KIMM Fast Tracks AI System, Robotics Development

Korea Institute of Machinery and Materials (KIMM) has developed a new robot task artificial intelligence system designed to learn and perform everyday repetitive tasks by observing human demonstrations.

The technology, developed by a research team led by Dr. Jeong-Jung Kim at KIMM’s Research Institute of AI Robotics, enables robots to complete common activities such as organizing items, clearing tables and manipulating objects. The system is designed to automate labor-intensive routine work across environments including homes, offices, retail stores and logistics facilities.

At the core of the platform is a hierarchical task execution framework that allows robots to break complex jobs into sequential steps and carry them out systematically. The AI learns tasks by observing how humans perform them, converting those demonstrations into data that the robot can use to replicate the process.

The system integrates three key technologies: task extraction that converts human demonstrations into usable datasets; virtualized training environments that simulate real-world conditions; and hierarchical execution AI that allows robots to perform tasks reliably.

KIMM said the approach differs from many existing robot task systems, which often rely on single-task datasets or simulation-only testing. The new framework integrates the full development pipeline—from dataset construction to real-world testing using physical robots.

During testing, the system achieved success rates above 90% across multiple tasks, demonstrating strong reliability and adaptability even when environmental conditions change.

According to KIMM researchers, the technology could expand the role of service robots in sectors such as retail merchandising, warehouse logistics and general workplace support.

The research is part of KIMM’s RoGeTA (Robot General Task AI) framework program running from 2024 to 2029, which aims to develop robot intelligence capable of supporting a wide range of daily service tasks. The research team also plans to release task datasets and virtualized models of real environments to help accelerate development of future service robots.

The KIMM Research Team that Developed the Robot General Task AI Framework. Image courtesy Korea Institute of Machinery and Materials (KIMM)