Researchers at Seoul National University will develop robots that sense touch like humans and artificial intelligence that erases unnecessary memories on its own.
The university's College of Engineering said on the 14th that two research projects — one led jointly by professors Hong Yong-taek and Choi Jong-hyun of the Department of Electrical and Computer Engineering, and another by professor Jeon Se-young — were selected for Samsung's Future Technology Fostering Program for the second half of 2026.
The team led by Hong and Choi will develop robot AI that touches and senses objects and adjusts its own actions, much as a person does. It was the only project chosen in the program's next-generation robotics category.
Existing robots face a limitation when gripping or assembling objects: the contact surface between the hand and the object is hidden from cameras. Such sensory blind spots make it difficult to detect slight slippage or temperature changes, leading to task failures or damaged components.
To overcome this, the team will develop a high-resolution electronic skin that deforms flexibly like human skin while sensing pressure, temperature and heat flow. The researchers will also study technology that combines tactile data from the electronic skin with camera images and language information so robots can judge their own actions — easing their grip when an object slips, or releasing a component that proves too hot.
The team further plans to verify whether the system can respond appropriately to objects it has not been trained on and to manufacturing environments that are difficult to predict. The goal is to establish a foundation for intelligence technology applicable not only to industrial robots but also to humanoid robots.
The team led by Jeon will develop selective forgetting AI that erases unnecessary memories. Existing AI focuses on retaining as much learned information as possible, which means important clues can be buried as data accumulates, and existing capabilities can degrade during the learning process.
The team will develop technology that lets AI judge for itself how likely information is to be used later, keeping what it needs and deleting what it does not. The technology is expected to lay the groundwork for AI assistants that maintain context over long tasks and for on-device AI that runs on less memory and power.