Dylan Turpin
Papers
4
Total Citations
115
H-Index
4
About
Dylan Turpin is a roboticist pushing the boundaries of dexterous manipulation and autonomous assembly. His research centers on differentiable simulation and self-supervised learning to solve two of robotics' hardest challenges: generating robust multi-finger grasps and assembling unknown objects from their parts. Turpin's landmark work, *Grasp’D* (2022, 53 citations), introduced a differentiable contact-rich grasp synthesis framework that enables gradient-based optimization for multi-fingered hands, dramatically improving grasp quality over traditional methods. He further advanced this with *Fast-Grasp'D* (2023, 24 citations), which accelerates dexterous grasp generation by making contact dynamics fully differentiable. In parallel, his *Neural Shape Mating* (2022, 34 citations) pioneered a self-supervised approach to object assembly, using adversarial shape priors to learn how geometric parts fit together without human labels—a critical step toward autonomous repair, manufacturing, and household robotics. By replacing heuristic search with differentiable optimization, Turpin's work has opened new avenues for robots to interact with the physical world with unprecedented precision and adaptability. His contributions directly impact the future of manufacturing, assistive robotics, and automated assembly.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4