 2025

Over the past few months, I’ve been exploring how physical constraints—like force closure and contact stability—can be introduced directly into the learning process for in-hand manipulation. While most recent progress in dexterous robotic skills has focused on end-to-end reinforcement learning using vision or tactile images, I found myself wondering: what if we teach policies to care about physical grasp stability?