Human Robot Comparison in Rapid Structural Inspection

Abstract

Amid the rapid development of robotic technologies and artificial intelligence, Human Robot Collaboration (HRC) has gained momentum in a variety of civil engineering applications. However, with robots only obeying redefined algorithms without human-like intelligence, there has not been a widely accepted method that enables a complete integration of human robot team in knowledge-and-experience-based tasks, such as building inspection. To enhance the efficiency of inspection tasks, a deeper insight must be gained into the advantages and limitations of human intelligence and robotic algorithms, respectively, and how these two can be seamlessly integrated. As the first step, in this paper we designed an experiment to compare human and robot performance in a building inspection task. A quadrupedal robot is simulated in ROS (Robot Operating System) Gazebo, which automatically navigate in and scan the buildings with SLAM (simultaneous localization and mapping) and RRT (rapidly exploring random tree) algorithms, while human experts finishing the same inspection task in Virtual Reality. The total identified structural defaults, inspection time, and routes are recorded and compared. The result shows that there is an apparent pattern difference of human route plans with considerably better accuracy and efficiency in building inspection.