Qingtao Liu
My work sits at the intersection of dexterous manipulation and multimodal representation learning. I enjoy building systems where tactile and visual signals come together to make robots act with intuition. Long-term goal: teach robots to learn from humans and serve people as collaborative partners.
About
I am a 5th-year PhD student at the College of Control Science and Engineering, Zhejiang University, advised by Prof. Qi Ye and Prof. Jiming Chen. I develop generalizable hand–object representations (DexRepNet++/T-RO 2025, IJRR 2025), visual–tactile datasets and benchmarks (VTDexManip/ICLR 2025), and multi-task manipulation policies (ICRA/IROS/CoRL) that close the gap between robot and human dexterity.
- Representation: DexRepNet++ learns spatial hand–object geometry for dexterous grasp/manipulation.
- Data & benchmarks: VTDexManip provides large-scale visual–tactile pretraining for dexterous hands.
- Policies: Multi-task RL and visual–tactile pretraining to enable human-like skills across objects and tasks.
News
- 2025.11 · DexRepNet++ accepted to IEEE T-RO.
- 2025.10 · Co-organized Dexterous Grasp Motion Challenge @ ICCV 2025 HANDS Workshop.
- 2025.07 · Contact2Motion accepted to IJRR (student second author).
- 2025.06 · Two papers accepted to IROS 2025.
- 2025.01 · VTDexManip accepted to ICLR 2025.
Publications
Recent and representative papers.
Journals
Conferences
Projects
- NSFC Center: Autonomous Intelligent Unmanned Systems · Dexterous hand skill learning for multi-task generalization in unstructured environments.
- Central Univ. Fund: Humanoid tactile skin sensor · Led tactile sensor benchmarking (since 2024).
- ZJU–Borui Joint Lab · SDF-based arm-hand online obstacle avoidance and replanning (2023).
- Dexterous Grasp Motion Challenge @ ICCV 2025 · Organizer and baseline code contributor.