Variety of Robot Behaviors for Human-Robot Interactions Using Attention

Abstract

Abstract Human-robot interaction (HRI) is essential in order for robots to play a role in everyday life. To achieve this, one problem is the fixation of robot’s reaction. In many existing HRI systems, the behavior of robots is fixed to each situation which is assumed beforehand. The human can easily predict the robot’s reaction in such case and becomes to feel the robot as not an life-like agent but a toy. To solve this problem, we propose a novel model of attention, named Feature Drift. With Feature Drift, a robot can determine its attention spontaneously and change its attention over time. By changing robot’s attention, the robot becomes to react variously, not fixed. This study evaluates the behavior of Feature Drift model in simulation and designs an HRI system. We also test the HRI system, and then confirms that the system can generate various reactions.