A new paradigm to study social and physical affordances as model-based reinforcement learning - Action, Cognition, Interaction et Décisions Encorporées
Article Dans Une Revue Cognitive Robotics Année : 2024

A new paradigm to study social and physical affordances as model-based reinforcement learning

Résumé

Social affordances, although key in human-robot interaction processes, have received little attention in robotics. Hence, it remains unclear whether the prevailing mechanisms to exploit and learn affordances in the absence of human interaction can be extended to affordances in social contexts. This study provides a review of the concept of affordance in psychology and robotics and proposes a new view on social affordances in robotics and their differences from physical affordances. We moreover show how the model-based reinforcement learning theory provides a useful framework to study and compare social and physical affordances. To further study their differences, we present a new benchmark task mixing navigation and social interaction, in which a robot has to make a human follow and reach different goal positions in a row. This new task is solved in simulation using a modular architecture and reinforcement learning.
Fichier principal
Vignette du fichier
Chartouny2024affordances.pdf (3.03 Mo) Télécharger le fichier
Origine Fichiers éditeurs autorisés sur une archive ouverte
licence

Dates et versions

hal-04678813 , version 1 (27-08-2024)

Licence

Identifiants

Citer

Augustin Chartouny, Keivan Amini, Mehdi Khamassi, Benoît Girard. A new paradigm to study social and physical affordances as model-based reinforcement learning. Cognitive Robotics, 2024, 4, pp.142-155. ⟨10.1016/j.cogr.2024.08.001⟩. ⟨hal-04678813⟩
6 Consultations
0 Téléchargements

Altmetric

Partager

More