A simulated environment for long-term interactions with episodic memory

The expansion of robots use in the domestic environment demanded the development of robotic systems with the capacity to maintain long-term interaction with human beings in the execution of everyday tasks. This need produced an increase in research in Human-Robot Interaction (HRI) in particular in the area of social robotics. However, in a long-term interaction, the use of social robots requires the ability to learn users' preferences and adapt their behavior according to changes in the environment, presenting a natural and friendly interaction with human beings, generating engagement and enabling future interactions. In this article, we present a cognitive system with episodic memory for a robotic agent in a domestic environment. The cognitive system selects and stores the most important interactions with users, learning with each person's preference. The system was tested and validated using the snack and drink suggestions scenario.