[기획]"사람은 퇴근하세요" 인간 공존 시대 연 휴머노이드 근황_산업뉴스 [산업방송 채널i]

[Anchor Commentary] Industrial Broadcasting has prepared a series of special reports in collaboration with the Korea Electronics and Technology Institute (ETRI) to examine the innovative technologies that will change the future of our industry. For the first installment, we visited the "Gangnam Robot Test Field," which has emerged as the central axis of Korea's robot ecosystem as advanced robots hone their practical capabilities. We covered the current state of our robot technology, including humanoid robots that resemble humans, poised to lead the global market. Reporter Baek Ga-hye reports. [Report] A robot hand picks up logistics items and places them into boxes. Although it looks like a common robotic arm, research is currently in full swing on "humanoid robots" that resemble humans—seeing with their eyes and thinking with their heads. While robots of the past performed only simple, repetitive tasks following pre-programmed trajectories, we have now entered an era of intelligence where they independently assess their surroundings. Robots have reached a stage where they learn data on their own to engage in thinking akin to that of the human brain. The Korea Electronics and Technology Institute is conducting advanced research to transplant large language models (LLMs) and AI-based visual technologies into robots. Core software control programs designed to enable robots to perceive situations like humans, going beyond simple machines, are being developed here and distributed to actual robot companies. On-site response and satisfaction are also very high. [Interview – Hwang Jeong-hoon / Head of Intelligent Robotics Research Center, Korea Electronics & Technology Institute] Recently, there has been significant research focused on integrating humanoid robots into logistics or manufacturing to perform tasks like humans. We are also conducting such research. People often refer to this as "Robot Foundation Models" these days; we are creating these models by combining AI that allows robots to visualize data from LLMs with behavioral intelligence. The greatest strength of humanoid robots is flexibility. This is because they can be immediately deployed into spaces where human workers previously worked, enabling simultaneous collaboration, without completely overhauling existing factory infrastructure or work structures. To implement precise movements, sensor cameras mounted on the robot's wrists and head perceive the surroundings in real time. This involves perfectly converting the changes in joint angles and the flow of gaze that occur when a person intentionally extends their hand into control signals for the robot. Ultimately, for robots to work as precisely as humans, the quality of motion data and the control technology that seamlessly transmits it in real time are key. [Interview – Hwang Yu-hyun / Senior Researcher, Korea Electronics & Technology Institute] In the case of humanoid robots, I believe the advantage is that they can be deployed directly into spaces where humans currently work. The robot's joint information and gaze are located at its wrists and head. Based on that gaze and the robot's joint information, it recognizes, "I need to perform this action in this situation," and operates accordingly. When collecting data, human intent is important, but sending that intent to the robot as a control signal in real time is crucial. This AI collaborative intelligence technology is seamlessly integrating not only into logistics but also into traditional manufacturing sites. Robots are inheriting the know-how accumulated over decades by field artisans, opening a new breakthrough for foundational industries facing increasingly severe labor shortages. This technology, known as imitation learning, goes beyond simply replicating the trajectory of a worker's movements. It precisely detects even the minute pressure and resistance coming from the artisan's fingertips while they are working. Research is currently underway to train robots with this core torque data, enabling them to respond flexibly on their own even when unexpected situations occur. [Interview – Cho Chang-no / Senior Researcher, Korea Electronics & Technology Institute] We are measuring force and torque at the end of our robots. By utilizing this information, we can identify problems that arise during operations. We are developing technology that allows us to respond once we identify the issue. We are collecting work data, motion data, and force data from skilled artisans, processing that data, and training the robots. Through this, we are researching ways for the robots to accurately reflect the movements of the artisans. This is a robot that inspects the manufacturing process of "ferrite magnets," the black magnets commonly used in our daily lives. It uses AI and 3D depth sensors to accurately identify minute cracks, twisting, and surface defects in products that skilled workers previously had to manually spot with t...