RT Journal Article
T1 Bioinspired decision-making for a socially interactive robot
A1 Perula Martínez, Raúl
A1 Castro González, Álvaro
A1 Malfaz Vázquez, María Ángeles
A1 Alonso Martín, Fernando
A1 Salichs Sánchez-Caballero, Miguel
AB Nowadays, robots and humans coexist in real settings where robots need to interact autonomously making their own decisions. Many applications require that robots adapt their behavior to different users and remember each user’s preferences to engage them in the interaction. To this end, we propose a decision making system for social robots that drives their actions taking into account the user and the robot’s state. This system is based on bio-inspired concepts, such as motivations, drives and wellbeing, that facilitate the rise of natural behaviors to ease the acceptance of the robot by the users. The system has been designed to promote the human-robot interaction by using drives and motivations related with social aspects, such as the users’ satisfaction or the need of social interaction. Furthermore, the changes of state produced by the users’ exogenous actions have been modeled as transitional states that are considered when the next robot’s action has to be selected. Our system has been evaluated considering two different user profiles. In the proposed system, user’s preferences are considered and alter the homeostatic process that controls the decision making system. As a result, using reinforcement learning algorithms and considering the robot’s wellbeing as the reward function, the social robot Mini has learned from scratch two different policies of action, one for each user, that fit the users’ preferences. The robot learned behaviors that maximize its wellbeing as well as keep the users engaged in the interactions.
PB Elsevier
SN 1389-0417
YR 2019
FD 2019-05-01
LK https://hdl.handle.net/10016/35571
UL https://hdl.handle.net/10016/35571
LA eng
NO The research leading to these results has received funding from the projects: Development of social robots to help seniors with cognitive impairment (ROBSEN), funded by the Ministerio de Economía y Competitividad (DPI2014-57684-R); and RoboCity2030-III-CM, funded by Comunidad de Madrid and cofunded by Structural Funds of the EU (S2013/MIT-2748).
DS e-Archivo
RD 27 jul. 2024