Qixuan Huang 黄启烜

Robots that learn by watching us, and by trying for themselves.

M.S. Robotics student at Georgia Tech. Actively seeking Summer 2027 research internships.

Senior design, spring 2026. Our dual-arm rig solving a cube with two fixed grippers, at 4× speed. Project page
Fall 2026, in simulation. Two dexterous hands turning a cube in MuJoCo, with finger tricks I scripted by hand. Sharpa Wave hand models from MuJoCo Menagerie.
Qixuan Huang

About

I am an M.S. student in Robotics at Georgia Tech. Before that, I completed a dual-degree program in 2026, earning a B.S. in Computer Engineering from the University of Illinois Urbana-Champaign and a B.E. in Electrical and Computer Engineering from Zhejiang University.

As an undergraduate I worked with Prof. Liangjing Yang, Prof. Katie Driggs-Campbell, and Prof. Gaoang Wang. My research focuses on robot learning, dexterous manipulation, and perception.

Research question

How should a robot represent a dexterous skill it learns from a few human demonstrations, so the skill stays predictable and can adapt when a new setting brings uncertainty?

News

Aug 2026
Started my M.S. in Robotics at Georgia Tech.
May 2026
Completed my senior thesis, RecoverTwin: train-free failure recovery for frozen dual-arm policies on RoboTwin 2.0.
May 2025
Joined the Human-Centered Autonomy Lab at UIUC as an undergraduate research assistant.
Apr 2025
CityGen accepted to the CVPR 2025 Workshop on Urban Scene Modeling.
Jun 2023
Joined the CVNext Lab at ZJU, working on 3D city generation.
The dual-arm robot holding a Rubik's cube corner-down between its two white printed grippers
Project

Dual-Arm Robotic Rubik's Cube Solver

Senior design, team lead · Jan – May 2026. A MuJoCo cube model, Kociemba moves turned into clamp and layer-turn commands, and a serial bridge to Arduino motors. It completed a full solve on the physical prototype. Most of our hardware time went into friction and claw compliance, and that is a big part of why I moved to dexterous hands.
RecoverTwin architecture: a frozen base policy and an analytic failure detector feed an action gate; on detection, frozen localization, six-DoF grasp sampling, ranking and IK planning produce a recovery action.
Senior thesis

RecoverTwin: Train-Free Failure Recovery for Dual-Arm Manipulation

A train-free supervisor that notices when a frozen dual-arm policy stalls at the grasp and re-grasps before handing control back. It helped on some single-arm grasping tasks in RoboTwin 2.0 and hurt on bimanual ones.
CityGen: semantic and height fields stacked under a generated 3D city layout
CVPR-W 2025

CityGen: Infinite and Controllable City Layout Generation

Jie Deng, Wenhao Chai, Jianshu Guo, Qixuan Huang, Wenhao Hu, Jenq-Neng Hwang, Gaoang Wang
CVPR 2025 Workshop on Urban Scene Modeling (USM3D)
@InProceedings{Deng_2025_CVPR,
  title     = {CityGen: Infinite and Controllable City Layout Generation},
  author    = {Deng, Jie and Chai, Wenhao and Guo, Jianshu and Huang, Qixuan and Hu, Wenhao and Hwang, Jenq-Neng and Wang, Gaoang},
  booktitle = {Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR) Workshops},
  year      = {2025}
}
CityCraft: a generated 3D city stacked above its building mask and 2D layout
Under review

CityCraft: A Real Crafter for 3D City Generation

Jie Deng*, Wenhao Chai*, Junsheng Huang*, Zhonghan Zhao*, Qixuan Huang, Mingyan Gao, Jianshu Guo, Shengyu Hao, Wenhao Hu, Jenq-Neng Hwang, Xi Li, Gaoang Wang
arXiv 2024
@article{deng2024citycraft,
  title   = {CityCraft: A Real Crafter for 3D City Generation},
  author  = {Deng, Jie and Chai, Wenhao and Huang, Junsheng and Zhao, Zhonghan and Huang, Qixuan and Gao, Mingyan and Guo, Jianshu and Hao, Shengyu and Hu, Wenhao and Hwang, Jenq-Neng and Li, Xi and Wang, Gaoang},
  journal = {arXiv preprint arXiv:2406.04983},
  year    = {2024}
}

Course project HyperGemini: Autonomous Parallel Parking · Jan – May 2025 · ROS, GEM e2, 4/5 physical trials

Experience & Education

Research experience

  • Jan 2026 – May 2026
    RecoverTwin (Senior Thesis)
    Advisor: Prof. Liangjing Yang
    Train-free failure recovery for frozen dual-arm policies on RoboTwin 2.0.
  • May 2025 – Sept 2025
    Human-Centered Autonomy Lab, UIUC
    Undergraduate RA · Advisor: Prof. Katie Driggs-Campbell
    A behavior-tree prototype for recovering from failed ACT rollouts, which informed the design of RecoverTwin.
  • Jun 2023 – Sept 2024
    CVNext Lab, Zhejiang University
    Undergraduate RA · Advisor: Prof. Gaoang Wang
    Data pipelines (100k+ map and satellite patch pairs) and Blender rendering for CityCraft and CityGen.
  • Jun 2023 – Aug 2023
    Ultrafast Photonics Laboratory, ZJU
    Undergraduate RA · Advisor: Prof. Chaoyuan Jin
    A quantum-inspired Boids simulator; one coupling variant cut mean foraging time by 7.5%.

Education

  • 2026 – May 2028 (expected)
    M.S. in Robotics
    Georgia Institute of Technology · ECE home school
  • 2022 – Jun 2026
    B.S. in Computer Engineering, with honors
    University of Illinois Urbana-Champaign · Dean's List
  • 2022 – Jun 2026
    B.E. in Electrical & Computer Engineering
    Zhejiang University (dual degree) · GPA 4.00/4.3 · Second-Prize Scholarship, 2023 (top 8%)

Experiments

Continuously updated · All experiments

Format adapted from Wenhao Chai's Experiments.

Life

TravelTravelling with a camera. Next must-go destination: Switzerland.
SportsBasketball, Rugby and F1. Sanhao Cup basketball champs ×3, ZJU INTL.
GamesAll kinds of simulator games, such as MSFS and Transport Fever 3.