Two ICRA 2018 papers accepted

  • Koert, D.;Maeda, G.; Neumann, G.; Peters, J. Learning Coupled Forward-Inverse Models with Combined Prediction Errors [BibTeX]

This paper revisits concepts from sensorimotor learning algorithms based on mixture of experts, MOSAIC in particular. This work proposes a novel formulation to learn one-to-many mappings, such as the ones found in the Inverse Kinematics of redundant arms.  It reports experiments using a real KUKA lightweight arm where different IK solutions are found as a function of the context. The figure below is part of the paper and gives the overall motivation.

doro_icra2018

  • Lioutikov, R.; Maeda, G.; Veiga, F.; Kersting, K.; Peters, J. Inducing Probabilistic Context-Free Grammars for the Sequencing of Robot Movement Primitives [pdf][BibTeX]

Movement Primitives are a well studied and widely applied concept in modern robotics. Composing primitives out of an existing library, however, has shown to be a challenging problem. We propose the use of probabilistic context-free grammars to sequence a series of primitives to generate complex robot policies from a given library of primitives. The rule-based nature of formal grammars allows an intuitive encoding of hierarchically and recursively structured tasks. This hierarchical concept strongly connects with the way robot policies can be learned, organized, and re-used. However, the induction of context-free grammars has proven to be a complicated and yet unsolved challenge. In this work, we exploit the physical nature of robot movement primitives to restrict and efficiently search the grammar space. The grammar is learned with Markov Chain Monte Carlo optimization over the posteriors of the grammars given the observations. Restrictions over operators connecting the search define the corresponding proposal distributions and, therefore, guide the optimization additionally. In experiments, we validate our method on a redundant 7 degree-of-freedom lightweight robotic arm on tasks that require the generation of complex sequences of motions out of simple primitives.

3rd Hand Robot Project

From November 2013 to October 2017 I managed one of the work packages of the European Funded Project 3rd Hand. Related to the 3rd Hand, I also led the efforts of the IAS group into the topic of human-robot collaboration.

During this period, we proposed a new representation of primitive movements that accounts for both robot and human collaborative skills. We called it Interaction Probabilistic Movement Primitives. Investigations towards this new representation led to a number of publications in top-notch journals and conferences [IJRR, AURO, ISRR, HUMANOIDS, etc.] including a nomination for the best paper award at ICRA 2015 with this paper.

We also contributed to other aspects related to collaborative skills, such as ergonomics, joint skill learning, learning from observations, interactive learning, etc. and organized a workshop at IROS in 2016.

You can find the complete list of publications of the project here.

Workshop on Human-Robot Collaboration. IROS 2016

In 2016 I organized an IROS workshop held in Daejeon, Korea [website].

The workshop was organized with Luka Peternel (HRI2, ADVR, IIT, Italy), Leonel Rozo (IIT, Italy), Serena Ivaldi (INRIA Nancy Grand-Est, France), Claudia Pérez D’Arpino (MIT, USA), Julie A. Shah (MIT, USA), Jan Babič (JSI, Slovenia), Tamim Asfour (KIT, Germany) and Erhan Oztop (Ozyegin University, Turkey).20161010_175413_previewIMG_3661_preview