Original Source
Development of 'Robot AI' That Sees and Learns Like Humans, Expected to Reduce Labor in Households and Logistics
Development of AI Robot Technology and Its Core Principles
A research team at the AI Robot Research Center of the Korea Institute of Machinery and Materials (KIMM) has developed a 'robot task AI' that autonomously performs repetitive everyday tasks by observing and learning human behavior. This technology can sequentially process complex tasks, such as organizing items and tidying tables, just like a human. The core lies in the robot learning by observing human task demonstrations, and to achieve this, 'task extraction technology' that converts human movements into precise data and 'task virtualization technology' that trains robots in a virtual space have been applied. Furthermore, a 'hierarchical task execution AI' that allows the robot to divide complex tasks into multiple steps, infer, and execute them autonomously, has been integrated.
High Success Rate and Broad Application Prospects
Unlike existing robot technologies that were limited to data or virtual environment verification, this AI is characterized by its ability to operate in the real world. As a result of applying and testing the technology on actual robot systems, the research team demonstrated a high success rate of over 90% in various work environments and versatility in flexibly adapting to changes in the surrounding environment. Kim Jeong-jung, head of the AI-powered Machinery Research Lab, explained that robots will reliably support repetitive daily tasks, easing the burden on humans. This technology is expected to be widely utilized in labor-deficient areas such as household and office space services, retail shelf arrangement, and logistics site item picking.
*Source: Machine News (2026-03-13)*
