[E-Daily Reporter Kim Hyun-ah ] Artificial intelligence (AI), which was previously used to detect product defects, is expanding into the realm of robot motion control. SAIGE, which has been developing AI for quality inspection in manufacturing facilities, is joining the “ Doosan Robotics(454910) ” consortium to develop next-generation collaborative robot technology based on domestically produced AI semiconductors. By extending its existing vision AI technology to robot task planning and motion control, the company is launching its physical AI business in earnest.

SAIGE announced that it has been selected as a participating company in the “Development of Next-Generation Collaborative Robot Technology Based on On-Device AI Semiconductors” project, part of the “K-On-Device AI Semiconductor Technology Development Project” supported by the Ministry of Trade, Industry and Energy and the Korea Institute for Advancement of Technology (KEIT).

Doosan Robotics is leading this project. The goal is to develop a system in which robots independently perceive their surroundings, make decisions, and move without relying on external PCs or the cloud. The plan is to develop a domestically produced AI system-on-chip (SoC) that combines a neural processing unit (NPU) with real-time control capabilities and to commercialize a collaborative robot system equipped with this technology by 2031. The project duration is 54 months.

The consortium includes Doosan Robotics and Sage, as well as Aidin Robotics, Mobilint, the Korea Electronics Technology Institute (KETI), the Korea Institute of Industrial Technology (KITECH), and universities such as Sejong University, Korea University, Yonsei University, and Dongguk University. Doosan Robotics and Aidin Robotics will develop the robot body and grippers, while Mobilint and KETI will develop the AI semiconductors. Sage, KITECH, and the participating universities will be responsible for the AI software to be installed on the robots.

Sage is developing software for task planning, motion generation, and real-time control of collaborative robots, which runs directly on the AI semiconductor. Its role is to determine what tasks the robot will perform and how, and to implement the process of actually moving the robot’s arm and hand.

The team plans to reduce the computational load and optimize execution performance by ensuring the software runs directly on the robot’s internal AI semiconductor, without relying on external servers. In tasks such as picking up objects or assembling parts, control delays can affect accuracy and safety; therefore, technology that operates stably even with limited computational resources is crucial.

The developed technology will be demonstrated in logistics and manufacturing settings. The company plans to implement an autonomous operation system in which the robot recognizes the target object, determines the work conditions, and, if an anomaly occurs during the task, modifies the plan to continue the operation. The goal is to create robots that do not merely repeat predetermined movements but adjust their operations according to on-site conditions.

What kind of company is SAIGE? SAIGE

was founded in 2017 by a research team from Seoul National University’s Robotics and Automation Laboratory. The company has been providing AI-based machine vision solutions to manufacturing sites—such as those for rechargeable batteries and semiconductors—to detect product defects and identify process anomalies. Its flagship product, “SAIGE VISION,” supports the automation of quality inspections using deep learning-based image analysis technology.

Co-founder and CEO Park Jong-woo is a professor in the Department of Mechanical Engineering at Seoul National University and a robotics engineer who has conducted research in the fields of robot motion control and optimization. He previously served as editor-in-chief of “IEEE Transactions on Robotics,” the robotics journal of the Institute of Electrical and Electronics Engineers (IEEE). CEO Park leads the company alongside co-CEO Hong Young-seok under a co-CEO system.

This project is significant in that it extends the AI technology Sage has accumulated in manufacturing settings to the actual movement and control of robots. It broadens the company’s business scope beyond simply detecting product defects using camera footage to include technology that enables robots to perceive their surroundings, plan tasks, and execute them.

Last December, Sage signed a memorandum of understanding with RS Automation Co., Ltd., an industrial motion control company, to jointly develop a physical AI-based robotics and automation platform. With its participation in this government-funded project, the company is expected to accelerate its efforts to integrate vision AI and motion control technologies.

Park Jong-woo, CEO of Sage, stated, “Through this project, we will develop AI for robot motion generation and control that runs directly on domestically produced AI semiconductors, and we will expand the industrial field experience we have accumulated through vision AI into the robotics domain to establish Sage as a physical AI company.”

Sage is also expanding its vision AI business from the manufacturing sector into the public and infrastructure sectors. In June of this year, the company was selected as a “Global ICT Future Unicorn” and for the Ministry of SMEs and Startups’ “Unicorn Bridge” program, and it signed agreements with Incheon International Airport Corporation and the Korea Agency for Infrastructure Safety Management in the field of AI-based safety monitoring.

The success of this project hinges on securing operational accuracy and real-time control performance in actual industrial settings. Whether Sage can demonstrate that its robots operate stably even with limited computational resources—and whether these robots can be successfully applied to manufacturing and logistics sites—is expected to determine the expansion of Sage’s physical AI business.