SK Telecom is building a leading "AI-RAN" network that combines artificial intelligence (AI) computing with base stations and will validate physical AI services such as patrol robots, autonomous transport vehicles, and humanoids at industrial sites. The goal is for the network to share not only data transmission but also the robots' AI computation to reduce latency and power consumption.
SK Telecom said on the 14th it was selected as an implementing agency for the "Hyper AI network infrastructure" demonstration project led by the Ministry of Science and ICT and the National Information Society Agency (NIA). The project will run for two years through 2027 and test applicability in manufacturing, logistics, and industrial safety in Incheon, Pangyo, and Pyeongtaek.
◇ Base stations handle AI computation too… compare performance of four vendors' equipment
AI-RAN is a next-generation network technology that equips mobile base stations with AI computing resources to provide communications and computing simultaneously. Because base stations share high-performance computation that robots or Autonomous Driving equipment previously had to handle on their own, device hardware burden and battery drain can be reduced.
In this project, SK Telecom will deploy AI-RAN equipment from four manufacturers, including Samsung Electronics, HFR, Ericsson, and Nokia. It plans to operate equipment from different vendors within the same project and compare performance, stability, and service suitability.
AI compute units will be configured in various ways, including CPUs and GPUs. The placement of user plane function (UPF), which handles data forwarding in AI servers and the 5G core, will also be designed in multiple options to measure which architecture best fits ultra-low-latency physical AI services.
It will also apply 5G standalone (SA), network slicing, service management and orchestration (SMO), and AI-based network autonomy technologies. SMO is a technology that manages and controls an open RAN environment composed of multiple vendors' equipment from a single system.
◇ Three demonstrations: patrol robots, autonomous transport, humanoids
The demonstrations span three areas: quadruped patrol robots, unmanned autonomous transport, and humanoid low-power mode. They will quantitatively verify high-volume video transmission, ultra-low-latency communication, mobility, distributed computing, and network reliability.
The quadruped robot patrols hazardous factory areas that are hard for people to access and transmits high-definition video in real time. AI-RAN analyzes the transmitted footage with AI models to detect risks such as fires or equipment anomalies and alerts the control system.
The unmanned autonomous transport service aggregates LiDAR sensor data installed at the factory to the base station, recreates the site in a digital space, and then issues remote driving commands to vehicles. Because each vehicle does not need to be equipped with a high-performance compute unit, routes and obstacles can be analyzed centrally, which SK Telecom said can lower logistics automation expense.
In the humanoid demonstration, part of the complex AI computation performed by the robot is offloaded to the base station. It will verify ways to reduce onboard computation and power consumption to extend battery life and deliver high-level services with relatively low-cost hardware.
◇ Starting in Incheon and Pangyo, expanding to KG Mobility's factory
In the first year, an AI-RAN leading network will be built at SK Incheon Petrochem's business site and in Pangyo. In Incheon, based on Samsung Electronics' equipment, it will test an industrial safety control service using quadruped robots and mobile closed-circuit (CC)TV. The focus is AI monitoring of high-risk facilities and blind spots at chemical plants.
In Pangyo, it will create a physical AI living lab applying HFR equipment and validate the unmanned autonomous transport service. In the second year, the technologies verified in Pangyo will be moved to KG Mobility's Pyeongtaek plant for application at real automobile production and logistics sites. The Pyeongtaek plant will be equipped with an AI-RAN network and SMO using Ericsson's 5G technology.
The project is led by SK Telecom, with Ericsson-LG and HFR participating on the equipment side. Intellivix, Seoul Robotics, and Klever are responsible for developing physical AI services. SK Incheon Petrochem and KG Mobility are participating as demand-side institutions, while Samsung Electronics and Nokia will supply equipment and collaborate on technology.
SK Telecom plans to develop performance indicators based on data obtained from the demonstrations and reflect them in international standardization discussions at O-RAN and 3GPP. It aims to support the government's "AI expressway" initiative by connecting AI data centers, AI-RAN base stations, and robots at industrial sites with an ultra-low-latency network.