Severin Fichtl

University of Aberdeen, University of Southern Denmark, Maersk (Denmark)

Papers

Total Citations

H-Index

About

Severin Fichtl is a researcher whose work lies at the intersection of developmental robotics, computer vision, and cognitive systems, with a central focus on how robots can learn to understand and manipulate objects in human-like environments. His primary research area is the learning of spatial relationships and relational affordances—the idea that the outcome of a manipulation action depends not just on a single object, but on the spatial relationships between pairs of objects (e.g., "inside," "on top," "behind"). Fichtl’s major contribution is the development of vision-based methods, such as relational histogram features, that allow robots to extract these meaningful spatial features from 3D visual data and use them to predict action outcomes. His most cited work, "Learning spatial relationships from 3D vision using histograms" (18 citations), lays the foundation for this approach. He has also explored how robots can bootstrap knowledge of object-pair affordances through transfer learning, mimicking the developmental progression seen in human infants. Though his citation counts are modest, Fichtl’s work is notable for its principled, developmental perspective on robot learning, addressing a fundamental challenge in autonomous manipulation: understanding the preconditions for means-end actions in dynamic, unstructured environments.