Florian T. Pokorny
University of California, Berkeley, KTH Royal Institute of Technology
Papers
32
Total Citations
1,091
H-Index
17
About
Florian T. Pokorny is a researcher whose work spans robot learning, grasp planning, topological methods in robotics, and learning from demonstrations. He is perhaps best known for his foundational contributions to the Dexterity Network (Dex-Net) project, with his co-authorship of "Dex-Net 1.0" (2016) accumulating over 370 citations and establishing cloud-based grasp planning as a powerful paradigm for robust manipulation under uncertainty. His research consistently addresses the practical challenges of deploying robots in unstructured environments, from grasping objects in cluttered warehouses to surgical task segmentation in robot-assisted minimally invasive surgery. A distinctive thread running through Pokorny's career is the application of algebraic topology — particularly persistent homology and simplicial complexes — to robotics problems such as trajectory classification and gripper design for objects with holes, work that has attracted a dedicated following in the computational robotics community. He has also made significant contributions to imitation and reinforcement learning, developing algorithms like SWIRL and SHIV that reduce the burden on human supervisors while enabling robots to efficiently learn complex, multi-stage tasks. With a portfolio exceeding 800 total citations across these diverse areas, Pokorny represents a rare combination of mathematical rigor and practical engineering insight that continues to influence both theoretical and applied robotics research.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4
- 5
- 6
- 7Grasping objects with holes: A topological approach43 citations · 2013
- 8Multi-armed bandit models for 2D grasp planning with uncertainty37 citations · 2015
- 9Multiscale Topological Trajectory Classification with Persistent Homology35 citations · 2014
- 10