Ferromagnetic Surfaces with a Quadrupedal Robot

Agile and versatile climbing on ferromagnetic surfaces with a quadrupedal robot
​[Science Robotics'22, Cover Paper] :
This paper introduces MARVEL, an untethered quadrupedal climbing robot designed for agile locomotion on floors, walls, and ceilings in ferromagnetic environments. Its key hardware innovations include electropermanent magnet feet combined with magnetorheological elastomers for strong adhesion and traction, and torque-controlled actuators. These features enable high-speed climbing, versatile gaits, and transitions across complex surfaces and obstacles. 

Dual Reduction Ratio Planetary Drive

DRPD, Dual Reduction Ratio Planetary Drive for Articulated Robot Actuators [IROS'22] : 
This paper proposes a robot actuator with a switchable reduction mechanism based on a 3K compound planetary drive. By fixing either the carrier or the ring gear using a pawl brake mechanism with cams and micro servos, the actuator achieves two reduction ratios: 6.91 and 44.93. The prototype demonstrates reliable gear shifting between low and high reduction modes, enabling a wider operational range.

Design of KAIST HOUND, a Quadruped Robot Platform

Design of KAIST HOUND, a Quadruped Robot Platform for Fast and Efficient Locomotion with Mixed-Integer Nonlinear Optimization of a Gear Train [ICRA'22] :
This paper presents a design method for an efficient and agile quadruped robot using mixed-integer optimization, which includes the number of gear teeth to determine the optimal gear ratio for a 3 m/s running-trot. The approach also considers thermal dissipation in the motor controller to account for heat generation during high-speed motion. Based on the optimized design, the 45 kg KAIST Hound robot was developed and successfully demonstrated stable 3 m/s running using a nonlinear model predictive controller (NMPC). Its robustness was further validated through experiments such as climbing a 22° slope, walking 3.2 km, and traversing a 35 cm obstacle.

Optimized Actuator Design

Torque Control Actuator Design [IROS'17, Best Student Paper Finalist] : 
A general goal of our legged robots is to obtain agile and aggressive motion, which requires high GRF(Ground Reaction Force) and fast speed. This requirement has to be met under a appropriately selected gear ratio and motor specification. We have dealt with this problem as nonlinear programming for the optimized performance of the robot.

Integrating System

Robots are always in real world, so our laboratory also do the experiments of all the theories we've developed. For this, we make our own hardware platform from the scratch, of course. Our researchers have naturally obtained various technical skills for making the hardware, which cannot be easily learned in papers.