Walk the Talk: Gestures in Mobile Interaction

This study aims at describing navigation guidelines and concerning analytic motion models for a mobile interaction robot, which moves together with a human partner. We address particularly the impact of gestures on the coupled motion of this human-robot pair. We pose that the robot needs to adjust its navigation in accordance to its gestures in a natural manner (mimicking human-human locomotion). In order to justify this suggestion, we first examine the motion patterns of real-world pedestrian dyads in accordance to 4 affective components of interaction (i.e. gestures). Three benchmark variables are derived from pedestrian trajectories and their behavior is investigated with respect to three conditions: (i) presence/absence of isolated gestures, (ii) varying number of simultaneously performed (i.e. concurring) gestures, (iii) varying size of the environment. It is observed empirically and proven quantitatively that there is a significant difference in the benchmark variables between presence and absence of the gestures, whereas no prominent variation exists in regard to the type of gesture or the number of concurring gestures. Moreover, size of the environment is shown to be a crucial factor in sustainability of the group structure. Subsequently, we propose analytic models to represent these behavioral variations and prove that our models attain significant accuracy in reflecting the distinctions. Finally, we propose an implementation scheme for integrating the analytic models to practical applications. Our results bear the potential of serving as navigation guidelines for the robot so as to provide a more natural interaction experience for the human counterpart of a robot-pedestrian group on-the-move.