Career History

Thesis: Accelerating Policy Learning for Robust Control of Robotic Manipulators and Aerial Vehicles via Physics-Informed Guidance

​Supervisor: Professor Dong Eui Chang

Experience Summary: Recipient of KAIST scholarship. Throughout my MS program, I also worked as a graduate research assistant at Control Lab. I specialized in the intersection of modern control theory and learning-based algorithms for robotic systems. ​My work spanned theoretical studies and designs to real-world hardware deployment on manipulators, drones, and mobile robots.

​

Primary Research: 

• Marine-Manipulator Control: I was an integral part of a research project involving robust and safety-constrained control of a marine-manipulator system under adverse sea conditions. I played key roles in all major aspects of the project since the inception, notable contributions include complete system identification, digital twin development, sim2real gap quantification and mitigation, control system design, RL training/optimization and highly optimized hardware implementation on Jetson Orin+STM32.

• Robust Drone Control: I performed independent research on robust drone control that combines model-based and geometric control strategies with reinforcement learning algorithms for robust control under heavy disturbances.

• Hybrid Control: I also did collaborative research on robot independent hybrid control strategies utilizing model-based approaches to improve learning-based solutions for sample efficiency in  RL (model-based baselines, approximate analytic solutions, residual RL), parameter estimation and tuning (using RL, differentiable simulations, and observer designs), and on reducing sim-to-real gap for RL policies (robust simulation design, disturbance observers, L1-adaptation, and offline fine-tuning approaches). 

​

Technical Experience:

• Control Theory: Nonlinear systems (refer Khalil Nonlinear Systems/Control), optimal control (LQR, MPC, Krener NST), and geometric control on manifolds.

• Learning and AI: Deep RL, Foundation models (VLAs, LLMs), World models, and Neural ODEs for system identification.

• Differentiable Simulations: Nvidia Warp/Newton, JAX and Pytorch for differentiable simulation based policy learning and system identification.

• Manipulator Control: Control strategies for manipulators including FK-IK, trajectory design, observer design, and system modeling from scratch.

• Hardware Implementations: High-performance control loops in C/C++, TensorRT and CUDA implementations on edge devices like Jetson for RL policies.

• Simulation/RL Tools: MATLAB/Simulink, IsaacSim, RPG-Flightmare, Gymnasium, SB3.