Developing smart MAVs for autonomous inspection in underground constructions

Smart Micro Aerial Vehicles have transformed infrastructure inspection through high manoeuvrability, enabling efficient monitoring of tunnels, subterranean structures, and other environments inaccessible to ground equipment. This study proposes a novel autonomous framework featuring: a hierarchical perception-planning architecture that autonomously identifies critical structural components with 0.962 F1-score for column recognition and 85% wall recognition accuracy; an adaptive exploration strategy constructing obstacle maps and dynamically adjusting inspection paths in real-time; and experimental validation in a 4,000m² underground facility demonstrating positional RMSE below 0.2m at 1.75m/s flight speeds. In summary, this study presents an autonomous MAV framework capable of efficient tunnel inspection without human intervention.