Real-Time Robot Personality Adaptation based on Reinforcement Learning and Social Signals

Hannes Ritschel, Elisabeth André

Proceedings of the Companion of the 2017 ACM/IEEE International Conference on Human-Robot Interaction (HRI)

Verlag: ACM


Recent findings in Human-Robot Interaction (HRI) indicate that the adaptation of a robot's behaviors to the human's personality profile makes interaction more engaging, but also that it depends on the task context whether a similar or opposing robot personality is preferred. This late breaking report presents our ongoing work on an approach using Reinforcement Learning and social signals for figuring out and adapting to the human preferences, i.e. desired personality profile. Our scenario involves a "Reeti" robot in the role of a story teller talking about the main characters in the novel "Alice's Adventures in Wonderland" by generating descriptions with varying degree of introversion/extraversion. The learning process is running in real-time during the interaction and allows for simultaneous adaptation without explicitly asking the user about its preferences.