Haoyu Wei

Haoyu Wei | 卫浩宇

I am a first-year M.S. student in the Department of Automation at Tsinghua University, advised by Prof. Jiwen Lu. In 2025, I obtained my B.Eng. in the Department of Automation from Beihang University.

I work on robotics. My current research focuses on:

Efficient Manipulation that optimizes asynchronous inference and improves imitation learning paradigms.
Visual Navigation that enables robots to autonomously explore environments according to visual observations and instructions.
Dexterous Manipulation that builds tactile-aware general manipulation models with dexterous hands.

*Equal contribution, ^†Project leader.

Publications

F2F-AP: Flow-to-Future Asynchronous Policy for Real-time Dynamic Manipulation
Haoyu Wei, Xiuwei Xu†, Ziyang Cheng, Hang Yin, Angyuan Ma, Bingyao Yu, Jie Zhou, Jiwen Lu
arXiv, 2026
[arXiv] [Project Page]

To address observation lag in asynchronous inference, we use optical-flow prediction and contrastive learning to predict future observations in the feature space, align actions with their true execution timeline, and improve success rates on highly dynamic tasks.

CMP: Robust Whole-Body Tracking for Loco-Manipulation via Competence Manifold Projection
Ziyang Cheng, Haoyu Wei, Hang Yin, Xiuwei Xu, Bingyao Yu, Jie Zhou, Jiwen Lu
arXiv, 2026
[arXiv] [Project Page]

Competence Manifold Projection (CMP) aims to solve the problem that learning holistic whole-body control policies for tracking global end-effector poses remains fragile against Out-of-Distribution (OOD) inputs induced by sensor noise or infeasible user commands.

GC-VLN: Instruction as Graph Constraints for Training-free Vision-and-Language Navigation
Hang Yin*, Haoyu Wei*, Xiuwei Xu, Wenxuan Guo, Jie Zhou, Jiwen Lu
arXiv, 2025
[arXiv] [Project Page]

We propose a general framework for mobile manipulation, which can be divided into docking point selection and fixed-base manipulation. We model the docking point selection stage as an optimization process, to let the agent move and touch target keypoint under several constraints.