Learning from Demonstration and Correction via Multiple Modalities for a Humanoid Robot

Brenna Argall^* and Aude Billard^*

^(*) Learning Algorithms and Systems Laboratory (LASA), École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland
E-mail: brennadee.argall@epfl.ch, aude.billard@epfl.ch

Abstract

This paper reports ongoing work that employs multiple demonstration modalities in order to accomplish motion control learning in a multi-staged policy adaptation process. A novel interface for providing tactile guidance to correct learned motion control behaviors is introduced. This interface extends our prior work by making use of a more sophisticated set of tactile sensors, developed by the ROBOSKIN consortium.