The News, Explained
SK Telecom published details of its AI-RAN vehicle-moving demonstration and next steps on October 15. The technology received the Most Innovative 5G Communication Service award at Network X in Vienna on October 14. That local event date is separate from the press release’s publication date in Korea. The technical story is how sensors at a site work with AI processing at a communications base station to move vehicles. Source
AI-RAN is network infrastructure that provides communications and AI services together. SKT says it developed and tested a base station that maintained stable communications quality while also processing AI services. Its AVM, or autonomous vehicle marshalling, service uses that network to coordinate moving and arranging vehicles. The announcement describes managing movement within a defined site, such as a parking facility or a vehicle-handling operation. Source
The system divides the work among sensors that observe the surroundings, a base station that plans movements and vehicles that follow commands. LiDAR sensors installed at the site send point-cloud data: collections of three-dimensional coordinates describing objects’ positions and shapes. The AI-RAN base station uses those observations to construct a real-time digital twin, a virtual representation of the site. It calculates paths intended to avoid collisions and sends driving commands to the vehicles. Observation and path planning therefore sit in infrastructure outside the cars. Source
The vehicles need preparation too. SKT says ordinary vehicles without their own autonomous-driving capability can be moved without a driver after modules and software are installed. The site’s sensors and AI base station work as one system with the vehicle’s control equipment. SKT also describes using a view of the whole site to coordinate routes and prevent vehicles from blocking one another in narrow passages. These are the company’s account of the design and its intended benefits. Source
The reported demonstration took place in an actual parking facility and used autonomous valet parking. According to SKT, collecting sensor data, making AI decisions and delivering driving commands all finished within the time required for unmanned vehicle control. The demonstration vehicle completed parking without human intervention. The announcement reports that trial’s outcome; it provides no numerical processing time or safety-performance metric applicable across different sites. Source
The next stated step is industrial testing. Under the national Hyper-AI network project, SKT plans to apply and test the technology at a vehicle factory in Korea in 2027. It also intends to expand toward logistics sites such as ports and airports and develop physical AI services, in which AI helps control real vehicles or robots. Fleet-control services and subscription revenue models depending on scale are possibilities the company describes if the technology is commercialized. Source
OYOPICK’s Take
The distinctive idea is to share intelligence between a vehicle and its surroundings. If ordinary cars can be equipped for control while site infrastructure observes the environment and coordinates routes, physical AI could reach beyond the newest autonomous vehicles. A shared view of the site makes coordination between vehicles an important design problem alongside planning each vehicle’s own movement.
SKT’s planned 2027 factory trial is the next stated checkpoint for that possibility. Real operations may combine movements by several vehicles with people’s routes through the same site. Safely coordinating those movements as communications conditions or the site layout change would be an important criterion for industrial use. If testing extends from the parking demonstration to these operating conditions, we hope the technology can develop into useful support for people who repeatedly move vehicles as part of their work.