An AI-guided robotic system to assist in inspecting aging infrastructure has been developed at Drexel. The system, detailed in the journal Automation in Construction, aims to address the growing backlog of repairs needed for buildings, roads and bridges across the United States.
Led by Arvin Ebrahimkhanlou, PhD, assistant professor of civil, architectural and environmental engineering, and doctoral student Ali Ghadimzadeh Alamdari, the multi-scale system combines computer vision with deep learning algorithms to identify and assess cracks in concrete structures. The process begins with a high-resolution stereo-depth camera feed analyzed by a convolutional neural network trained to detect crack patterns.
Once a potential problem area is identified, a robotic arm uses laser scanning to create a detailed 3D image of the damaged section. This is combined with LiDAR scans of the surrounding structure to produce a “digital twin” model, allowing for precise measurement and ongoing monitoring of cracks.
In laboratory tests, the system outperformed existing technologies in detecting and measuring minute cracks, with sensitivity down to less than 0.01 millimeters. While human inspectors would still make final decisions on repairs, the robotic assistants could significantly reduce workload and minimize oversights.
The researchers envision integrating this technology into a broader autonomous monitoring framework, potentially including drones and unmanned ground vehicles. They plan to conduct real-world testing and collaborate with industry and regulatory bodies to refine the system for practical applications in maintaining structural integrity across various types of infrastructure.
