😎 About me

I am a control engineer and researcher specializing in data-driven optimal control for robotic systems. My research focuses on the intersection of machine learning and model predictive control (MPC), particularly on developing frameworks that combine Neural ODE-based dynamics learning with nonlinear MPC to execute trajectory tracking and manipulation tasks.

Background: I initially earned a degree in mathematics with a focus on financial engineering, and worked in quantitative finance at KISpricing (2022–2023). However, I became increasingly interested in bridging the gap between theoretical guarantees and practical robustness through data-driven approaches. This led me to transition into data-driven control engineering, where I have found my true research passion.

Experience: I have extensive experience designing and implementing data-driven optimal controllers across diverse systems. In simulation environments, I have designed and validated controllers for mobile robots, thermal process control systems, and manipulator systems. On real physical systems, I have successfully controlled mobile robots and magnetic levitation systems, demonstrating the practical viability of my approaches.

Current Research Direction: Building on this experience, I am motivated by the challenge of identifying nonlinear dynamics efficiently from data collected across diverse physical systems—not limited to any specific robotic platform. My goal is to develop controllers that learn from minimal data while maintaining the theoretical rigor and stability guarantees required for real-world deployment.

Technical Approach: To achieve this, I leverage control-theoretic tools such as LMI-based design and Lyapunov stability theory to account for model mismatch inherent in learned dynamics. This ensures that controllers remain stable and robust even in the presence of learning errors and unmodeled phenomena.

Research Goal: Through this integrated approach, I am addressing two critical challenges simultaneously: the data efficiency limitations of traditional learning-based methods, and the control performance degradation caused by dynamics model mismatch. My research aims to develop controllers that are both data-efficient and theoretically guaranteed to perform reliably in real-world environments.

🤖 Experiments

I am a control and learning engineer working on data-driven optimal control for robotic systems, with a focus on integrating Neural ODE models into nonlinear MPC and validating performance through real-world experiments.