Skip to main content


Period: 2015-2021
PI: Danica Kragic Jensfelt

In this project, we will put focus on the modeling of dynamical systems that have the purpose of interacting with the environment, systems that based on their perceptual capabilities (sensors), control their body and primarily their arms and hands. Compared to humans or primates, the sensing and dexterity of today’s robotic hands and arms are extremely limited. We can also draw a parallel with prosthetic hands that commonly have a single degree of freedom allowing them to interact with only a limited set of objects. Replicating the effectiveness and flexibility of human hands requires a fundamental rethinking of how to exploit the available mechanical dexterity and potentially suggests new, more efficient designs. To achieve this, we need to integrate several sensory modalities to estimate the state of the environment, to understand the requirements of a specific task, and to provide reasoning based on the available kinematic and dynamic properties for planning and acting in uncertain, dynamic environments. The theoretical work needed to address these problems is interdisciplinary and ranges from mathematical and statistical modeling to signal processing and data mining. More generally, we propose that moduli spaces and Geometric Invariant Theory (GIT), which has been studied in mathematics, can provide a framework for understanding equivalence classes and their deformations. While in robotics, one commonly concentrates on formulating problems in terms of vector spaces, moduli spaces and GIT can take the form of manifolds and/or algebraic varieties. Using this, the goal is thus to reduce the complexity of the problems and employ learning techniques that are otherwise not applicable when classical representations for perception and action are used.