🎓 Ph.D. Candidate @ Tongji University
🚗 Department of Automotive Engineering
🔬 Research Interests: 3D Generation · Diffusion Models · NeRF · Point Cloud Registration

I am a Ph.D. student in Vehicle Engineering (Intelligent Driving and Artificial Intelligence) at Tongji University (985 / Double First-Class), Shanghai.
Before that, I completed my B.Eng. in Vehicle Engineering (Automotive Electronics and Intelligence) at the same university.

My research lies at the intersection of Computer Vision, Computer Graphics, and Machine Learning, focusing on LiDAR-based 3D generation, multi-modal fusion, and scene reconstruction.
I am passionate about exploring how generative models (e.g., diffusion, neural fields) can enhance perception and simulation in autonomous driving.

• National Inspirational Scholarship, First Class (Top 5%)
• Outstanding Student Scholarship, First Class (Top 5%)

• Outstanding Student Scholarship, First Class (Top 5%)

• Yunnei Power Scholarship

• National Inspirational Scholarship, First Class (Top 5%)
• Outstanding Student Scholarship, First Class (Top 5%)
• Social Activity Scholarship

• Awarded "Outstanding Student of Tongji University" 4 times (Top 5%)

• "Excellent Graduate of Tongji University" (Top 4%)

(Under Review at AAAI 2026, First Author)
2024.06 – 2025.06
Unified text, image, and point cloud into a shared 3D layout representation, converted into semantic and depth control signals.
Adopted ControlNet to guide unconditional LiDAR generation based on Stable Diffusion, achieving high-quality “zero-to-simulation” generation with both flexibility and geometric consistency.

(Accepted at NeurIPS 2025, Fourth Author)
2023.09 – 2024.05
Proposed a hybrid framework alternating between neural reconstruction and geometric pose optimization.
Introduced selective weighted geometry constraints to improve pose-free NeRF optimization stability.

(Accepted at NeurIPS 2026, Fifth Author)
2024.05 – 2025.05
Developed MUP (Multimodal Unified Pose-free), a LiDAR–camera fusion framework for large-scale view synthesis without pose supervision.
Provided continuous depth supervision through LiDAR fields, improving robustness of pose-free image synthesis.

(Submitted to ICRA 2026, Third Author)
2024.09 – 2025.08
Explored 4D trajectory-conditioned video generation using HexPlane spatiotemporal representations combined with Stable Diffusion, DreamGaussian, and Stable Video Diffusion.
Achieved geometrically and temporally consistent video synthesis aligned with trajectory inputs.

NIO – Machine Learning Intern, Battery Cloud Department
2022.09 – 2023.01

• Used Python & MySQL to monitor vehicle data and analyze battery SOC/SOH.
• Designed PowerBI dashboards for real-time battery health visualization.

NIO – Product Experience Intern, Product Department
2022.09 – 2023.01

• Extracted user driving data and behavior patterns.
• Assisted in feature design through data-driven analysis.

• 🥇 National First Prize, China Computer Design Competition (2021)
• 🥇 Gold Award, China “Internet+” Innovation & Entrepreneurship Competition (2021)
• 🥉 Third Prize, Shanghai Computer Application Ability Competition (2020–2021)
• 🏅 Mathematical Modeling Competition, City-level Third Prize (2020)

• Intelligent Pavement for Autonomous Delivery (Leader) — 2020.03–2021.05
• Multi-user Intelligent Cockpit based on Human-size Adaptation Algorithm (Co-leader) — 2021.02–2021.08
  • Designed an adaptive cockpit system for shared vehicles (Top 1.3%, National Award)
• Battery Cell Research for FSAE Race Car (Co-leader) — 2021.03–2022.04

• Programming: Python, Linux, Git, MySQL, Web Development
• Languages: English, German, Italian
• Interests: Motorcycling, Badminton, Swimming, Singing

• Email: 2311399@tongji.edu.cn

“Exploring generative models for LiDAR-driven 3D perception and autonomous scene synthesis.”

⭐ © 2025 Haiyun Wei | Tongji University