MPML

Allow me to introduce myself. My name is Philippe Komma - but please call me Philippe if you like to contact me. Born and grown up in the beautiful city of Esslingen, I started my academic career at the University of Tübingen. For my studies, I chose Bioinformatics which - in short - is a complete studies of computer science with the minor subject of biology. Soon after entering university, I was able to devote my passion for computer graphics as research assistent at the department of graphical interactive systems (GRIS) which was headed by Wolfgang Straßer - the inventor of the Z-buffer - at that time. A couple of years later with my Diploma thesis in my pocket I had the opportunity to do my PhD at the department of cognitive systems at the University of Tübingen. I gladly accepted the job offer since cognitive science was the core area of my minor subject biology. The main topic of my PhD was the classification and clustering of varying terrain types a mobile robot is navigating on. During this time, I experimented with and enhanced fancy algorithms fulfilling the terrain identification task both in Matlab and C++. Furthermore, by means of my collaboration in varying third party funded projects, I came into contact with various car manufacturers and suppliers such as the Daimler AG and Borg Warner BERU Systems. Retrospectively, the experiences I gained from these projects have aroused my interest for the automotive domain and provided the motivation to start my job at the BMW Group in Munich after finishing my PhD.
And well, this is exactly the place where I stranded - a Swabian in the middle of the Bavarian heart. I guess, there are more inferior places to end up - all the more so since my current position as software developer offers a great versatility including

• software implementations of change requests by means of hand code or model-based develepment with ASCET and Mathwork's Real-Time Workshop
• tool- and HiL-based software testing
• tool development for software analysis such as automatic test case generation
• a tight collaboration in the field of function development
• the monitoring of software deliveries for suppliers which requires the coordination of various change request implementations along with their integration into a complete software package

Looking back at the things I've done, I don't need any crystall ball to predict that my future will be as exciting as my former days. And I can't wait to see what will happen next...

• 10/1999-06/2006: Studies of Computer Science at the University of Tübingen
• 11/2000-07/2005: Research assistant at the Department of Graphical-Interactive-Systems at the Wilhelm Schickard Institut of Computer Science in Tübingen
• 07/2005-12/2005: Diploma thesis at the Telecommunications Software and Multimedia Laboratory, Helsinki University of Technology
• 10/2006-04/2012: Research assistant at the Department of Computer Architecture, University of Tübingen
• 01/2008-02-2008: Collaboration at the headquarters of BorgWarner BERU Systems in Auburn Hills, Michigan
• since 01/2012: Software developer at the BMW Group in Munich

During my employment as scientific researcher at the University of Tübingen I focused on the topic of autonomous terrain classification and clustering. In outdoor robotic this is an important domain of research since along with a growing number of applications in unstructured outdoor environments, mobile robots are faced with increased requirements with respect to their driving safety and operational demands. Issues arise due to varying terrain types, each possessing different navigability characteristics such as the degree of hazardousness. Hence, a safe robot traversal necessitates the classification of the present ground surface from sensor data. In this thesis, a retrospective approach based on tactile sensing has been considered for the terrain discrimination task. That is, the ground surface classification relies on sensor data being collected while the robot traverses the environment. Here, the sensor data is represented by means of preprocessed acceleration patterns which emanate from terrain-wheel-chassis interactions and directly relate to the mechanical properties of the ground surfaces.
The main contribution of this thesis comprises the integration of contextual information into the terrain classification procedure. Here, contextual information denotes the temporal dependencies between consecutive measurements which are likely to arise from the same terrain type. It is demonstrated that the modeling of these dependencies in a principled way results in a significant improvement of the classification performance.
Note that the terrain-dependent adaption of the driving style according to extracted ground surface characteristics is not only a play of mind. Recent developments in the automotive industry of (e.g. Land Rover or Ford) have shown that an adaption of driving parameters improves the driving safety and behavior significantly.

• Philippe Komma and Andreas Zell. Markov random field-based clustering of vibration data. In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2010), pages 1902-1908, Taipei, Taiwan, October 2010. Best Paper Award Nominee.

• Philippe Komma and Andreas Zell. Clustering vibration data using a temporally coherent Expectation Maximization approach. In 7th Symposium on Intelligent Autonomous Vehicles ( IAV 2010), pages 1-6, Lecce, Italy, September 2010.

• Philippe Komma and Andreas Zell. Posterior probability estimation techniques embedded in a bayes filter for vibration-based terrain classification. Springer Tracts in Advanced Robotics, 62:79-89, 2010.

• Philippe Komma and Andreas Zell. Posterior probability estimation techniques embedded in a Bayes filter for vibration-based terrain classification. In 7th International Conference on Field and Service Robots (FSR 2009), pages 1-10, MIT, Cambridge, Massachusetts, USA, July 2009.

• Philippe Komma, Christian Weiss, and Andreas Zell. Adaptive Bayesian filtering for vibration-based terrain classification. In IEEE/RSJ International Conference on Robotics and Automation (ICRA 2009), pages 3307-3313, Kobe, Japan, May 2009.

• Philippe Komma and Andreas Zell. Towards real-time and memory efficient predictions of valve states in diesel engines. In IEEE Workshop on Computational Intelligence in Vehicles and Vehicular Systems (CIVVS 2009), pages 8-15, Nashville, TN, USA, April 2009.

• Philippe Komma, Christian Weiss, and Andreas Zell. Improved vibration based terrain classification using temporal coherence. In 40th International Symposium on Robotics (ISR), pages 359-364, Barcelona, Spain, March 2009.

• Philippe Komma, Jan Fischer, Frank Duffner, and Dirk Bartz. Lossless volume data compression schemes. In Simulation and Visualization 2007 (SimVis 2007), pages 169-182, Magdeburg, Germany, March 2007.