Impression's predictive models for animated robot

Abstract

Some studies in the field of HRI show that the impressions given by a robot depends on motions. The robot that is developed for communication with human needs to give the appropriate impression of the users. Yet robot motion gives a different impression to the observers. This means that robot designers and programmers have to make trial and errors to generate behaviors conveying the appropriate impressions. This study developed a predictive model of user's qualitative impression scores, and evaluates the effectiveness of the model using experimental data. In the experiment, 2 kinds of humanoid robots presented 6 behaviors and participants rated the qualitative impression. We compared the obtained scores with the predictive scores calculated by our model. As a result, it is possible to predict users' impression scores for robot behaviors is 89.6% of the cases.