DGIST research team wins Best Paper Award at IROS 2026

A research team led by Professor Ki-seop Kim at DGIST has received the Best Paper Award at the IEEE/RSJ IROS 2026, a premier international conference for intelligent robotics. Out of 1,933 submitted papers, the team's research was uniquely selected for the top honor.

The team developed ‘LT-Mem’, a new long-term memory system that allows robots to accumulate and remember the history of changes in object locations as they visit the same space multiple times. This addresses a major limitation in existing technology, where robots often lose past information when updating maps or fail to recognize that an object seen at different times is the same entity.

By combining precise spatial perception with systematic memory management, the system significantly reduces the data processing burden on artificial intelligence. Compared to traditional large language model methods that require re-interpreting vast video records, LT-Mem achieves higher reasoning performance while using only approximately one-sixteenth of the data tokens.