Benchmarking Human versus Robot Performance in Emergency Structural Inspection

Abstract

Human–robot collaboration (HRC) has gained significant momentum in various civil engineering applications. A representative application is robot-assisted emergency structural inspection, i.e., a human–robot partnered on-site investigation of structural damage and possible causes after any human-caused or natural incidents. Human experts are good at making informed decisions that reflect structural failure patterns, whereas robots can help in searching larger areas and in hazardous places. An effective HRC team would leverage the advantages of human and robot agents. However, deeper insights and direct evidence are still needed to build a baseline model, and in particular to identify the relative advantages and limitations of human and robot agents in emergency structural inspections. To establish a basis for efficient HRC emergency inspection team design, this paper presents a benchmark to compare human and robot performance in a simulated emergency structural inspection task following an earthquake disaster. The total amount of identified structural damage, inspection time, and route patterns of humans and robots are compared. The results show that humans outperformed autonomous robots in most inspection metrics, possibly owing to prior knowledge about the patterns of structural failure, and presented a different but more efficient route pattern. The findings are expected to inspire better HRC team design for emergency structural inspection tasks and other relevant knowledge-based HRC tasks.