Invariant Representations of Space Curves for Tactile Gesture Recognition

Type of thesis: 



Muxfeldt et al.: Exploring tactile surface sensors as a gesture input device for intuitive robot programming, IEEE ETFA 2016

Starting Data: 


Abstract/Topic : 

Tactile gestures can successfully be used to control and program industrial robots. To this end, the links of a robot are equipped with tactile surface sensors (TSSs), which are utilized as both collision detection and tactile input devices. By executing a sequence of single- and multi-touch gestures on these TSSs, even novice users can intuitively create complex robot programs to solve typical tasks occurring in industrial production environments.
A key component of the gesture recognition approach is an invariant representation of the gesture input which is based on the differential geometry of space curves that are generated from the measured contact forces. The goal of this master’s thesis is to devise and implement new approaches to accelerate the generation of space curves from the input data and to extend and generalize the existing invariant representation. Moreover, a thorough performance evaluation based on an existing data set shall be performed.

Required Qualification: 

Willingness to study differential geometry; working knowledge of analysis and linear algebra; programming skills in Matlab, C++, or Python