A robot artificial intelligence capable of performing everyday repetitive tasks similar to humans has been developed in South Korea.
The Korea Institute of Machinery and Materials (KIMM) AI Robot Research Institute announced on the 12th that a research team led by Kim Jung-joong, director of the AI Machinery Research Division, has developed a robot task AI called "RoGeTA" that performs everyday tasks such as organizing items, clearing tables, and manipulating objects.
The robot AI consists of task extraction technology that converts human work demonstrations into data, task virtualization (simulation) technology that recreates real spaces in virtual environments for learning and verification, and hierarchical task execution AI that enables robots to perform tasks step by step. Through this system, the robot learns from human demonstrations, undergoes repeated learning and verification in virtual environments, and can then execute complex tasks in stages.
For example, when commanded to "do the recycling" in a home, the robot—having been pre-trained in a virtual environment reflecting the actual interior structure of the house—remembers and moves to where trash is located. It then independently determines how to sort items by category and completes the recycling task.
Regarding this development, senior researcher Ko Du-yeol explained, "To enhance the performance of robot task AI, accurately capturing real human work data at the data collection stage is most important. To achieve this, we configured an interface that guarantees high degrees of freedom while enabling precise work data collection."
Existing robot task technologies often remained limited to single-task-focused datasets or simulation verification. In contrast, RoGeTA differentiates itself by comprehensively integrating the entire process: building datasets including various everyday tasks, real-space virtualization, hierarchical task execution AI, and actual robot system integration with real-environment verification.
With its high versatility and adaptability, RoGeTA is expected to replace labor-intensive and repetitive work across various fields—from homes and office spaces to retail store shelf organization and picking and sorting operations at logistics sites.
KIMM President Ryu Seok-hyun explained, "Robot performance is largely divided into mobility and manipulation capabilities, and RoGeTA excels in manipulation task capabilities." He emphasized the differentiation from Chinese humanoid technology, stating, "Conversely, even a robot with mobility advanced enough to perform kung fu cannot do household chores if it lacks manipulation capabilities."
President Ryu added, "Our goal is first for robots to possess capabilities equivalent to a Level 2 household manager certification so they can reduce domestic labor burdens. We will continue to advance our data with the target of RoGeTA reaching this level by 2030."