Skip to main content

Research Vision

Mission

My mission is to bridge the gap between rigorous control theory and modern robot learning, creating embodied AI agents that are not only intelligent and adaptable but fundamentally safe and reliable in complex, unstructured environments.

Core Focus Areas

Embodied AI & Robotics

Designing intelligent robotic systems that can perceive, learn, and adapt to complex, unstructured environments through advanced learning and simulation.

VLAVLMRLSim2Real

Advanced Control Theory

Applying rigorous control methodologies to guarantee system performance, stability, and absolute safety under severe disturbances and model uncertainties.

Safety ControlRobust ControlOptimal ControlData-Driven Control

Roadmap

Why am I learning these things? This is the path to building true physical intelligence.

Classic Control

solidified

Foundations of frequency domain analysis, PID controllers, and classical filters.

Modern Control

solidified

State-space representations, State Feedback (FB), LQR, Observers, and Kalman Filtering.

Robust & Nonlinear Control

solidified

Sliding mode control and Lyapunov stability theory for managing bounded uncertainties.

Multi-Robot Control

solidified

Formation control and distributed consensus algorithms for multi-agent systems.

Safety-Critical & Optimal Control

solidified

Model Predictive Control (MPC) and Control Barrier Functions (CBF) for mathematical safety guarantees.

Data-Driven Control

current

Bridging theory and data via System Identification, Koopman operators, and data-driven dynamics.

Robot Learning

current

Deep Reinforcement Learning (RL), Imitation Learning, and Sim2Real transfer techniques.

Foundation Models

upcoming

Exploring Vision-Language Models (VLM) and Vision-Language-Action (VLA) architectures for semantic reasoning.

Advanced Architectures

upcoming

Next-generation representations including World Models and Joint Embedding Predictive Architectures (JEPA).

Physical Intelligence

ultimate goal

Embodied AGI—general-purpose robots capable of robust physical interaction in open-world environments.

Publications

Safety-Critical Formation Tracking Control of Multi-Robot Systems via CLF-CBF-QP

2025 International Conference on Advanced Mechatronic Systems (ICAMechS)

This paper proposes a novel Hierarchical Fallback CLF-CBF-QP framework for safety-critical formation tracking. The approach employs CLF for tracking and High-Order CBFs for safety, featuring a three-stage fallback strategy that resolves QP infeasibility by prioritizing safety over performance while maintaining computational efficiency.

IEEE XploreCode (Coming Soon)

Distributed Robust Time-Varying Formation Control of Multi-Agent Systems Under Disturbances

2024 SICE Festival with Annual Conference (SICE FES)

This work considers the problem of time-varying formation tracking control of second-order multi-agent systems under disturbances. A distributed robust time-varying formation control law is proposed including distributed finite-time estimators of the leader states and sliding mode time-varying formation controllers.

Talks

SICE Festival with Annual Conference (SICE FES) 2024

Distributed Robust Time-Varying Formation Control of Multi-Agent Systems Under Disturbances

Presented our work on distributed finite-time estimators and sliding mode controllers for multi-robot formation tracking.