I have been a PhD student at the IAS group at the TU Munich since 2007. Previously, I studied Electrical Engineering and Information Technology at the RWTH Aachen, Germany, and Robotics and Embedded Systems at the ENSTA in Paris, France. I spent one year as an intern at the Accenture Technology Labs in Sophia Antipolis, France, where I was working on visual localization and mapping. From February to May 2009, I was a visiting scholar in the Human Sensing Lab at Carnegie Mellon University where I was working with Fernando De la Torre on models of human everyday activities.

Current projects

Segmentation of Human Motions in Everyday Manipulation Tasks We are investigating methods for segmenting sequences of human motion into meaningful classes like "Reaching" or "Grasping". The input data is provided by the MeMoMan full-body pose tracker. Our approach is based on Conditional Random Fields (CRF) which perform the segmentation combining pose-related features (if hands are extending, beyond a threshold, etc) with information from a sensor network and the environment model (object picked up, cupboard opened).
Hierarchical Action Models In order to perform abstract reasoning over observed action sequences, their ordering and parameters, one has to abstract from sequences of motion segments into higher-level action classes. The segmentation and classification step produces a sequence of motion segments which are represented as instances of the respective motion classes in a knowledge base. They are combined with additional observations (which object was manipulated, where was it taken from,...), and abstract action specifications are matched against this sequence (stating e.g. that a transport action includes picking up an object, moving to another place and putting it down). For comparing actions and for learning the structure of activities we are using statistical relational learning methods, in particular Bayesian Logic Networks.
Importing Natural-Language Task Descriptions from the WWW The goal of this project is to transform task descriptions from websites like ehow.com, which are written in natural language, into formal representations. These formally represented instructions can be used in action recognition (by matching them against observed activities) as well as in robot planning (by generating an executable robot plan). The system uses state-of-the-art natural-language parsing techniques and resolves the meaning of words using the WordNet database and the Cyc ontology. Challenges include the ambiguities in natural language and incomplete action specifications. Our approach is to first import the task descriptions into a formal representation in the knowledge base, and to identify and add missing information using common-sense knowledge and facts learned from observations.
Creation of a Knowledge Base for Mobile Household Robots Mobile household robots performing advanced object manipulation tasks in a kitchen environment need a large amount of knowledge including spatial knowledge about the world, e.g. the positions of obstacles, items of furniture or objects to be manipulated, ontological knowledge about objects, their types, relations, and properties, common-sense knowledge, for instance that objects inside a cupboard are not visible from outside unless the door is open, knowledge about the functions of objects like the main task a tool serves for or the sequence of actions required to operate a dishwasher. This project aims at the integration of various sources of knowledge from public ontologies like Cyc, databases of common-sense like OpenMind Indoor Commonsense or the semantic environment maps developed at our chair.
Knowledge Representation, Visualization and Data Mining Tools These tools allows for the manual exploration of data before automating the process using the integrated data mining modules, custom features for different calculations, visualisations, links to data sources like MySQL databases and the interface to the PROLOG knowledge processing system. Several visualisations demonstrate the robot's knowledge about the world, for instance about objects, trajectories, human poses or observed positions.
Acquisition of Grounded Action Models for Mobile Manipulation Actions Most objects in a household environment are designed for a certain purpose, so it is only natural to describe them by their function instead of their appearance. We are learning action models from observation that describe mobile robot manipulation actions and their parameters like the objects manipulated, tools used, the place where the robot stood when performing the actions, the hand that was used et cetera. These models are completely integrated into our knowledge base and are represented as concepts in our knowledge base. By completely integrating the learned models with taxonomic and common-sense knowledge we can perform powerful analysis and prediction of the robot's actions.

Research Topics

• Ontological Representation and Reasoning

• Human Motions in Manipulation Tasks

• Human Motion Capture, Tracking and Analysis

Teaching General

Selected Publications

• Moritz Tenorth , Michael Beetz , KnowRob --- Knowledge Processing for Autonomous Personal Robots, 2009, IEEE/RSJ International Conference on Intelligent RObots and Systems. To appear.,

  (BibTeX) (PDF)

• Moritz Tenorth , Jan Bandouch , Michael Beetz , The {TUM} Kitchen Data Set of Everyday Manipulation Activities for Motion Tracking and Action Recognition, 2009, IEEE Int. Workshop on Tracking Humans for the Evaluation of their Motion in Image Sequences (THEMIS). In conjunction with ICCV2009,

  (BibTeX) (PDF)

• Moritz Tenorth , Daniel Nyga, Michael Beetz , Understanding and Executing Instructions for Everyday Manipulation Tasks from the World Wide Web, 2009,

  (BibTeX) (PDF)

• Freek Stulp , Andreas Fedrizzi , Franziska Zacharias, Moritz Tenorth , Jan Bandouch , Michael Beetz , Combining Analysis, Imitation, and Experience-based Learning to Acquire a Concept of Reachability, 2009, 9th IEEE-RAS International Conference on Humanoid Robots,

  (BibTeX) (PDF)

• Dejan Pangercic , Rok Tavcar, Moritz Tenorth , Michael Beetz , Visual Scene Detection and Interpretation using Encyclopedic Knowledge and Formal Description Logic, 2009, Proceedings of the International Conference on Advanced Robotics (ICAR).,

  (BibTeX) (PDF)

• Moritz Tenorth , Michael Beetz , Towards Practical and Grounded Knowledge Representation Systems for Autonomous Household Robots, 2008, Proceedings of the 1st International Workshop on Cognition for Technical Systems, M{\"u}nchen, Germany, 6-8 October,

  (BibTeX) (PDF)

• Michael Beetz , Freek Stulp , Bernd Radig , Jan Bandouch , Nico Blodow , Mihai Emanuel Dolha , Andreas Fedrizzi , Dominik Jain , Ulrich Klank , Ingo Kresse , Alexis Maldonado , Zoltan-Csaba Marton , Lorenz Mösenlechner , Federico Ruiz-Ugalde , Radu Bogdan Rusu , Moritz Tenorth , The Assistive Kitchen --- A Demonstration Scenario for Cognitive Technical Systems, 2008, IEEE 17th International Symposium on Robot and Human Interactive Communication (RO-MAN), Muenchen, Germany,

  (BibTeX) (PDF)

• Zoltan-Csaba Marton , Nico Blodow , Mihai Emanuel Dolha , Moritz Tenorth , Radu Bogdan Rusu , Michael Beetz , Autonomous Mapping of Kitchen Environments and Applications, 2008, Proceedings of the 1st International Workshop on Cognition for Technical Systems, Munich, Germany, 6-8 October,

  (BibTeX)