LG CNS has unveiled "PhysicalWorks," a robot transformation (RX) platform that integrates the entire process of robot deployment — from training to operation — into a single management system. It marks the first time a Korean company has launched an end-to-end platform covering robot training through operation under its own brand.
LG CNS made the announcement at an "RX Media Day" held on the 7th in Magok, Seoul. At the event, the company introduced the two core platforms of PhysicalWorks: "PhysicalWorks Forge" and "PhysicalWorks Baton."
PhysicalWorks Forge is a platform that handles the full process — from collecting data needed for robot training, to validating robots, to deploying them on-site — in one place. Moving beyond the conventional method of having robots repeatedly mimic human movements thousands of times, the platform boosts efficiency by using simulation data that recreates real worksites and tasks in a three-dimensional (3D) virtual environment. The company plans to additionally adopt methods that convert videos of human work into training data or utilize motion capture.
Artificial intelligence (AI) automatically selects, organizes, and processes data useful for robot training. For example, during the training process in which a robot learns to pick up products in a factory, the AI selects only successful motions or removes unnecessary segments to enhance training efficiency. Once trained, robots have their task feasibility and stability verified through simulation in a 3D virtual environment, then are optimized for actual sites before deployment. According to LG CNS, this can shorten the time required to train a robot and deploy it on-site from several months to one or two months.
PhysicalWorks Baton is a platform that assigns tasks to various types of robots — including bipedal, quadrupedal, and wheeled models — and integrates their control and monitoring. Its key feature is the ability to operate robots from different manufacturers and of various forms within a single system. It addresses the problem of management becoming increasingly complex as robot types grow, since each manufacturer uses different control methods and operating interfaces. Using mathematical optimization technology, it automatically distributes tasks among robots and optimizes movement paths to prevent collisions.
Agentic AI adjusts workflows in real time by reflecting changes in task progress and equipment status. If a conveyor belt stops, logistics routes are automatically reconfigured; if a specific robot halts, its task is instantly transferred to another robot. The platform has also expanded the scope of automation so that robots can naturally perform non-standard tasks mid-process — tasks that were difficult to manage with conventional manufacturing execution systems (MES). When PhysicalWorks Baton is applied to a robot operating environment of around 100 units, including autonomous mobile robots (AMRs) and automated guided vehicles (AGVs), productivity is expected to improve by more than 15% and operating costs to be cut by up to 18%.
"Different Manufacturers, One Operation" — LG CNS Unveils Robot Platform
LG CNS has built a virtuous cycle connecting training, validation, and operation by linking PhysicalWorks Forge and PhysicalWorks Baton. Robots trained and validated on PhysicalWorks Forge are operated and controlled through PhysicalWorks Baton, and the data accumulated in the process is fed back into training. PhysicalWorks Forge is currently running robot proof-of-concept (PoC) projects with more than 20 client companies. PhysicalWorks Baton is being used to integrate the monitoring of four types of robots — patrol, barista, luggage-carrying, and cleaning — in the Busan Smart City national pilot project.
"The essence of RX lies not simply in acquiring robots, but in how quickly robots can be settled on-site to work and how sustainably they can operate in changing environments," LG CNS President Hyun Shin-kyun said. "Based on our full-stack capabilities — from establishing robot deployment strategies optimized for customer sites, to securing industry-specific robot foundation models (RFM), and covering robot training, application, and operation — we will create a new standard for the commercialization of physical AI and ultimately realize a robot-centered autonomous operating system."
