My name is Yiming Li. I am currently a second year PhD student in the Robot Learning & Interaction group at the Idiap Research Institute and École Polytechnique Fédérale de Lausanne (EPFL), supervised by Dr. Sylvain Calinon.
I am broadly interested in planning, manipulation, and learning for robotic systems that interact with their surroundings and humans. Currently, my focus is on differentiable representations in manipulation planning tasks.
Why differentiable representations? Gradients play a crucial role in AI’s training and optimization processes. Gradient-based optimization techniques are widely used in learning, planning, and control. However, these components are usually computed separately (for instance, learning and planning typically focus on the task space, while control commands are in the joint space). On the other hand, end-to-end approaches directly output control commands but function as a black box. Differentiable robot representation bridges these approaches, allowing for gradient-based optimization from high-level scene understanding to low-level control in a unified framework. Differential geometries and equations are also appealing in solving robot manipulation problems. They are well-studied in mathematics but usually hard to scale to high-dimensional robot systems. Neural PDE/ODE solvers seem promising.
During my master’s, I worked on learning-based approaches for robotic grasping in cluttered environments.
I welcome the opportunity to listen to others and exchange ideas. Please feel free to drop me an email if you want to discuss anything with me!
🔥 News
-
2024.07: 🎉🎉 Our paper Configuration Space Distance Fields for Manipulation Planning has been nominated as a Best Paper Finalist at RSS 2024! We’re thrilled to receive this recognition!
-
2024.05: 🎉🎉 Our paper about configuration space distance fields is accepted to RSS 2024!
-
2024.04: 🎉🎉 Our paper about online learning signed distance fields using piecewise polynomials is accepted to RA-L!
-
2024.01: 🎉🎉 Our paper about representing robot geometry as distance fields is accepted to ICRA 2024!
-
2022.10: I start my PhD in the Robot Learning & Interaction group!
-
2022.08: Our paper about human-to-robot handovers is accepted to IEEE Transactions on Cognitive and Developmental Systems (T-CDS)!
-
2022.08: Our paper about variational grasp generation for dextrous manipulation is accepted to RA-L!
-
2022.01: Our paper about anthropomorphic hand grasping in clutter is accepted to ICRA 2022!
-
2021.07: Our paper about 6-DoF grasp pose estimation is accepted to IROS 2022!
📝 Selected Publications
Please visit my Google Scholar page for full publications.
HGC-Net: Deep Anthropomorphic Hand Grasping in Clutter
Yiming Li, Wei Wei, Daheng Li, Peng Wang, Wanyi Li, Jun Zhong
In Proc. IEEE Intl Conf. on Robotics and Automation 2022 (ICRA 2022).
Learning Human-to-robot Dexterous Handovers for Anthropomorphic Hand
Hoanan Duan, Peng Wang, Yiming Li, Daheng Li, Wei Wei
IEEE Transactions on Cognitive and Developmental Systems(T-CDS) 15 (3), 1224-1238.
DVGG: Deep Variational Grasp Generation for Dextrous Manipulation
Wei Wei, Daheng Li, Peng Wang, Yiming Li, Wanyi Li,Yongkang Luo, Jun Zhong
IEEE Robotics and Automation Letters (RA-L) 7 (2), 1659-1666.
Simultaneous Semantic and Collision Learning for 6-dof Grasp Pose Estimation
Yiming Li, Tao Kong, Ruihang Chu, Yifeng Li, Peng Wang, Lei Li
In Proc. IEEE/RSJ Intl Conf. on Intelligent Robots and Systems 2022 (IROS 2022).
📖 Educations
- 2022.10 - now, École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland.
- 2019.06 - 2022.06, Institute of Automation, Chinese Academy of Sciences, Beijing, China.
- 2015.09 - 2019.06, Tongji University, Shanghai, China.
📞 Contact
- Email: ymli.cn@gmail.com
- Telephone: (+41) 77 278 38 48
- Address: Idiap Research Institute, Centre du Parc, Rue Marconi 19, CH-1920 Martigny, Switzerland