Zixuan Huang

I am Zixuan Huang, a third year PhD student in Robotics at University of Michigan studying learning for robot manipulation. I'm very fortunate to be advised by Prof. Dmitry Berenson. Prior to UMich, I was a Master student in Robotics at Carnegie Mellon University advised by Prof. David Held.

I obtained a B.S. in Computer Science from the City University of Hong Kong, where I explored privacy-preserving machine learning and non-photorealistic rendering under the supervision of Prof. Antoni B. Chan, Prof. Cong Wang and Prof. Jing Liao

CV  /  Github  /  Google scholar

Email: zixuanh[at]umich.edu

profile photo
Research

My current research focuses on the intersection of robotics, computer vision, reinforcement learning and planning. My goal is to enable robot to learn, adapt and operate autonormously in the unstructured envrionments.

fast-texture Implicit Contact Diffuser: Sequential Contact Reasoning with Latent Point Cloud Diffusion
Zixuan Huang, Yinong He*, Yating Lin*, Dmitry Berenson
In submission (new)
Website / Paper

We enable the robot to reason about the changing contacts between objects and environment by predicting a sequence of NDF point clouds.

fast-texture Subgoal Diffuser: Coarse-to-fine Subgoal Generation to Guide Model Predictive Control for Robot Manipulation
Zixuan Huang, Yating Lin, Fan Yang, Dmitry Berenson
IEEE International Conference on Robotics and Automation (ICRA) 2024
Website / Paper

We use diffusion model to generate subgoals at appropriate temporal resolution dynamically to guide MPC

fast-texture Self-supervised Cloth Reconstruction via Action-conditioned Cloth Tracking
Zixuan Huang, Xingyu Lin, David Held
IEEE International Conference on Robotics and Automation (ICRA) 2023
Website / Paper

fast-texture Mesh-based Dynamics with Occlusion Reasoning for Cloth Manipulation
Zixuan Huang, Xingyu Lin, David Held
Robotics: Science and Systems (RSS) 2022
Website / Paper / Code

fast-texture Learning Visible Connectivity Dynamics for Cloth Smoothing
Xingyu Lin*, Yufei Wang*, Zixuan Huang, David Held
Conference on Robot Learning (CoRL 2021)
Website / Paper / Code

Learning a particle-based dynamics model on the visible part of the cloth enables efficient planning for cloth smoothing. We show that implicit occlusioning by graph imitation further improves the performance.

fast-texture Style Mixer: Semantic-aware Multi-Style Transfer Network
Zixuan Huang*, Jinghuai Zhang*, Jing Liao
Computer Graphics Forum 38(7):469-480, Proc. Pacific Graphics 2019
Paper / Code

A framework for semantic style transfer by patch attention.



Awesome template stolen from Jon Barron