Adaptive Supervisory Control of a Communication Robot That Approaches Visitors

Abstract

In the future, we believe communication robots will be working in shopping malls, talking to people, and providing information for them. In a mall, a robot must choose appropriate visitors as targets to approach. But since robotics technology has not matured yet, fully autonomous communication robots that can deal effectively with problematic situations are not expected; annoying robots who obstruct people in malls must be avoided. Instead, we chose a supervisory control scheme where a robot requests permission from a human operator to approach selected visitors. To reduce the time required by the operator to give approval, the system adaptively learns the operator’s choice to reduce the number of approval requests based on the classification of visitor walking behaviors and their attitudes toward the robot. We evaluate the effectiveness of our adaptive supervisory control system in a field trial at a shopping mall. The experimental results demonstrated that 86.5% of the targets requested by the system for approval were appropriate. Moreover, operating time was reduced: the operator only controlled the robot 34.9% of the experiment time.