A micro drone captures footage of a rail line at the Osong test track. [Korea Railroad Research Institute]
A micro drone captures footage of a rail line at the Osong test track. [Korea Railroad Research Institute]

AI-equipped autonomous drones will take over safety inspections of railway infrastructure — such as bridges and power transmission towers — that are difficult for workers to access.

The Korea Railroad Research Institute announced Sunday that it has confirmed the feasibility of automating track-obstruction inspections using micro drones and AI, and will now begin developing the core technologies needed to make the system operational.

The research aims to detect obstacles and hazards on rail lines through drone footage and AI analysis, eliminating the need to deploy workers onto the tracks.

Railway track inspections often require workers to approach the tracks directly, creating operational burdens and safety risks. Fixed CCTV systems alone also struggle to provide continuous coverage of blind spots along the line — including tunnel entrances, areas around structures, drainage zones and cut sections.

Before formally launching the research, the institute conducted a proof-of-concept study confirming that micro drone footage could be analyzed by AI, with detection results — including video, location and time data — fed into a control system.

The project will use drones not merely as cameras but as mobile inspection sensors that convert track hazards into data. The system will be tested and validated at the institute's Osong test track.

A key challenge is linking drone footage collection, AI-based obstruction detection and inspection history management into a single integrated system — one that accounts for GNSS (global navigation satellite system) dead zones such as tunnels. The project centers on three core technologies: railway-specific autonomous drone flight, AI-based automated obstruction detection, and digital twin-based inspection history management.

The institute plans to expand the research further, building an automated track safety verification system based on railway physical AI that integrates drones, fixed CCTV footage, quadruped robots and digital twins.

Choi Jong-hyeok, a senior researcher at the institute, said the core goal of the project is "to support workers by identifying track hazards before they even approach the scene, by connecting drone footage to AI analysis and control system alerts." He added that the institute would pursue practical deployment through phased validation at the test track and at actual railway sites.


nbgkoo@heraldcorp.com