Advancing intuitive human-machine interaction with human-like social capabilities for education in schools




Europe Flag

This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 765955

Identifiez-vous pour accéder à la partie privée du site ou le modifier. Déconnexion.

ESR9 - Learning from and about humans for robot task learning

 picture_BIED_ANIMATASwebI joined the Institut des Systèmes Intelligents et de Robotique (ISIR) at Sorbonne Université as a Ph.D. student in Social Robotics in March 2018, where I’m supervised by Prof. Mohamed Chetouani. Within the ANIMATAS project, I’m working on a Learning-by-Teaching approach allowing children to increase their understanding of a task by showing how to solve it to a robot. Therefore, I´m equally interested in the manner how humans teach, how robots can learn from that teaching, and how this process can be used to foster learning of children.

Before joining ISIR I received a B.Sc. and M.Sc. degree in Electrical Engineering from TU Darmstadt, Germany. In the Intelligent Autonomous Systems Institute (IAS) headed by Prof. Jan Peters, I participated in a project on a bipedal walker and in a project on Brain-Computer-Interfaces.Finally, I conducted my Master’s thesis in cooperation with Honda Research Institute Europe (HRI) in Offenbach, Germany, on the Learning from Demonstration approach using a DARwIn-OP2 robot.

Main host institution: Sorbonne Université

Supervisor: Mohamed Chetouani (UPMC), in association with Amit Pandey (SBR)

Secondment institution: SBR

Objectives: There is growing of interest for Robot Learning from Demonstration with the aim of developing machines that are able to take advantage of human interactions. Interpersonal skills and the nature of the task impact the performances. Our main objective is to investigate the role of interpersonal interaction in task learning/sharing situations and in particular the impacts on interactive learning schemes. We will focus on social learning mechanisms that could allow a robot to learn a task from interactions with humans (in particular teachers) but also inform about humans (e.g. identity, personality traits, engagement and expertise). We will explore socially guided machine learning techniques (neural networks, reinforcement learning). Interpersonal interactions (including task sharing) will be explicitly modeled at the feature (e.g., synchrony between actions) or decision (e.g., rewards) levels. Multi-task learning framework will be employed for learning social traits and tasks. The final aim is to propose strategies for bootstrapping robot task learning. The models will be evaluated in experimental settings with human teachers demonstrating various sets of tasks (object recognition, sequence learning) to robots. 


Expected results: Completed PhD dissertation; computational model allowing robot to learn from adult teachers demonstration and guidance; peer-reviewed publications 


For further information, contact: Mohamed Chetouani

ANIMATAS – MSCA – ITN – 2017 - 765955 2