Emotion expression in robots

A study on the effect of a virtual gravity vector on the perceived emotional content of robot movement

More Info
expand_more

Abstract

Adding emotion to motion, or creating emotionally meaningful movement is one of the next steps in robotics. Non-verbal communication between humans and robots has the potential to raise the current interaction level to a next stage. This study is a contribution to the field of emotion expression in robotics. Several parameters are known for influencing the emotional content of a movement in humans. In this study the influence of a changing virtual gravity parameter is investigated. Simulations were created of a virtual arm executing various daily tasks. Each task was executed under three different virtual gravity conditions: gravity vector pointing down, no gravity and gravity vector pointing up. On a website, a user study was created for people to rate the videos of the movement of a robotic arm in terms of its emotional content. The robotic arm performed ten different tasks. Two response tools were used for the participant to rate the videos: the AffectButton and the Self-Assessment Manikin. A total of 275 participants took part in the user study. The overall effect of gravity on the emotional content of the arm movement was found to be rather limited. However, in a number of cases a small significant effect was measured. Gravity had a small significant effect on dominance measured in the AffectButton: the arm movements, when the gravity vector was in the up and downward direction, were rated with a higher level of dominance than in the situation of no gravity. Also, gravity had a small significant effect on pleasure measured in the Self-Assessment Manikin: downward gravity resulted in a higher level of pleasure than in the neutral and upward case. The overall effect of task was generally more significant. When comparing individual tasks, significant difference were found on the ratings of the three variables pleasure, arousal and dominance, in both the AffectButton and the Self-Assessment Manikin.

Files

Thesis_Suzanne_Weller.pdf
(pdf | 11.1 Mb)
License info not available