The Korea Electrotechnology Research Institute (KERI) has developed autonomous manufacturing technology in which AI interprets manufacturing process instructions given via human voice and links them to robot movements. This technology aims to reduce the burden of existing automation methods, where engineers had to rewrite code every time the process changed.

The Korea Electrotechnology Research Institute (KERI) announced that a team led by Dr. Lee Ju-kyung at the Artificial Intelligence Research Center has developed "autonomous manufacturing multi-agent AI" technology in collaboration with Changwon National University. This research was conducted as part of the government's "Glocal University 30" project and KERI's own research project, "Development of Original AI-Robot Convergence Technologies for the Implementation of Agentic Self-driving Lab."

In manufacturing sites, robot operations and process conditions must be reset whenever a product or part changes. Since conventional automation robots operate according to pre-programmed rules and codes, specialized personnel had to intervene whenever the environment changed. For small and medium-sized manufacturing companies in regions lacking skilled labor, this process resulted in a burden of time and cost.

To solve this problem, the research team applied an agentic AI architecture based on Large Language Models (LLM) to the manufacturing process. When a user gives instructions in natural language, the language AI identifies the intent, and the vision AI analyzes the location and state of objects using a camera. Subsequently, the robot control AI converts this into an actual operation scenario and performs the process.

In particular, this technology focuses on connecting verbal instructions with actual workspace coordinates and object information. For instance, when a specific part is instructed, the AI identifies the target and its location based on camera data and converts it into a work procedure that the robot can perform.

KERI explained that utilizing this technology can reduce the process reset time, which previously took several days to a week, to within one hour. It is evaluated as having potential for use by small and medium-sized manufacturing companies engaged in multi-product, small-batch production, as it can also lower the burden of software modifications resulting from process changes.

Dr. Lee Ju-kyung of KERI stated, “Local small and medium-sized enterprises (SMEs) and middle-sized companies are struggling to adopt AI due to cost and manpower issues,” adding, “We will promote technology transfer by developing a technology that can intelligentize existing manufacturing lines without significant cost.” Moving forward, KERI and Changwon National University plan to cultivate the students who participated in this research into AI convergence talents capable of being deployed to local industrial sites.