Main research interests

I am interested in fundamental and applied research in robotics, machine learning and artificial intelligence. Currently, I mainly work on problems around navigation, perception and human-robot interaction. I often use techniques from signal processing, optimal control, statistics and optimization to this end. Some concrete research topics include:

  • (Inverse) reinforcement learning and optimal control for behavior development
  • Interactive & life-long learning (esp. for mobile robots)
  • Activity recognition for mobile systems (esp in crowds)
  • Dynamical systems for control (e.g. DMPs)
  • Game theory (Markov games)
Current project
  • SPENCER → [ALU-FR, DE]

    Social situation-aware perception and action for cognitive robots, an EU FP7-ICT project. In this project, we aim to break new grounds in interactive and cognitive systems acting in crowded environments and especially in domains including human behavior modelling and learning, individual and group perception, socially-aware task and motion planning, socially normative mapping and navigation, etc. Further details here.
Selected past projects
  • ...

    Multi-scale representation of 3D objects → [JUB, DE]

    For my masters thesis, I have developed SCSS (Sliced Curvature Scale Space) for representing 3D obejcts and developed feature extraction methods as well as recognition of objects. I also did dimensionality reduction via PCA and NMF and objection classification using SVMs.
  • ...

    Real-time visual head tracking for HRI → [Idiap, CH]

    I visited Idiap Research Institute in Martigny in 2012/2013 and worked on real-time visual tracking on the Nao robot exploiting depth from stereo vision. We adopted sequencial monte carlo (particle filtering) approach. Additional detection filtering using depth information from disparity maps. Work done under the HUMAVIPS EU project.
  • ...

    Simple and General Hands → [Robotics Institute - CMU, USA]

    In Summer 2010, I visited the Manipulation Lab at Robotics Institute and worked on developing a control architecture for managing Grasp Learning and Control experiments using ROS. We experimented with marker localization using RVMs using only finger encoder data from the four fingers of our simple hand design. Project page.
  • ...

    CRAM (Cognitive Robot Abstract Machine) → [TUM, DE]

    Summer 2011, I visited IAS Group at TU Munich. I worked on generalizing robot motion plans by leveraging techniques from LfD (Learning from Demonstration), control theory and motion planning for the CRAM platform. I performed experiments on the PR2 robot to demonstrate learning by imitation using SEDS and DMPs. The IAS group has since relocated to the Intitute for Artificial Intelligence at Uni-Bremen.
  • ...

    Manipulation → [DFKI-Robotics Innovation Center, DE]

    During my studies, I worked as a HiWi (Research Assistant) at the DFKI Lab in Bremen. We developed the robot AILA. I also worked on multi-fingered grasp manipulation of shoes for footwear industries. I was particularly involved in building and setting up systems for the investigation such as simulators. I also set up speech on AILA robot using CMU Sphinx and added a complimentary text-to-speech system.
  • ...

    Robot Chess Player (Ammbot) → [JUB, DE]

    Oliver Dunkley and I developed Ammbot to play chess against humans for our bachelor thesis. Originally an independent study project started by Oliver, Alexandra, Flavia, Misa and I, we worked on motion planning, control and perception problems in the chess playing setup. The funding for the independent study project was provided by DAAD-PROFIN program.
  • Other Projects (mostly during Bsc years) → [Various places]

    I was also involved in a number of projects during my bachelor studies spanning many areas including;
    • Google summer of code 2009: GUI development using Qt and C++
    • Rasdaman: C++ development for an open source raster database system
    • Mapinteractor: Developing a mobile app using Python and GTk on the OpenMoko framework for use in map interaction experiments