Cognitive architecture for intuitive and interactive task learning in industrial collaborative robotics
François Hélénon, Stéphane Thiery, Éric Nyiri, Olivier Gibaru · 2021
This paper introduces a cognitive architecture, implemented in python3, designed with industrial collaborative robotics specifications in mind, to engage in a mixed-initiative teacher/learner setting called interactive task learning: a human can teach the robot, with natural and multimodal communication means, how to perform a task. The architecture has been built around explainable, modular representations (relational graphs and behavior trees) to ease the upgradability of the system and AI modules to adapt to realistic and complex settings. A first prototype based on speech and gesture communication means is proposed and has been validated on an industrial system to learn an unknown task. A link to a video of this validation is attached in the article.