EN FR
EN FR


Section: New Results

Interactive Learning and user adaptation

Interactive learning from unlabeled instructions

Participants : Grizou Jonathan [correspondant] , Itturate Inaki, Montesano Luis, Pierre-Yves Oudeyer, Manuel Lopes.

Interactive learning deals with the problem of learning and solving tasks using human instructions. It is common in human-robot interaction, tutoring systems, and in human-computer interfaces such as brain-computer ones. In most cases, learning these tasks is possible because the signals are predefined or an ad-hoc calibration procedure allows to map signals to specific meanings. In this work, we addressed the problem of simultaneously solving a task under human feedback and learning the associated meanings of the feedback signals. This has important practical application since the user can start controlling a device from scratch, without the need of an expert to define the meaning of signals or carrying out a calibration phase. We proposed an algorithm that simultaneously assign meanings to signals while solving a sequential task under the assumption that both, human and machine, share the same a priori on the possible instruction meanings and the possible tasks. This work was published in a conference paper [45] and a journal paper will be submitted in January 2015.

We communicated about this work to the human-robot interaction (HRI) community. A robot equiped with our algortihm would be able to interact with a human without knowing in advance the specific communicative signals used by the human. This work was published in the HRI Pionneer workshop [46] .

This work was presented during the thesis defense of Jonathan Grizou entitled: Learning from Unlabeled Interaction Frames, on October 24, 2014. The video, slides, and thesis manuscript can be found at: http://jgrizou.com/projects/thesis-defense/

Calibration-Free BCI Based Control

Participants : Grizou Jonathan [correspondant] , Itturate Inaki, Montesano Luis, Pierre-Yves Oudeyer, Manuel Lopes.

We applied previous work on interactive learning from unlabeled instructions [45] to Brain-Machine Interaction problem, leading to a Calibration-Free brain computer interfaces. So far in such brain-computer interfaces (BCI), an explicit calibration phase was required to build a decoder that translates raw electroencephalography signals from the brain of each user into meaningful instructions. Our method removes the calibration phase, and allows a user to control a device to solve a sequential task. We performed experiments where four users use BCI to control an agent on a virtual world to reach a target without any previous calibration process. Our approach is promising for the deployments of BCI applications out of the labs. This work was published in a conference paper [44] and a journal paper will be submitted in January 2015.

This work was presented during the thesis defense of Jonathan Grizou entitled: Learning from Unlabeled Interaction Frames, on October 24, 2014. The video, slides, and thesis manuscript can be found at: http://jgrizou.com/projects/thesis-defense/