CV简历
HE Tianlun (Allen) — academic CV.
HE Tianlun (Allen)
Embodied AI · Robot Learning · Computer Vision
Contact Information
| Name | HE Tianlun (Allen) |
| Professional Title | Embodied AI · Robot Learning · Computer Vision |
| theaj@connect.ust.hk | |
| Phone | +86 18216000982 |
| Website | https://allenhtl.com |
Professional Summary
Year-3 undergraduate at the Hong Kong University of Science and Technology (Integrative Systems and Design, Minor in Robotics) — top 10% of cohort. Embodied-AI researcher focused on vision-language-action (VLA) models, world models for manipulation, and sim-to-real reinforcement learning. Co-author of an ICML 2026 publication. Two-year Mechanical Lead at HKUST RoboMaster ENTERPRIZE; RMUC 2024 International Champion.
Experience
-
2023 - 2025 Hong Kong
Head of Mechanical Department
HKUST RoboMaster Team ENTERPRIZE
Mechanical lead for the team representing HKUST in the DJI-organised RoboMaster collegiate robotics championship.
- Team management — directed an 11-person mechanical team; owned seasonal technical roadmap, task breakdown, and parallel multi-robot R&D scheduling.
- Cross-department interface with electrical, vision, and hardware; standardised mech-electrical interfaces and routing to cut integration rework.
- Personal contributions — independent design of serial-leg infantry, rudder-wheel infantry gimbal, and quadcopter UAV (CAD → simulation → manufacture → integration → debug).
- Open-source impact — multiple core mechanical designs published to the RoboMaster community (article #542).
- Awards while in role — RMUC 2024 International Champion · RoboMaster 2025 University League Champion · RMUC 2025 National runner-up · RMUL 2024 regional runner-up.
-
2024 - 2025 China
Hardware Development Intern (Electronics / Hardware)
WeShareTech Limited
Laboratory chemical-reagent management — winter co-development project.
- Designed an RFID-based smart-weighing system for chemistry-lab reagent management; automated reagent identification and weight logging.
- Owned mechanical structure design (scale body), sensor + RFID module selection, and the system-integration plan.
- Built and validated the prototype end-to-end, from CAD to bench testing.
Education
-
2023 - 2027 Hong Kong, China
B.Eng., School of Engineering
The Hong Kong University of Science and Technology
Integrative Systems and Design (Minor in Robotics)
- Major — Integrative Systems and Design
- Minor — Robotics
- HKSAR Talent Development Scholarship, 2024/25
Publications
-
2026 MVISTA-4D: View-Consistent 4D World Model with Test-Time Action Inference for Robotic Manipulation
Jiaxu Wang, Yicheng Jiang, Tianlun He, Jingkai Sun, Qiang Zhang, Junhao He, Jiahang Cao, Zesen Gan, Mingyuan Sun, Qiming Shao, Xiangyu Yue
ICML 2026
A 4D world model that takes single-view RGBD input and produces multi-view-consistent 3D/4D scene representations, enabling an imagine-then-act decision paradigm for manipulation. End-to-end loop from sensor capture, multi-view reconstruction, dataset construction, and model training, to real-robot validation.
Projects
-
Memory VLA — long-horizon manipulation
Hierarchical task decomposition + fixed-capacity Implicit Memory Bank with cross-attention modulation. Research lead, 2026.02 – ongoing.
- Architecture — hierarchical task decomposition, sub-task boundary detector, fixed-capacity Implicit Memory Bank, memory-action cross-attention interface.
- Implementation — full PyTorch training & inference pipeline; multi-GPU distributed training; sim-to-real validation on a real arm.
- Evaluation — RMBench / VLABench / LIBERO-Long against Mem-0, MEM, MemER, CronusVLA baselines.
-
MVISTA-4D — View-Consistent 4D World Model (ICML 2026)
Research assistant, 2025.12 – 2026.02. Built the physical robot platform, multi-camera calibration / time-sync / cross-view alignment pipeline, and 4D manipulation dataset. Validated imagine-then-act on real hardware. Accepted to ICML 2026.
-
Indoor Structural Semantic Segmentation (Sonata + PTV3)
Researcher, 2025.12 – 2026.01. Point-cloud segmentation for ceiling / wall / floor / bay-window / glass classes. Built a Blender-based 3D point-cloud labelling plugin and end-to-end training pipeline (mIoU evaluation, class-imbalance tuning).
-
6-DOF Underwater Autonomous Robot
Full-stack — mechanical + embedded + algorithm, 2025.08 – 2025.12. Six-DOF cascaded PID (position + velocity loop), YOLO-based underwater target detection, mechanical sealing and buoyancy trim — end-to-end perception → planning → execution loop.
-
RL-based Bipedal Locomotion (sim-to-real prep)
Research assistant, 2024.12 – 2025.06. PPO training in IsaacGym with curriculum-shaped rewards; built a motor torque test bench and current-torque identification model for sim-to-real transfer.
-
Omnidirectional Autonomous Wheelchair
Team lead, 2024.09 – 2025.06. Rudder-wheel chassis + LiDAR-based autonomous navigation. Owned mechanical design, kinematics solver, embedded integration, and team coordination across mechanical / embedded / navigation tracks.
Awards
-
2024 RoboMaster 2024 RMUC International Region — Champion 🥇
DJI / RoboMaster
Top placement in the international region of RMUC, RoboMaster’s flagship national championship.
-
2025 RoboMaster 2025 University League — Region Champion (RMUL) 🏆
DJI / RoboMaster
-
2025 RoboMaster 2025 National (RMUC) — Second Place
DJI / RoboMaster
-
2024 HKSAR Government Talent Development Scholarship 🎓
HKSAR Government Scholarship Fund
Awarded for outstanding talent and development potential in non-academic dimensions.
-
2024 RoboMaster 2024 University League — Region Runner-up (RMUL)
DJI / RoboMaster