Yangzheng Wu

Team Lead, Embodied AI Senior Researcher at Huawei Canada Research Institute

profile/profile.jpg

Markham, Ontario, Canada

aaron1994wu@gmail.com

I am an embodied AI and computer vision researcher working on vision-language-action models, reliable robot perception, and 6DoF object pose estimation. I am currently a Team Lead and Embodied AI Senior Researcher at Huawei Canada Research Institute.

My recent work focuses on long-horizon robot behavior, robustness of vision-language-action systems, and simulation-to-real learning. During my PhD at Queen’s University, supervised by Dr. Michael Greenspan, I developed keypoint voting methods for RGB-D, point-cloud, and self-supervised 6DoF object pose estimation.

Highlights include technical leadership for The NorthStar Team, ranked 4th worldwide in the NeurIPS 2025 BEHAVIOR-1K Challenge, and main contribution to the best-performing method on ITODD in the BOP 2022 6DoF Rigid Object Pose Estimation Challenge.

selected publications

  1. arXiv
    Spatial Grafting: Grounding 3D Features for Flow-Matching Robot Policies
    Dingsheng Liu, Yangzheng Wu, Mahboubeh Asadi, and 5 more authors
    arXiv preprint arXiv:2609.35249, 2026
    HTML
  2. arXiv
    StageGuard: Learning Stage Transitions for Long-Horizon Robot Tasks via Agentic Distillation
    Jinbang Huang, Yuanzhao Hu, Zhiyuan Li, and 6 more authors
    arXiv preprint arXiv:2609.20791, 2026
    HTML
  3. RSS Workshop
    Data and Evaluation Protocols for Robust Robotic Manipulation
    Zhiyuan Li, Charles Eret, Xuan Zhao, and 2 more authors
    In Data-Centric Robotics Workshop at Robotics: Science and Systems (RSS), 2026
    HTML
  4. ICRA
    Distracted Robot: How Visual Clutter Undermine Robotic Manipulation
    Amir Rasouli, Montgomery Alban, Sajjad Pakdamansavoji, and 4 more authors
    In IEEE International Conference on Robotics and Automation (ICRA), 2026
    HTML
  5. arXiv
    How VLAs (Really) Work In Open-World Environments
    Amir Rasouli, Yangzheng Wu, Zhiyuan Li, and 4 more authors
    arXiv preprint arXiv:2604.21192, 2026
    HTML
  6. arXiv
    Do World Action Models Generalize Better than VLAs? A Robustness Study
    Zhanguang Zhang, Zhiyuan Li, Behnam Rahmati, and 11 more authors
    arXiv preprint arXiv:2603.22078, 2026
    HTML
  7. arXiv
    Mem2Ego: Empowering Vision-Language Models with Global-to-Ego Memory for Long-Horizon Embodied Navigation
    Lingfeng Zhang, Yuecheng Liu, Zhanguang Zhang, and 19 more authors
    arXiv preprint arXiv:2502.14254, 2025
    HTML
  8. ECCV
    Pseudo-keypoint RKHS Learning for Self-supervised 6DoF Pose Estimation
    Yangzheng Wu and Michael Greenspan
    In European Conference on Computer Vision, 2024
    HTML
  9. WACV
    Learning Better Keypoints for Multi-Object 6DoF Pose Estimation
    Yangzheng Wu and Michael Greenspan
    In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2024
    HTML
  10. 3DV
    Keypoint Cascade Voting for Point Cloud Based 6DoF Pose Estimation
    Yangzheng Wu, Alireza Javaheri, Mohsen Zand, and 1 more author
    In International Conference on 3D Vision, 2022
    HTML
  11. ECCV
    Vote from the Center: 6 DoF Pose Estimation in RGB-D Images by Radial Keypoint Voting
    Yangzheng Wu, Mohsen Zand, Ali Etemad, and 1 more author
    In European Conference on Computer Vision, 2022
    HTML Code