Personal tools
 

Intro

I have started my PhD at the Intelligent Autonomous Systems Group in 2008 under the supervisory of prof. Michael Beetz, PhD. My main research interest lies in robotic perception of 2D/3D data and interpretation of thereof. In particular I have a desire to decipher complex indoor scenes (e.g. arbitrary kitchen table sets) using various means  of suitable representations (e.g. first order logic) and/or a priori knowledge sources such as wikihow.com. 
 
I am an active member of IEEE and RAS societies and regularly submit and publish to following conferences: IROS, ICRA, ICAR, Humanoids, RSS. I actively develop and contribute code to Robot Operating System (ROS), open-source, meta-operating system for personal robots. Please find my packages in our public tum-ros repository.
 
I obtained my B.Sc. and M.Sc. degrees at the Faculty for Electrical Engineering, UL and  Department of Electrical Engineering and Information Technology, TUM respectively. Any additional information about me, including a complete CV, is available upon request.

Student Jobs

Students wanted for BSc/MSc/Internships(Praktika)!!!    Drop-in or send an email!

Research Threads

This section contains excerpts from my immediate and past work as reported at various robotics conferences. In the future I plan to enrich it with the pointers to videos and (as far as possible) code too.
 
General 3D Modelling of Novel Objects from a Single View
In this paper we present a method for building models for grasping from a single 3D snapshot of a scene composed of objects of daily use. We employ fast shape estimation, probabilistic model fitting and verification methods capable of dealing with different kinds of symmetries, and combine these with a triangular mesh of the parts that have no other representation to model previously unseen objects of arbitrary shape. Our approach is enhanced by the information given by geometric clues about different parts of objects which serve as prior information for the selection of the appropriate reconstruction method. While we designed our system for grasping based on single view 3D data, its generality allows us to also use the combination of multiple views and thus perform refining and building geometric models. This greatly simplifies the task for other systems that also require geometric models for grasping (for locating objects in camera images for example).
In submission IROS2010. [VIDEO], [pdf]		 CAD Modelling
Combining Perception and Knowledge Processing for Everyday Manipulation
This paper describes and discusses the K-CoPMan (Knowledge-enabled Cognitive Perception for Manipulation) system, which enables autonomous robots to generate symbolic representations of perceived objects and scenes and to infer answers to complex queries that require the combination of perception and knowledge processing. Using K-CoPMan, the robot can solve inference tasks such as identifying items that are likely to be missing on a breakfast table. To the programmer K-CoPMan, is presented as a logic programming system that can be queried just like a symbolic knowledge base. Internally, K-CoPMan is realized through a data structure framework together with a library of state-of-the-art perception mechanisms for mobile manipulation in human environments. Key features of K-CoPMan are that it can make a robot environment-aware and that it supports goal-directed as well as passive perceptual processing. K-CoPMan is fully integrated into an autonomous mobile manipulation robot and is realized within the open-source robot library ROS. In submission RSS2010.		 K-CopMan
Real-time CAD Model Matching for Mobile Manipulation and Grasping
Humanoid robotic assistants need capable and comprehensive perception systems that enable them to perform complex manipulation and grasping tasks. This requires the identification and recognition of supporting planes and objects in the world, together with their precise 6D poses. In this paper, we propose a 3D perception system architecture that can robustly fit CAD models in cluttered table setting scenes for the purpose of grasping with a mobile manipulator. Our approach uses a powerful combination of two different camera technologies, Time-Of-Flight (TOF) and RGB, to robustly segment the scene and extract object clusters. Using an a-priori database of object models we then perform a CAD matching in 2D camera images. We validate the proposed system in a number of experiments, and compare the system's performance and reliability with similar initiatives. Presented at Humanoids2009 [pdf]		 CAD Matching
Optimization of Simulated Production Process Performance using Machine Learning
This paper investigates integration of the supervised machine learning algorithms (Model Trees, Neural Networks) into a production plan realized in a physics-based realistic simulator. Proposed novelty is in that the learning capability is integrated into the control process which allows for online learning and on the fly control code modification. Running the process in a simulated environment enables hazardless experimenting with the system's setup and integral acquisition of data. Yielded optimization times obtained through learning outperform times of a production process solely based on averaging. Presented at ETFA2009 [pdf]		 System architecture
Visual Scene Detection and Interpretation using Encyclopedic Knowledge and Formal Description Logic
In this system paper we report on our experience while working with a top-down guided 3D CAD model-based vision algorithm, being executed by an autonomous robot on objects (tableware and cutlery) in an Assistive Household environment. Top-down guidance is shaped upon how-to instructions which are parsed and extracted from the wikihow.com webpage - one of the world's largest resource of natural language task descriptions. Therein we selected a \emph{How to set a table} entry and thus constructed this paper upon conversely interpreting the table setting for a meal. The robot's knowledge base is represented in Description Logics (DL) using the Web Ontology Language, and the inferences are obtained by virtue of SWI-Prolog queries. The whole proposed system is controlled by a modern, leading-edge Reactive Plan Language (RPL) which is the basic planning feature in the Assistive Household. Presented at ICAR2009 [pdf]		 Improved search

 

Current Research Projects

  • CoP : The project aims at the unification of vision-based sensing in the CoP (Cognitive Perception) in the learning and planning system. On the one hand, CoP manages the interpretation of different kinds of sensors and on the other hand it automatically acquires and maintains the knowledge about the world and objects in the world. CoP selects sensors and sensor interpretation algorithms based on their expected utility. To this end, CoP learns and improves intersensor and inter-algorithmic models for method seleciton from experience. Especially the vision system, the major sensor we use, provides several automatic model improving techniques. Improved models accelerate the perception process and provide more robust results.
  • KnowRob / Knowledge 4 CoTeSys : The project targets at building knowledge representation and processing systems for mobile robots by combining description logics knowledge bases with data mining, (self-) observation modules and imported knowledge from the World Wide Web.
  • EnvMod : We focus on point cloud based representation and reasoning techniques for building accurate and meaningful 3D maps for mobile robots in both indoor and outdoor environments. One of our main application and deployment scenario is the Assistive Kitchen. However, all our methods were carefully crafted with generality in mind, therefore they have been also successfully applied to outdoor urban, aerial, and underwater datasets.

Former Research Projects

  • CogMaSh : CogMaSh aims to create a manufacturing system that achieves similar levels of flexibility, robustness and improvement through experience as found in human machine shops. In the production of prototypes, customized products and small or mid-size series, human workers with their problem solving abilities, experience and cognitive capabilities are still the single way to provide the required flexibility, adaptability and reliability.


Selected Publications

  • 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)
For the full list of my Publications please see the Publications section.
Document Actions