Intrinsically Motivated Multi-Task Reinforcement Learning With Open-Source Explauto Library and Poppy Humanoid Robot - Inria - Institut national de recherche en sciences et technologies du numérique
Vidéo Année : 2017

Intrinsically Motivated Multi-Task Reinforcement Learning With Open-Source Explauto Library and Poppy Humanoid Robot

Afficher 

Résumé

Our demonstration presents an open-source hardware and software platform which allows non-roboticists researchers to conduct machine learning experiments to benchmark algorithms for autonomous exploration and active learning. In particular, in addition to showing the general properties of the platform such as its modularity and usability, we will demonstrate the online functioning of a particular algorithm which allows efficient learning of multiple forward and inverse models and can leverage information from human guidance. A first aspect of our demonstration is to illustrate the ease of use of the 3D printed low-cost Poppy humanoid robotic platform, that allows non-roboticists to quickly set up and program robotic experiments. A second aspect is to show how the Explauto library allows systematic comparison and evaluation of active learning and exploration algorithms in sensorimotor spaces, through a Python API to select already implemented exploration algorithms. The third idea is to showcase Active Model Babbling, an efficient exploration algorithm dynamically choosing which task/goal space to explore and particular goals to reach, and integrating social guidance from humans in real time to drive exploration towards particular objects or actions. [Forestier and Oudeyer, 2016] Forestier, S. and Oudeyer, P.-Y. (2016). Modular active curiosity-driven discovery of tool use. In 2016 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Daejeon, Korea. [Lapeyre et al., 2014] Lapeyre, M., Rouanet, P., Grizou, J., Nguyen, S., Depraetre, F., Le Falher, A., and Oudeyer, P.-Y. (2014). Poppy Project: Open-Source Fabrication of 3D Printed Humanoid Robot for Science, Education and Art. In Digital Intelligence 2014, page 6, Nantes, France. [Moulin-Frier et al., 2014] Moulin-Frier, C., Rouanet, P., Oudeyer, P.-Y., and others (2014). Explauto: an open- source Python library to study autonomous exploration in developmental robotics. In ICDL-Epirob-International Conference on Development and Learning, Epirob.

Dates et versions

medihal-01476320 , version 1 (24-02-2017)
medihal-01476320 , version 2 (14-04-2017)

Licence

Identifiants

  • HAL Id : medihal-01476320 , version 1

Citer

Sébastien Forestier, Yoan Mollard, Damien Caselli, Pierre-Yves Oudeyer. Intrinsically Motivated Multi-Task Reinforcement Learning With Open-Source Explauto Library and Poppy Humanoid Robot. 2017. ⟨medihal-01476320v1⟩
459 Consultations
81 Téléchargements

Partager

More