Research

Quadruped Robot

Legged Robots & Humanoids — Our quadruped and humanoid team bridges mechanism design with dynamic locomotion control. We develop novel actuator concepts — including worm gear-based series elastic actuators (WSEA), parallel elastic actuators (PEA), and multi-layer high-torque-density actuators — that give legged robots the compliance and power density needed for real-world mobility. On the control side, we combine model predictive control with reinforcement learning in simulation (MuJoCo, Isaac Sim) and transfer to physical quadruped platforms. A unique research thread is our humanoid waist mechanism using parallel kinematics, enabling human-like trunk motion for future humanoid robots.

Ongoing Research Topic

Humanoid Waist Mechanism: Parallel mechanism design for human-like waist motion with reduced actuator torque requirements — includes kinematics analysis and optimization
WSEA (Worm gear-based Series Elastic Actuator): Compact, backdrive-safe actuator design for legged robots
PEA (Parallel Elastic Actuator): Energy-efficient actuator with passive compliance for dynamic locomotion
Quadruped Dynamics with Spine: Modeling and control of quadruped robots incorporating articulated spine joints for enhanced agility
Multi-Layer Actuator Development: Novel stacked actuator architecture for high torque density in compact form factors
Quadruped RL + MPC: MuJoCo and Isaac Sim-based reinforcement learning combined with model predictive control for locomotion
Hardware/Simulation Integration: Real robot platforms with Isaac Sim digital twin for sim-to-real transfer

Dual-arm Robot

The dual-arm robot team is dedicated to developing advanced robotic systems that replicate and refine expert manual skills for effective transfer to learners. Our research focuses on the design and real-time control of a 14-degree-of-freedom (DOF) plant that mimics human dexterity. We emphasize precise motion control through accurate impedance rendering in response to external forces and robust multi-task control using a hierarchical disturbance observer. Additionally, we aim to implement embodied intelligence in robots by employing sophisticated feature extraction algorithms.

Ongoing Research Topic

Development of 14-DOF dual arm robot & real-time control GUI
Accurate impedance rendering for external forces in precise motion control
Robust multi-task control employing a hierarchical disturbance observer
Implementation of embodied intelligence in robots through feature extraction algorithms

Intelligent Robot

The Intelligent Robotics Team is at the forefront of innovation and creative application in robotic technology. Our research encompasses the development and control of heterogeneous manipulators with multi-articular structures, achieving features previously unseen in traditional robots. We employ machine learning for precise dynamics modeling and conduct in-depth analysis and reproduction of intricate tasks performed by artists in both visual arts and music.

Ongoing Research Topic

Development and control of Heterogeneous manipulator with Multi-articular structure
Machine Learning for Dynamics modeling
Analysis and Reproduction of Art & Music
PPLM-DOB (Physics-informed Probabilistic Local Model DOB): A novel framework integrating GMM-based learned dynamics into the DOB nominal model, with TS fuzzy descriptor stability analysis — TII journal submission
CATALYST (ICRA 2026): Cognitive-to-Autonomy-inspired Two-Stage Training Data Generation with physics-informed excitation trajectory design for manipulator dynamics learning
EffiDynaMix: Efficient dynamics modeling using GMM/GMR with physically consistent data generation for robot inverse dynamics

Vehicle Control

The Vehicle Control Team focuses on pioneering control technologies to enhance vehicle performance and efficiency. Our research aims at maximizing regenerative energy in 4WD/4WS electrified vehicles by integrating connected and autonomous vehicle (CAV) technology with four-wheel drive and steering systems, thereby optimizing energy recovery during braking. To improve ride comfort by minimizing vibrations and shocks, we are developing an innovative active suspension system based on Series Elastic Actuator (SEA) technology, utilizing real vehicle data. Additionally, we are advancing a three-wheeled mobile robot equipped with precise lateral velocity control algorithms to enhance stability and maneuverability, grounded in vehicle dynamics principles.

Ongoing Research Topic

Development of Regenerative Energy Maximization Driving Technology for Electrical 4WD/4WS Vehicles Based on CAV Technology
Lateral Velocity Control of a Three-Wheeled Mobile Robot Based on Vehicle Dynamics
Comfort Enhancement Control of an Vehicle using a Series Elastic Actuator (SEA) Based Suspension System

Precision Control

The Precision Control Team is dedicated to developing advanced algorithms for the precise control of various systems. Our focus includes creating model-based control algorithms to achieve accurate frequency response, automatic parameter tuning for controller automation, and data-driven precision control that operates without requiring model information. Additionally, we are researching the application of Gaussian processes, a machine learning technique, to enhance various control methods. The control algorithms we develop are applied to systems such as the two-inertia system, linear motor, and hybrid gantry stage, ensuring high precision and performance.

Ongoing Research Topic

Data-driven optimization of model-based control with optimal instrumental variable
Disturbance rejection with multi-dimensional Gaussian process regression
Frequency Response Function of Linear Parameter Varying System
Data-enabled Predictive Control

Page updated

Google Sites

Report abuse