Youngji Kim

Korea Advanced Institute of Science and Technology, Naver (South Korea)

Papers

Total Citations

H-Index

About

Youngji Kim is a robotics researcher whose work lies at the intersection of probabilistic state estimation, mapping, and deep learning for robot perception. Kim’s most impactful contribution addresses a foundational assumption in simultaneous localization and mapping (SLAM): the monotonic increase of pose uncertainty during exploration. Their 2017 paper, with 20 citations, rigorously demonstrates why uncertainty propagation on Lie groups preserves monotonicity, a critical insight that challenges and refines long-held beliefs in the SLAM community. Building on this, Kim developed a novel multitask multilayer Bayesian mapping framework (2022, 17 citations) that extends beyond metric-semantic maps to provide richer, scalable environmental representations for robots. In deep learning-based localization, Kim’s work on kidnap resolution achieved a 22% and 51% relative improvement over prior state-of-the-art position regression CNNs. Most recently, Kim introduced WayIL (2024), a method for image-based indoor localization using abstract wayfinding maps, tackling the challenge of geometric discrepancies between human-readable maps and robot sensor data. Through these contributions, Kim has advanced both the theoretical foundations and practical capabilities of robot mapping and localization.