Jangwon Kim

About

Jangwon Kim is a rising researcher in reinforcement learning and robotics, with a focus on bridging the gap between simulation and real-world control. His work addresses critical challenges in training stable, deployable agents, particularly for dynamic systems like quadrotors. In his highly cited 2025 paper, "Overcoming intermittent instability in reinforcement learning via gradient norm preservation," Kim introduces a novel method to stabilize training by maintaining gradient norms, directly tackling the common failure mode of catastrophic forgetting during learning. This contribution has already garnered 2 citations, signaling its immediate relevance to the RL community. Additionally, his work on "Transformer-based dynamics model for sim-to-real reinforcement learning control of a quadrotor with limited experimental data" (1 citation) demonstrates a practical approach to transferring policies from simulation to physical hardware with minimal real-world data, a key bottleneck in robotics. Kim’s research is notable for its dual focus on theoretical stability guarantees and empirical deployment, making his findings valuable for both algorithm designers and practitioners. As a young scholar, his early citations and innovative methods position him as a promising voice in the future of reinforcement learning and autonomous systems.

Research Focus

Key Achievements

Top Papers

1. 1
   Overcoming intermittent instability in reinforcement learning via gradient norm preservation

   2 citations · 2025
2. 2
   Transformer-based dynamics model for sim-to-real reinforcement learning control of a quadrotor with limited experimental data

   1 citations · 2025

Key Collaborators

Contact & Links