Dr. Johannes Kulick

My research interests touch especially the area of active learning in the field of robotics. So I try to apply machine learning techniques to the problem of autonomous robotics. This includes especially -- but is not limited to -- Bayesian techniques. In our curious robot project we experiment on autonomous exploration on our PR2 robot and use our active learning insights directly on real world problems. We are especially interested on degrees of freedom in the world and how robot could explore them and their structure to achieve task goals.

Earlier I was involved in the RoboCup Soccer team Fumanoids, where I wrote my master's thesis about humanoid gait generation.

I support the 'Erklärung der Initative für Hochschulen für den Frieden' (declaration of the initiative for universities for peace) against research for military purpose.

Publications

2017

• M. Baum, M. Bernstein, R. Martín-Martín, S. Höfer, J. Kulick, M. Toussaint, A. Kacelnik, and O. Brock: Opening a lockbox through physical exploration, In Proc. of Int. Conf. on Humanoids Robots, 2017, pdf

2016

• J. Kulick: Information Driven Exploration in Robotics, Ph.D. Thesis, University Stuttgart, 2016, pdf
• J. Kulick, R. Lieck, M. Toussaint: Cross-Entropy as a Criterion for Robust Interactive Learning of Latent Properties., In Proc. of Future of Interactive Learning Machines Workshop at NIPS, 2016, pdf

2015

• J. Kulick, R. Lieck, M. Toussaint: The Advantage of Cross-Entropy over Entropy in Iterative Information Gathering, arXiv e-pub: 1409.7552v2, 2015, pdf
• J. Kulick, S. Otte, M. Toussaint: Robots Solving Serial Means-Means-End Problems, In Proc. of Combining AI Reasoning and Cognitive Science with Robotics Workshop at Robotics: Science and Systems, 2015, pdf
• J. Kulick, S. Otte, M. Toussaint: Active Exploration of Joint Dependency Structures, In Proc. of the Int. Conf. on Robotics and Automation (ICRA), 2015, pdf, interactive poster

2014

• J. Kulick, R. Lieck, M. Toussaint: Active Learning of Hyperparameters: An Expected Cross Entropy Criterion for Active Model Selection, arXiv e-pub: 1409.7552v1, 2014, pdf
• S. Otte, J. Kulick, M. Toussaint, O. Brock: Entropy Based Strategies for Physical Exploration  of the Environment's Degrees of Freedom, In Proc. of the Int. Conf. on Intelligent Robots and Systems (IROS), 2014, pdf

2013

• J. Kulick, M. Toussaint, T. Lang, M. Lopes: Active Learning for Teaching a Robot Grounded Relational Symbols, In Proc. of the Int. Joint Conf. of Artificial Intelligence (IJCAI), 2013, pdf

2012

• D. Seifert, S. Otte, J. Kulick, N. v. Schmude, L. Dohrmann, S. Heinrich, H. Moballegh, S. Mielke, L. Freitag, S. Hohberg, J. Auer, M. Losch, R. Rojas: Berlin United – FUmanoids Team Description Paper 2012, RoboCup 2012, pdf

2011

• J. Kulick: Ein stabiler Gang für humanoide, Fußball spielende Roboter, Masters thesis, Freie Universität Berlin, 2011, German, pdf
• D. Seifert, H. Moballegh, S. Heinrich, S. Oe, S. Mielke, N. von Schmude, T. Weißgerber, K. Würfel, R. Jonschkowski, J.Kulick, M. Fröhlich, J. Streckenbach, G. Lubitz, N. von Geyso, M. Schubert, Fu Yao, S. Puhlmann, O. Morillo, L. Freitag, R.Rojas: FUmanoids Team Description Paper 2011, RoboCup 2011, pdf

2009

• J. Kulick, M. Block, R. Rojas: General Game Playing mit Stochastischen Spielen, Technical Report, Freie Universität Berlin, 2009, German, pdf
• J. Kulick: World Description Language - A logical Language for agent-based Systems and Games, Bachelor thesis, 2009, Freie Universität Berlin, pdf

During my research I implemented a few algorithms and put them to libraries. In my spare time I also program on some smaller projects. You can find them here. For all the even smaller stuff check my GitHub account.

Mont Carlo Tree Search (MCTS) is a state of the art planning algorithm useful in many domains. I implemented various versions of it as a SciKit (basically a scientific python library).

Picture of Monte Carlo: Hampus Cullin, CC BY-SA 3.0

A simple, yet useful music player for musicians. It can set A-B-repeat very precisely. I'm also working on integrating some signal processing to time stretch songs. All this is for easily playing along your favorite music, even if your skills don't match the artists ones yet.

I forked scholar.py, a tool to use Google Scholar from the command line, and added some functionality (direct BibTex, EndNote... download etc.)

I moved the signal processing stuff of playalong to an own lib, because it a) might be useful for others and b) I have some other ideas for using it in other little projects. So far it only consists of a STFT and a phase vocoder.

mailless.org is the joint mail server of some friends of mine and myself. This is basically an 'admin' project. I don't really write software -- apart from scripts -- for it.

In 2008, during my Bachelors, I was a participant of the Google Summer of Code. I wrote some drag'n'drop widgets for Audacity.

Email:

You may consider contacting me encrypted. my gpg key.