Tobias Löw
Postdoctoral Scholar in the Personal Robotics Lab University of Washington
I work on geometric representations for robot manipulation using the intrinsic structure of a task to make robot control and optimization efficient, safe, and generalizable.
I am advised by Prof. Siddhartha Srinivasa. Before moving to Seattle I completed my PhD with EPFL in 2025. I studied mechanical engineering at ETH Zürich (BSc 2018, MSc 2020), and spent two years at CSIRO in Brisbane.
News
- Talk and workshop at IROS 2026
- Talking Robotics seminar on cooperative task spaces
- Workshop at ICRA 2026: Geometry in the Age of Data-Driven Robotics
Research Statement
Robots are moving out of structured industrial cells and into unstructured, human-centric spaces. The hard question there is no longer speed or precision, but how a robot should represent the things it perceives and interacts with.
I argue that these are largely problems of geometry. Whether a robot is grasping an object, avoiding an obstacle, or interpreting a depth image, its behavior is dictated by spatial relationships and a representation that respects those relationships makes the resulting computation simpler rather than harder. My work builds on conformal geometric algebra, which expresses geometric primitives (points, lines, planes, spheres) and the rigid-body transformations acting on them within a single algebraic structure.
This yields task formulations that carry over unchanged from single- to dual- to multi-arm systems, controllers in which objectives are written uniformly across primitives instead of case by case, and geometrically coherent approaches to problems from optimal control to tactile surface coverage. The framework is implemented in gafro, an open-source C++ library for robotics.
For more on this line of work, see geometric-algebra.tobiloew.ch.