Cheolhyeon Kwon

Research profile

Research summary

Cheolhyeon Kwon is a researcher at Ulsan National Institute of Science and Technology (UNIST), South Korea, whose bibliometric profile spans control theory and, most recently, field robotics. The OpenAlex-linked topic record (h-index 13, i10-index 19) is anchored in stability and control of uncertain systems (18 works) and smart grid security and resilience (18), with distributed control of multi-agent systems (14), autonomous vehicle technology and safety (12), and adaptive control of nonlinear systems (11) rounding out the corpus. Evolve's current field tagging splits evenly (0.5/0.5) between field robotics/robot learning and robotics (autonomous racing, motion prediction), reflecting a recent research pivot: two 2026 arXiv papers reviewed in the catalog apply uncertainty-aware machine learning to robot problems — kernel-metric-regularized deep kernel learning for opponent trajectory prediction in 1/10-scale autonomous racing, and generative-replay continual learning for traversability prediction on off-road robots. Both papers place Kwon in the last author position. Productivity is high and rising: 24 works in the recent three-year window and a record 13 in 2026, on a continuous publishing record since 2012. The collaboration network is compact (5 distinct recent coauthors). Evolve scores activity at 94 and reproducibility at 55; output and impact are not yet scored because citation and publication totals were not supplied in this snapshot.

Research keywords

Highlights

• Uncertainty-aware opponent trajectory prediction in 1/10-scale autonomous racing

  A kernel-metric-regularized deep kernel learning model predicts opponent vehicle trajectories with uncertainty quantification, validated on 1/10-scale racing hardware. Evolve's verdict: strong hardware results, but statistical reporting is thin; reproducibility graded medium.

• Generative-replay continual learning curbs forgetting in traversability prediction

  A generative-replay continual learning approach cuts catastrophic forgetting in off-road traversability prediction without storing raw data, with plausible real-robot gains; Evolve notes navigation trials of n=3 and unreported λ/τ hyperparameters. Reproducibility: medium.

• Control-and-security foundation behind the robotics pivot

  The historical topic record — stability and control of uncertain systems (18 works), smart grid security and resilience (18), distributed control of multi-agent systems (14), autonomous vehicle technology and safety (12), and adaptive control of nonlinear systems (11) — supplies the systems-and-control grounding for the recent uncertainty-aware robotics line.

Timeline

1. 2012First indexed work appears: countsByYear records a single publication in 2012, the start of a continuous 15-year output record.
2. 2018Annual output reaches 7 works, the highest single-year total recorded to that point.
3. 20228 works published, tied with 2024 for the second-highest annual count in the record.
4. 2024Another 8-work year, matching 2022 and setting up the 2026 surge.
5. 2026Record year with 13 works, including both Evolve-reviewed arXiv papers: opponent trajectory prediction in autonomous racing and continual learning for traversability prediction.

Papers in the Evolve catalog

• 2026-09-15

  Kernel-Based Metrics Learning for Uncertain Opponent Vehicle Trajectory Prediction in Autonomous Racing

• 2026-09-15

  Continual Learning for Traversability Prediction with Uncertainty-Aware Adaptation

Frequent co-authors