KT and Samsung Cut High-Speed Rail Dropouts by 29% With AI

Why trains are hard on networks

On high-speed trains running at 250 km/h or faster, the base station a device connects to changes at short intervals depending on where the train is. If the handover to the next base station does not happen in time, data use and calls can be disrupted.

What the companies did

The two companies analyzed radio signals and connection failure records collected along the rail route using AI. Instead of checking individual base stations on site, they identified quality problems that recurred across multiple sections through data.

The analysis showed that the main cause was a delay in switching: devices kept their link to the existing base station too long before moving to the next one. The companies then adjusted settings so that handovers happen earlier in the sections where problems were concentrated.

What comes next

Based on the trial results, KT plans to consider applying the technology to actual high-speed rail lines. It will also continue working with Samsung Electronics to jointly develop and validate AI radio access network (AI-RAN) technology.

Jeong Jin-guk, head of the Next Generation Communications Research Center at Samsung Research, said the trial showed that customers' perceived network quality can be improved even in fast-moving environments with poor reception. Lee Jong-sik, head of the Future Network Lab in KT's network division, said KT will keep developing AI-RAN technology aimed at tangible improvements in customer experience.