Huaiyuan Xu

Postdoctoral Fellow @ Department of Electrical and Electronic Engineering (EEE), The Hong Kong Polytechnic University (PolyU)

My research centers on Embodied Agents, aiming to develop better end-to-end perception-action systems operating in the physical world. I currently focus on three key areas:

  1. Vision-Centric Perception – Advancing understanding and forecasting in the 3D world through vision-based observation.

  2. Robustness Across Modalities – Ensuring the robust performance of AI models across various sensors used by agents.

  3. Efficiency For On-Board Deployment – Pushing the efficiency frontier of AI models to enhance cost-effectiveness.

Before coming to PolyU, I received my PhD degree from Tianjin University (TJU), advised by Prof. CHEN Xiaodong. I completed both B.Eng and M.Eng study at Tianjin University. Currently, I work closely with Prof. CHAU Lap-Pui and Dr. WANG Yi.

I am open to collaborations (anytime & anywhere & any type). If you are interested in collaborating with me, feel free to reach out via email at huaiyuan.xu@polyu.edu.hk.

news

May 1, 2026 🎉 One paper was accepted at ICML (CCF A).
May 18, 2025 🎉 One paper was accepted at Information Fusion (IF: 15.5).
Jan 23, 2025 🎉 One paper was accepted at ICLR (CCF A).
Sep 4, 2024 🎉 One paper was accepted at Information Fusion (IF: 15.5), with GitHub link.
Jul 2, 2024 🎉 One paper was accepted at IEEE Transactions on Intelligent Vehicles (IF: 14.3).

selected publications

  1. ICML’26
    RoboFlow4D: A Lightweight Flow World Model Toward Real-Time Flow-Guided Robotic Manipulation
    Sixu Lin, Junliang Chen, Huaiyuan Xu#, Zhuohao Li, Guangming Wang, Yixiong Jing, Sheng Xu, Runyi Zhao, Brian Sheil, Lap-Pui Chau, and Guiliang Liu
    In Forty-third International Conference on Machine Learning 2026
  2. ICLR’25
    OccProphet: Pushing Efficiency Frontier of Camera-Only 4D Occupancy Forecasting with Observer-Forecaster-Refiner Framework
    Junliang Chen*, Huaiyuan Xu*, Yi Wang, and Lap-Pui Chau
    In 2025 International Conference on Learning Representations 2025
    Article Website
  3. Inf. Fusion
    A Survey on Occupancy Perception for Autonomous Driving: The Information Fusion Perspective
    Huaiyuan Xu, Junliang Chen, Shiyu Meng, Yi Wang, and Lap-Pui Chau
    Information Fusion 2025
    Article Website
  4. T-IV
    C2L-PR: Cross-modal Camera-to-LiDAR Place Recognition via Modality Alignment and Orientation Voting
    Huaiyuan Xu, Huaping Liu, Shoudong Huang, and Yuxiang Sun
    IEEE Transactions on Intelligent Vehicles 2024
    Article Code
  5. T-CSVT
    Learning Semantic Alignment Using Global Features and Multi-Scale Confidence
    Huaiyuan Xu, Jing Liao, Huaping Liu, and Yuxiang Sun
    IEEE Transactions on Circuits and Systems for Video Technology 2023
    Article Code