Robot work AI, capable of observing, learning, and performing routine tasks like a human, has been developed and is garnering attention as a technology set to change the paradigm of repetitive labor. As robots can flexibly handle simple yet labor-intensive tasks such as organizing goods or tidying tables, their scope of application is expected to expand beyond homes and offices to include retail and logistics sites.

The Korea Institute of Machinery and Materials (KIMM), under the National Science and Technology Research Council, announced on the 12th that it has developed a robot task AI that performs various daily tasks by learning from human work demonstrations. This technology is characterized by a design that extracts data from the process of human work, recreates the actual space as a virtual environment for learning and verification, and then enables the robot to perform the task step by step.

While existing robot technology relied on data limited to specific tasks or simple simulations, the newly developed system integrates the entire process, from data construction and virtualization to actual robot integration and real-world environment verification. Through this, it has secured versatility capable of reliably performing various daily tasks, such as object manipulation and cleanup.

In particular, a hierarchical task execution structure was applied to enable even complex tasks to be processed sequentially. The research team achieved a high success rate of over 90% by integrating different types of tasks into a single framework and confirmed the feasibility of field application by linking with actual robot systems. It is evaluated that the high adaptability to environmental changes allows for expansion into various service robot fields.

This robotic task AI can be utilized not only for service tasks in homes and offices but also in highly repetitive environments, such as organizing retail displays and picking and sorting in logistics centers. In the future, we plan to further expand the scope of operations and enhance the ability to respond to changes in space and objects to improve its performance in real-world usage environments.

Kim Jeong-joong, Head of the Artificial Intelligence and Machinery Research Lab at the Korea Institute of Machinery and Materials (KIMM), stated, “We have secured generalized work capabilities applicable to all aspects of daily tasks through a method of learning via human demonstrations and thinking step-by-step,” adding, “Robots can stably support repetitive tasks, thereby increasing work efficiency and reducing the burden on humans.”