Sky Intelligence is embarking on joint research into robot vision and grasping technologies with the Seoul National University AI Research Institute, thereby accelerating the acquisition of physical AI technologies based on digital twins and synthetic data. This collaboration is significant given that data capable of learning diverse environments and simulation technologies are required for robots to recognize visual information and translate it into actual movements.

Sky Intelligence announced on the 4th that it has signed a Memorandum of Understanding (MOU) with the Seoul National University AI Institute (AIIS) to establish a framework for joint research and cooperation aimed at advancing robot vision and grasping technologies. Both parties plan to jointly conduct research on robot perception and manipulation technologies, pursue national R&D projects, exchange research personnel, and cooperate on securing AI learning and reasoning infrastructure.

This collaboration was pursued as securing real-world environmental data emerged as a key challenge in the implementation of physical AI. Due to the diverse working environments and target objects of manufacturing, logistics, and service robots, it is difficult to train them for all situations using only real-world data. Accordingly, the importance of synthetic data and simulation technology is growing as they generate various conditions in virtual environments and utilize them for actual robot learning.

Both parties are conducting joint research centered on vision technology, which enables robots to analyze visual information obtained through sensors such as cameras and translate it into actions, and grasping technology, which estimates the position and orientation of objects to stably pick them up. The research goal is focused on improving robot grasping accuracy and stability to a level applicable in actual industrial settings.

Sky Intelligence plans to provide NVIDIA Omniverse-based digital twin technology and industrial synthetic data generation pipelines for joint research. This method converts real-world objects and environments into high-precision digital assets and generates training data in a virtual environment that reflects the laws of physics. Through this, the company plans to support high-precision 3D CAD models, training and evaluation datasets, and data annotation.

The Seoul National University AI Research Institute will participate in the verification and advancement of robot vision and grasping technologies, leveraging its research capabilities and human network. Both parties will continue research cooperation through regular seminars and technical meetings, and are also considering the joint implementation of government R&D projects.

This collaboration can be seen as an attempt to reduce technical challenges in the commercialization of physical AI by combining the university's AI research capabilities with the company's digital twin and synthetic data technologies. It is expected that if the results of this joint research lead to actual application in industrial settings, it will also have an impact on expanding the domestic robotics AI ecosystem.