-1e293b?style=flat-square){kind=icon}
-15803d?style=flat-square){kind=icon}
I am Zhilun Hu (胡至伦), a Ph.D. student in the School of Mechanical Science and Engineering at Huazhong University of Science and Technology (HUST), advised by Prof. Weiming Shen (IEEE Fellow, Academician of CAE).
My research centers on Reliable Robotic Manipulation—investigating how physical agents actively acquire unoccluded observations, verify high-level action proposals against geometric constraints prior to execution, adapt policies through visuotactile interaction, and recover from execution anomalies.
我是胡至伦，华中科技大学机械科学与工程学院博士研究生，师从**沈卫明教授。我的研究致力于解决机器人操作中的可靠性问题（Reliable Robotic Manipulation），构建从主动视觉伺服感知**、双臂柔性操作先验视觉校验、视触觉强化学习策略自适应到故障恢复与经验记忆的完整闭环架构。
- Academic Homepage (Main): https://169884902hzl.github.io/hzlink-academicpages/
- Scholar Citations & BibTeX Profile: https://169884902hzl.github.io/hzlink-academicpages/scholar/ · ORCID:
0009-0006-9392-1125 - Curriculum Vitae & Full Publications: CV · Publications · Projects
| Pillar | Research Thrust | Core Technical Mechanism | Representative Work |
|---|---|---|---|
01 · PERCEPTION |
Active Perception & Visual Servoing | Dual-arm viewpoint planning & target–gripper relative image error feedback under occlusion | Active Observation (IEEE CASE 2026) |
02 · VERIFICATION |
Pre-Execution VLM Verification | Iterative image-overlay action critique, mask projection & depth-informed 3D geometry | VR-Loop (Under Review) |
03 · ADAPTATION |
Visuotactile Policy Learning | Cross-modal RGB + tactile force field fusion with RL policy adaptation on Piper hardware | T-Rex / Sticks (Ongoing) |
04 · RECOVERY |
Failure Recovery & Memory | Online value assessment, reflexive safe retraction & episodic experience memory retrieval | HELIX (Ongoing) |
|
Zhilun Hu, Yuqi Cheng, Yuxiang Tan, Wen-pan Li, Weiming Shen. Coordinating an active eye-in-hand RGB-D camera arm with a grasping arm via a two-stage viewpoint strategy, target–gripper relative visual servoing, and occlusion-triggered viewpoint exploration. |
|
Yuqi Cheng, Zhilun Hu, Cheng Jiang, Qingyu Peng, Weiming Shen. A closed-loop Verify-and-Refine Vision-Language Agent that renders proposed dual-arm actions onto image overlays, critiques crossing topology and mask constraints prior to execution, and synthesizes depth-informed 3D trajectories for rope untangling. |
Visuotactile Learning · Ongoing Research Coupling visual observations and tactile shear/normal force fields with reinforcement learning for contact-rich insertion tasks on a Piper robot platform. |
Failure Recovery · Ongoing Simulation Research Integrating online value/risk assessment, fast reflexive retraction feedback, and episodic experience memory retrieval to recover from manipulation failures. |