• Intelligent execution architecture enabling autonomous collaboration among heterogeneous robots from situational awareness to mission completion

• Agent-based collaboration platform where AI understands physical environments and determines optimal actions

• Combines ultra-high-speed networks, AI and robotics to expand across industries and target the global AX market

KT Corporation unveiled its Physical AI strategy and the K RaaS (KT Robot as a Service) platform at Mobile World Congress 2026, presenting a vision that connects robots, facilities and IT systems into a single intelligent ecosystem.

K RaaS goes beyond conventional robot control technology and serves as a robot orchestration platform capable of delivering operational Physical AI services in real-world environments. By integrating robots, facilities and legacy systems, the platform recognizes, analyzes and manages the entire lifecycle of services, enabling AI-driven automation to be applied directly to real business operations.

Designed for cloud environments, K RaaS can monitor and manage heterogeneous robots and equipment deployed worldwide in real time. Rather than automating individual robots, the platform orchestrates a holistic Physical AI system where service flows operate seamlessly as a unified structure.

The platform includes agents responsible for different roles. The Service Builder Agent enables enterprises to design and deploy robot-integrated services tailored to their environments without additional development.

Meanwhile, the K RaaS Agent provides a natural language interface through which users can check mission status, analyze operational data and automatically generate reports. Tasks that previously required monitoring dozens of control screens can now be accomplished through a simple conversational query.

KT has already secured Physical AI deployment references in industries such as semiconductor manufacturing plants, logistics centers and smart buildings by combining its ultra-high-speed network infrastructure with its generative AI model SOTA K and robot orchestration technologies.

KT emphasized that K RaaS is not merely a robot management or connectivity technology but a field-oriented Physical AI orchestration platform where AI understands the physical environment and drives optimal execution.

Next-Generation Robot Intelligence: VLA Agent that “Sees, Understands and Acts”

KT also showcased the VLA (Vision-Language-Action) Agent, a next-generation robot intelligence that integrates visual information and language understanding and translates them into real-world actions.

Built on a robot-agnostic architecture, the VLA Agent can be applied to both humanoid and mobile robots, enabling perception, reasoning and action capabilities. Unlike traditional service robots that rely mainly on manual operation, the VLA Agent allows robots to autonomously interpret user intent and contextual information through features such as wake words and gaze recognition.

The system also allows real-time monitoring of what the robot perceives, how it reasons and what action it executes. Camera footage and voice data are analyzed directly on the robot and immediately discarded without being stored. All processing occurs on-device, meeting the stringent security requirements of industrial environments.

During demonstrations, robots equipped with the VLA Agent accurately recognized user intent even in crowded environments. When a visitor makes eye contact, waves or calls “KT Robot,” the robot simultaneously recognizes the wake word and gaze and responds accordingly.

If a visitor says, “Please guide me to a window seat,” the robot evaluates factors such as group size and seat availability, calculates the optimal location and autonomously begins navigation. During movement, LiDAR sensors and depth cameras continuously scan the surroundings to avoid people and obstacles.

If information is insufficient, the robot proactively asks follow-up questions. For example, when a visitor says, “Please guide me to a seat,” the robot asks, “How many people are in your party?” If it detects a customer wearing a white shirt, it may even offer, “Would you like an apron?”

These capabilities demonstrate service intelligence that understands context and interacts proactively, rather than simply recognizing voice commands.

VLA-based Physical AI can be applied across multiple industries, including hospitality, retail and healthcare.

Autonomous Robot Collaboration Without Human Intervention

KT also demonstrated the Edge R2R (Robot-to-Robot) Agent, which operates in real customer environments.

The Edge R2R Agent performs three key functions:

• Providing integrated services across heterogeneous robots

• Orchestrating all agents and legacy systems on site

• Executing missions through real-time integration with the central platform

Visitors could observe how Physical AI operates through a smart automotive factory scenario.

A humanoid robot named Hugo, powered by VLA intelligence, inspects parts and detects abnormalities. It then determines that the next step requires item transportation. Once Hugo requests the task, the platform immediately calls the Warehouse Management System (WMS) to identify available logistics lines and assigns the transport mission to a mobile robot named Mobi.

Through Robot-to-Robot (R2R) collaboration and Agent-to-Agent (A2A) communication, the entire workflow is completed without centralized control or human intervention.

This demonstration highlights that the value of Physical AI lies not only in individual robot intelligence but in the platform’s agent-orchestration capability.

End-to-End Service Ecosystem: From Digital Orders to Physical Delivery

KT also introduced the K RaaS Order/Delivery Agent, a customer-facing service where AI agents collaborate to handle the entire process from ordering to robotic delivery.

When a user places an order via chat in a mobile app, the Order Agent analyzes the request and sends a delivery task to the platform. The platform assigns the most suitable robot.

The robot autonomously navigates by interacting with building systems such as elevators and security gates, while customers can monitor real-time order status and robot location. In this process, the platform, edge system, VLA intelligence and robots operate as a single ecosystem, completing the Physical AI workflow from digital order to physical delivery.

KT stressed that K RaaS is not simply a technology for connecting robots but a field-ready platform where AI understands and optimizes the physical world.

Oh Seung-pil, Executive Vice President and Head of Technology Innovation at KT, said:
“K RaaS learns from data generated in the field using neural-network-based models and continuously feeds those insights back into service improvement and operational optimization. As learning and execution iterate, performance improves in a virtuous cycle, enabling Physical AI to expand across industries such as manufacturing, logistics and smart buildings.”