CV

Education, experience, selected publications, and projects.

Contact Information

Name Qirui Hu
Professional Title Master's Student in Electronic Information
Email 19872217049@163.com
Phone +86 198 7221 7049
Location Beijing,
Website https://hqr-robotic.github.io

Professional Summary

Master’s student at the Institute of Computing Technology, Chinese Academy of Sciences. Research interests include embodied intelligence, imitation learning, and road perception in semi-structured outdoor environments.

Education

  • M.S.
    Institute of Computing Technology, Chinese Academy of Sciences
    Electronic Information
    • GPA: 3.82/4.0
    2024.09 - 2027.06

    Beijing, China

  • B.E.
    China Agricultural University
    Artificial Intelligence
    • GPA: 3.85/4.0; Major rank: 1/35
    • English: CET-4 565; CET-6 522
    2020.09 - 2024.06

    Beijing, China

Experience

  • Embodied Intelligence Intern
    LimX Dynamics
    Online data collection, model training, and simulation evaluation for VLA systems.
    • Personal contribution: Independently designed and developed the model-decoupled real-robot DAgger system FluxDAgger. As an open-source contributor, maintain the VLA engineering platform FluxVLA and the standalone evaluation framework FluxThemis.
    • FluxDAgger: Built the full stack for AgileX Piper dual-arm robots, from hardware control to the data loop. Redesigned the four-arm CAN-bus topology for independent arm control; decoupled policy inference, reward modeling, collection control, and data processing in a multi-process framework; and implemented multi-camera synchronization, online human takeover, corrective data collection, and Qwen3-VL reward visualization.
    • FluxVLA: Extended model and simulation capabilities by integrating FastWAM’s uncond, joint, and IDM paradigms together with Wan2.2, MoT, and ActionDiT components. Connected data preprocessing, distributed training, and LIBERO evaluation, and integrated LIBERO-Plus for zero-shot evaluation of standard LIBERO checkpoints, multi-task rollouts, and result aggregation.
    • FluxThemis: Built the LIBERO-Plus evaluation pipeline for a standalone multi-model, multi-simulator VLA framework. Used ROS to decouple model inference from simulation evaluation and separate Conda environments to isolate LIBERO, LIBERO-Plus, and robosuite dependencies. Covered 10,030 robustness tasks with metadata filtering, grouped success-rate statistics, and parallel multi-process rollouts.
    2025.11 - Present

    China

  • AI Algorithm Intern
    Beijing Humanoid Robot Innovation Center
    Teleoperation data collection, manipulation policy training, and visual generalization.
    • Teleoperation and data loop: Built a ROS-based Haption + UR5e teleoperation data collection system, designed robot manipulation tasks, collected multi-task datasets, and conducted real-robot training.
    • Policy and visual generalization: Reproduced Diffusion Policy, integrated hand-eye calibration into the policy pipeline, and explored multi-camera and BEV representations to improve cross-view and cross-background generalization of compact models.
    • Research output: Contributed to OmniD, an image-based BEV multi-view fusion policy study published on arXiv.
    2024.07 - 2025.01

    Beijing, China

Projects

  • Road Detection in Semi-Structured Outdoor Environments

    Research project on end-to-end road and boundary-line detection in outdoor scenes without artificial lane markings but with visible road trajectories.

    • Benchmark construction: Selected semi-structured scenes from DeepScene, YCOR, and GOOSE. Used an enhanced Labelme curve annotation tool to label left and right road boundaries with high-order Bézier curves and pair them with road instances, creating the SUNSET dataset and benchmark.
    • Method design: Contributed to TEAR, which uses interchangeable dual instance decoders to disentangle road and boundary-line instances and hierarchical bipartite matching to associate them.
    • Research output: TEAR of the SUNSET introduces a benchmark and end-to-end method for semi-structured road detection and was accepted by IEEE ICME 2026 (CCF B).
  • Gesture Recognition and Mobile-Robot Deployment on Horizon X3 Pi

    Undergraduate thesis on edge gesture recognition, inference optimization, and real-robot deployment.

    • Model training: Trained YOLOv5s on HaGRID, achieving 86.60% mAP, and compared it with models from the YOLO and DETR families.
    • Edge optimization: Deployed and optimized the model on Horizon X3 Pi, increasing inference speed from 20 FPS to 30 FPS with only a 0.36% mAP decrease.
    • Robot integration: Developed ROS2 motion-control nodes, integrated gesture recognition and human tracking, and validated the system on a real robot.

Publications

Awards

  • 2021, 2023
    National Scholarship (two awards)
    Ministry of Education, China
  • 2021, 2023
    Jinlongyu Scholarship (two awards)
    China Agricultural University

    Comprehensive rank 1st.

  • 2020
    First-Prize Academic Scholarship
    China Agricultural University
  • 2024
    Outstanding Graduate
    China Agricultural University

Competitions

  • 2021, 2022: First Prize (Beijing), National Undergraduate Mathematical Contest in Modeling, Group A.
  • 2022: First Prize (Beijing), National Undergraduate Mathematics Competition, non-math category.
  • 2021: Second Prize (Beijing), National Undergraduate Mathematics Competition, non-math category.
  • 2022: Meritorious Winner (M Award, Top 7%), Mathematical Contest in Modeling.

Student Activities

  • Director, External Communications Department, New Media Center (2021-2023): Produced and reviewed school news posts and managed the official WeChat account.
  • Member, Publicity Department, Graduate Student Union (2024-2026): Produced event posts, handled on-site photography, and organized publicity materials.