Harnessing Long-term Memory for Personalized Human-Robot Interactions

Long-term memory system along with a working memory paves the way for a personalized human-robot interaction. Memorability based on visual and auditory cues can potentially enhance the simple stimulus-driven interaction process towards a more human-like and natural one. Most of the existing interactive humanoids, to some extent, utilize the short-term or working memory for the robot to establish interaction. This work incorporates a stable long-term memory system for the socially interactive robot ROBIN, triggering a notion of episodic memory. A MySQL database system has been integrated into the existing robot behavior control system in order to store and retrieve essential information e.g., name, age, occupation etc. of an interaction partner. The proposed personalized interaction system utilizes the interlocutors' past events / experiences in different interaction scenarios. Realistic robot behaviour in the form of speech, gesture / posture and facial expressions has been generated. The user experiences collected with a questionnaire-based evaluation process indicate a notion of fluid and realistic human-robot interaction.