Prof. Dr. rer. nat. Felix v. Hundelshausen

my homepage

Prof. Dr. rer. nat. Felix von Hundelshausen 

 

felix

Universität der Bundeswehr München
Fakultät für Luft und Raumfahrttechnik
Institut für Technik autonomer Systeme
Werner-Heisenberg-Weg 39
85579 Neubiberg, Germany
e-mail: email1.jpgemail2.jpg
Tel : +49-(0)89-6004-4145
Fax: +49-(0)89-6004-3074

 

My research interests cover artificial intelligence, robotics, computer vision, and perception in general. From a practical point of view my dream is to build intelligent robots.  From a theoretical point of view my dream is to understand how the human brain works.

The robots I am currently working with are the autonomous vehicle MuCAR-3 at our institute and the humanoid robot LOLA of the applied mechanics group of Prof. Ulbrich at TUM. While the applied mechanics group of Prof. Ulbrich is interested in the science of and around walking, our group is interested in the perception system and we have succesfully forged our alogrithms together into a single working system.

I am responsible for the development of Lola's computer vision system and our research is funded by the DFG-project "Towards a general vision system for humanoid robots".

We are currently demonstrating Lola's cutting edge mechanical construction, walking and perceptual capabilities at Hannover Messe 2010. Our achievement is a new degree of flexibility: Lola is able to walk around in an environment that hasn't been defined and modelled before. Lola can avoid any obstacles and we demonstrate this, allowing the spectators to place any objects as obstacles on the ground.

Here are some links to press coverage and videos that demonstrate these capabilities.

Although I am fascinated by the robots themselves, the scientific justification for them is that they force new ideas to stay on the ground and to have practical implications.

When thinking about why machines aren't intellegent yet and why they have so big difficulties in perceiving and acting flexibly in the world, I believe that there is one key-problem that hasn't been solved sufficiently well up to now. Interestingly, this problem comes in many facettes and while most people find this problem being located in the field of pure perception it is of outmost importance for other fields like planning and machine learning, too. This problem is the "General Correspondence Problem".

The general correspondence problem is the problem of determining that two things correspond to another, although they might appear differently in their sensory form, e.g. because they underly variations, either of the objects themselves (different chairs are still all part of the category "chair"), contextural variations (e.g differnt backgrund clutter, occlusions etc.. ) mapping variations (e.g. differnt lighting conditions in vision, e.g acoustic conditions in speech recognition), receptive variations (different states of the receptors e.g due to adaption, etc.. ). Humans are masters in solving the correspondence problem:

We know that an object like a chair that is seen from different perspective and casts different images on the retina is still the same object. In the field of cognitive psychology this phenomenon is called object constancy. In computer vision, it means (simplifying slightly) that the computer can establish interframe correspondences. Apart from temporal correspondence we can recognize objects, that is we can establish correspondence between image structures and internal categories and/or models. But the correspondence problem is not restricted to vision. The same problem occurs in speech recognition, or tactile recognition of objects.  It also occurs when trying to understand a written story. Here, correspondences between an internal mental plot and the text have to be established.

In short, I belief that solving the coorespondence problem is the key to artificial intelligence.

What makes the correspondence problem so hard to solve, is that it cannot be solveed without solving a second problem which is the "segmentation problem". The segmentation problem means to identify the structures that are the objects of corresponence. For instance, when a human's speech is to be understood in the presence of background noise, the difficulty is to identify the accousitc parts that are actually of the person speaking and not of someone different, talking in the background. What makes things diffifcult is that the correspondence problem and the segmentation problem have to be solved simultaneously.

Since about  four years I am working on a new theory for solving the general correspondence problem in a general way, independent of its particular domain and I am

working hard to publish the theory within the next few years.

(some links)

Short vitae

From 2001 to 2004 I was PhD-student in the group of Prof. Raúl Rojas at the Free University of Berlin. During this time, I developed a computer vision system for the RoboCup Middle Size league. Our team placed 2nd at the world championships in 2005 using this system. From 2004 to 2005 I was Postdoc in the CORAL group of Prof. Manuela Veloso at Carnegie Mellon University (CMU), Pittsburgh, USA. Since 2006 I am in the group of Prof. H. J. Wünsche, the successor of Prof. E. D. Dickmanns (emeritus), carrying on research on perception for autonomous vehicles. Since 2008, I am responsible for the development of the computer vision system for the humanoid robot Lola, collaborating with the Applied Mechanics group of Prof. Heinz. Ulbrich at TU-Munich. This collaboration takes place in form of the DFG-project "Towards a general vision system for humanoid robots".

Research Community

· I was chairing one of the Robotic Vision sessions at ICJAI07 (International Joint Conference on Artificial intelligence)

· Member of the program comitteee of the International RoboCup Symposium 2005

· Member of the program comitteee of the program comittee of the International Conference on Informatics and Control, Automation and Robotics (ICINCO) 2005

· Member of the program committee for the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 2005

· Service as reviewer for the journal "Soft Computing and Automation"

· Service as reviewer for the journal "Image and Vision Computing"

· Service as reviewer for the journal "Computer-Aided Design"

Publications

Driving with tentacles: Integral structures for sensing and motion. F. von Hundelshausen, F., Himmelsbach, M., Hecker, F., Mueller, A., and Wuensche, H. 2008. J. Field Robot. 25, 9 (Sep. 2008), 640-673.

[pdf]

Team AnnieWAY's autonomous system for the 2007 DARPA Urban Challenge. Kammel, S., Ziegler, J., Pitzer, B., Werling, M., Gindele, T., Jagzent, D., Schröder, J., Thuy, M., Goebl, M., Hundelshausen, F. v., Pink, O., Frese, C., and Stiller, C. 2008.  J. Field Robot. 25, 9 (Sep. 2008), 615-639

[pdf]

Mesh-based Active Monte Carlo Recognition. F. von Hundelshausen, H.-J. Wuensche, Marco Block, Raul Kompass and Raúl Rojas, In Proceedings of the 20th International Joint Conference on Artificial Intelligence (IJCAI07), pp. 2231-2236, Hyderabad, India, 2007.

[pdf]

Active Monte Carlo Recognition. F. v. Hundelshausen and Veloso, M. In Proceedings of the 29th Annual German Conference on Artificial Intelligence. Springer Lecture Notes in Artificial Intelligence, pp. 229-243, 2006.

[pdf]

FOCUS: a generalized method for object discovery for robots that observe and interact with humans. M. Veloso, P. E. Rybski and F. von Hundelshausen, In Proceedings of the 2006 Conference on Human-Robot Interaction, pp. 102-109, Salt Lake City, UT, March, 2006.

[pdf]

Learning Visual Object Definitions by Observing Human Activities. M. Veloso, F. von Hundelshausen, P. E. Rybski, In Proceedings of Humanoids 2005, Japan, December, 2005.

[pdf]

Robot Localization through Region Tracking. F. v. Hundelshausen, In it-Technology, 47(5), pp. 258-265, Oldenburg Wissenschaftsverlag, May, 2005.

[pdf]

Computer Vision for Autonomous Mobile Robots. F. v. Hundelshausen. Dissertation, Department of Computer Science, Free University of Berlin , September, 2004.

[pdf][digital library][movie]

A constructive feature-detection approach for mobile robotics. In  RoboCup 2004: Robot Soccer World Cup VIII. Lecture Notes in Computer Science 3276, Springer 2005, pp. 72-83, Lisboa, July 5-7, 2004.

[pdf]

Tracking Regions. In RoboCup 2003: Robot Soccer World Cup VII, Lecture Notes in Computer Science 3020, Springer 2004 , Padova, July 10-11, 2003.

[pdf]

Matrix: A force field pattern matching method for mobile robots. Technical Report B-09-03, Freie Universität Berlin, June 2003.

[pdf]

Localizing a robot by the marking lines on a soccer field. In Ondřej Drbohlav,editor, Proceedings of the Computer Vision Winter Workshop, CVWW03, pp. 135-140, February 2003.

[pdf]

Tracking Regions and Edges by Shrinking and Growing. In Ondřej Drbohlav, editor, Proceedings of the Computer Vision Winter Workshop, CVWW03, pp. 33-38, February 2003.

[pdf]

FU-Fighters Omni 2001 (Local Vision) In RoboCup-01: Robot Soccer World Cup VLecture Notes in Computer Science 2377, Springer 2002.

[pdf]

A general algorithm for finding transitions along lines in colored images. Proceedings of the Computer Vision Winter Workshop, CVWW02, February 2002.

[pdf]

An omnidirectional vision system for soccer robots. master thesis (Diplom), Department of Computer Science, Free University of Berlin, April 2001

[pdf]

« November 2017 »
November
MoDiMiDoFrSaSo
12345
6789101112
13141516171819
20212223242526
27282930
Anmelden


Passwort vergessen?
mucar

"MuCAR-3" (Munich Autonomous Robot Car, 3rd generation), our autonomous robot car.

berlin

Berlin, Germany

pittsburgh

Pittsburgh, USA

cmu

Hammerschlag Hall, Carnegie Mellon University

munich

Munich, Germany

mucar

"MuCAR-3" (Munich Autonomous Robot Car, 3rd generation), our autonomous robot car.

berlin

Berlin, Germany

pittsburgh

Pittsburgh, USA

cmu

Hammerschlag Hall, Carnegie Mellon University

munich

Munich, Germany

mucar

"MuCAR-3" (Munich Autonomous Robot Car, 3rd generation), our autonomous robot car.

berlin

Berlin, Germany

pittsburgh

Pittsburgh, USA

cmu

Hammerschlag Hall, Carnegie Mellon University

munich

Munich, Germany