[Edaily Reporter Shin Yeong-bin ] A pilot project for “network-collaborative physical AI”—which utilizes the artificial intelligence (AI) computing resources of telecommunications base stations to support robots’ decision-making and task execution—is being launched for the first time at an industrial site in South Korea.

The KoreaInformation&Communication Agency (NIA) announced on the 15th that it is launching a field demonstration combining AI-based wireless access networks (AI-RAN) and Physical AI services as part of a competitive grant program by the Ministry of Science and ICT.

This project is a field demonstration that applies high-performance graphics processing unit (GPU) computing capabilities to communication infrastructure based on “Ieum 5G”—a 5G mobile communication private network—and links this with robots on-site.

AI-RAN is a next-generation wireless network technology that integrates AI computing resources into communication base stations to optimize the communication environment and support the AI computations required by nearby devices.

Eight domestic institutions and companies, including the National Information Society Agency (NIA), are participating in the demonstration. Wizcore is responsible for building the AI-RAN integrated platform and conducting the demonstration, while HFR is in charge of developing and integrating the related equipment.

FRTEK CO., LTD. and Wave Electronics Co., Ltd. are developing open wireless access network equipment, while the Korea Electronics Technology Institute is responsible for the integration between physical AI and the network. ROBOTIS(108490)is designing humanoid robots, and HYUNDAI STEEL(004020)is providing the demonstration site.

The demonstration will take place at HYUNDAI STEEL’s Dangjin Hot Rolling Mill. Previously, when equipment malfunctions occurred, remote intervention was difficult, forcing workers to enter the high-temperature site in person to address the issue.

On-site robots also faced limitations in performing complex AI inference and taking actions at the level of skilled workers due to constraints on the performance of their onboard semiconductors and battery capacity.

To address this, the NIA is establishing a response system that spans from the robot’s on-site perception to AI-RAN’s analysis and decision-making, and finally to the robot’s actual actions.

This structure reduces the hardware burden on the robots by having the network share the load of complex AI computations, thereby enabling them to perform high-difficulty tasks such as identifying and resolving equipment malfunctions.

The project will also demonstrate multi-robot collaborative control, in which multiple robots are connected to a single AI-RAN. This approach allows robots to share hazard information and work status in real time and coordinate their actions with one another.

Participating organizations will complete infrastructure deployment and on-site demonstrations by the end of this year. They also plan to conduct certified testing and verification by linking the domestic 5G-A testbed with the “O-RAN Global Plugfest,” an international open radio access network testing event.

After completing infrastructure deployment and verification, the NIA intends to refine the technology through two years of performance management and lay the groundwork for its expansion to other industrial sites.

Kim Hyung-chul, President of the NIA, stated, “This demonstration ushers in the era of Physical AI, which combines telecommunications, AI, and physical robots,” adding, “We will create field demonstration cases that domestic AI network and Physical AI companies can utilize to enter the global market.”