Discrimination of Humanoid Robots from Real Human in a Perception Sensor Network

Abstract

Since the appearance of humanoid robots is similar with real human, a method is proposed to discriminate these robots from real human in a perception sensor network (PSN) which uses multiple Kinects and PTZ cameras to provide the location and name of tracked human to these robots. To this end, the global map which is acquired by fusing and calibrating one local map built by a robot and the other local map by the PSN system is used to provide the location of the robot. Since the PSN updates periodically the location of the robot by subscribing the ROS (Robot Operating System) topics, the robot which is wrongly detected as human but yields the closest distance to the location can be correctly identified as a robot in the system. Therefore, the misunderstood cases when robots are recognized as human in the PSN system are removed effectively. The experimental results demonstrate the outperforming of the proposed method in various scenarios.