Research Statement
The insights obtained from the research at the intersection of machine learning and robotics will have a profound impact on the everyday life of all of us in the coming years and decades. By employing the cutting-edge machine learning methods, the robots and machines surrounding us will become more and more intelligent and autonomous. Besides household appliances and automobiles, this development will also affect the work at factories, where an increasing amount of labor will be transferred from workers to robots and machines. Especially in this area, the collaboration of humans with robots and the knowledge transfer from humans to robots are important challenges addressed by the ongoing and upcoming research. During my PhD studies, I want to focus my research on the pipeline of the transfer of skills from humans to robots, that encorporates the following questions:

  • How to transfer movements from a human to the robot (tele-operation, kinesthetic teach-in)
  • How to imitate the behaviour of a human (imitation learning)
  • How to improve the skills taught by a human (reinforcement learning)

Current Position
Ph.D. student at the CLAS lab at TU Darmstadt
Topic: Machine Learning for Semi-Autonomous Manipulation of Objects
Supervisor: Prof. Dr. Gerhard Neumann
FB 20 : Informatik

Educational Background
2015-nowPh.D. Student
 supervised by Prof. Dr. Gerhard Neumann
 CLAS lab, TU Darmstadt
2014Master Thesis
 Thesis Title: Embedding Kalman Filters into Reproducing Kernel Hilbert Spaces
 supervised by Prof. Dr. Gerhard Neumann and Prof. Dr. Jan Peters
 IAS lab, TU Darmstadt
2011-2014Master Student
 Specialized Masters program 'Autonomous Systems'
 TU Darmstadt, Germany
2011Bachelor Thesis
 Thesis Title: Learning Symbols for Hierarchical Control from Interaction: Controllability Control
 supervised by Prof. Dr. Marc Toussaint and Dr. Tobias Lang
2008-2011Bachelor Student
 Freie Universität Berlin, Germany
Internships and work positions
2012-2014Student Assistant at Bionic Robotics GmbH
 Participation in the development of the control software for a compliant
2011Intern at Carmeq GmbH
 Development of a tool for automatic generation of test cases for a hardware-in-the-loop-system
2010-2011Student Assistant at Autonomous Project
 Participation in the software development for an autonomous car
2013-2014Teaching Assistant "Grundlagen der Robotic" (Foundations of Robotics)
2014Teaching Assistant "Machine Learning - Statistical Approaches 1"
2013Participant at RoboCup and RoboCup German Open
 Member of the Team Hector Darmstadt
References, etc.
Gebhardt, G.H.W. (2014). Embedding Kalman Filters into Reproducing Kernel Hilbert Spaces, Master Thesis.   Download Article [PDF]   BibTeX Reference [BibTex]
Gebhardt, G. (2011). Learning Symbols for Hierarchical Control from Interaction: Controllability Control, Bachelor Thesis. Download Article


zum Seitenanfang