Motion sequence scheme for detecting mobile robots in an office environment

Abstract

The ability to detect a robot by sensing would be very useful, especially in multi-robot applications. It may not be realistic that all robots will have a special device aimed at the robot detection. In this paper, we propose a multi-robot system that enables the robot detection in office environments by a normal vision system, without additional infrastructure. The method utilizes a feature called "motion sequence (MS)" to identify a robot. This feature is robust against disturbances such as occlusions, invariant illuminations, etc., and represents characteristics of robot motions well. In addition, we introduce support vector machine (SVM) based classifier to detect errors in the robot detection as much as possible. To enhance the accessibility of MS as well as personal information of each robot, we propose a simple Web-based system by which robots transmit personal information via wireless network. The system is also used for interaction between robots and human users.