
Research
Three research areas — bimanual manipulation, biomedical robotics, and grippers — sharing one systems-level approach: mechanism, sensing, modeling, and control designed together.
Bimanual Manipulation Research Area 1
We design dual-arm robotic platforms that approach human-level manipulation capability. Our work spans mechanism synthesis for arms and wrists, kinematic and actuation co-design, and the real-time control architecture that runs them.
Dual-arm manipulationHumanoid platformsMechanism designCo-design
KAPEX: Co-Designed Dual-Arm Manipulator
Dual-arm manipulatorJoint-torque marginCo-design
Building a dual-arm robot that matches human-level motion characteristics is not a matter of selecting stronger actuators — the kinematic structure and the actuation system have to be designed together. KAPEX is a dual-arm manipulator developed under this co-design principle, with the joint arrangement chosen so that torque margins stay balanced across the arm’s working range. It serves as the lab’s testbed for bimanual manipulation.
Researchers
Sujin Lee
Youngjun LeeJi Won YoonSumin Lee
Effective Workspace Optimization for Robot Arms
Kinematic optimizationEffective workspaceShoulder design
Where a robot arm can reach and where it can actually work are not the same region. We optimize the shoulder base angle of a 7-DoF humanoid arm against torque efficiency, energy consumption and overload ratio together, rather than against geometric reach alone, and introduce effective workspace as the metric that captures the difference. Optimizing with a differential-evolution search expanded the effective workspace by 18.4 % on the KIST arm and 3.78 % on a Unitree G1 arm, giving a standardized way to compare manipulation efficiency across platforms.
- Enhancing Humanoid Robot Dynamics: An Optimization Framework for Shoulder Base Angle Adjustment
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2025 doi
ResearchersJi Won Yoon
Sujin Lee
Parallel Wrist for Confined-Space Manipulation
Parallel mechanism2-DoF wristCompact articulation
Serial wrist joints stack their actuators along the arm, which makes the distal link bulky exactly where compactness matters most. We are developing a two-degree-of-freedom parallel wrist for the KAPEX humanoid, placing the actuation proximally so that the distal structure stays slim while retaining the orientation range required for manipulation in confined spaces.
Researchers
Youngjun LeeJi Won Yoon
REIST: Reversed-Envelope Wrist
Wrist mechanismConfined-space manipulationRemote center of motion
A robot needs a larger operating volume than a human to perform the same task, and much of that excess comes from the wrist: as it reorients, the linkage sweeps outward. REIST is a 3-DOF wrist that reverses this motion envelope by realizing a virtually fixed wrist center, so the final kinematic link is compensated away and the assembly ends up shorter than the end-effector it carries. The prototype — 70 × 70 × 250 mm at 1.3 kg — uses spatial parallelograms and gimbals to achieve 19 mm of compensable length. On a dual-arm robot it produced a more compact envelope than a human wrist and measurably reduced unnecessary compensatory motion in the upper limb.
- REIST: A Novel Robotic Wrist With Reversed-Envelope Characteristics for Manipulation in Confined Spaces
IEEE/ASME Transactions on Mechatronics (Early Access), 2026 doi - Design and Analysis of a Robotic Wrist Mechanism with Bending Radius Compensation
파지점의 회전 반경을 보상하는 손목 메커니즘의 가동범위 분석
Transactions of the Korean Society of Mechanical Engineers A, 48(3), 199–203, 2024 doi
ResearchersYunseong Na
Sujin Lee
RAON-RT: Real-Time Control Architecture
Hard real-timeEtherCATReconfigurable architecture
A robotic platform is a networked system of many devices that must meet hard real-time deadlines together, yet research platforms are reconfigured constantly as new instruments and sensors are added. RAON-RT is a control architecture that keeps those two requirements compatible: devices are defined in software over EtherCAT and can be recomposed without rewriting the timing layer. The framework underlies the lab’s manipulation and surgical platforms.
- Design and Analysis of a Real-Time Control Architecture Towards Software-Defined EtherCAT Devices
Annual Conference of the IEEE Industrial Electronics Society (IECON), 2023 doi
ResearchersRaimarius Delgado
Woo Ryeol LeeGeonuk Kim
Biomedical Robotics Research Area 2
We build robotic systems for procedures at scales where unaided human motion reaches its limit — most notably supermicrosurgery, where vessels below 0.8 mm must be anastomosed. Our work covers the surgical platform itself, its end-effectors and sensing, the control architecture that runs it in real time, and the teleoperation interface through which surgeons work.
SupermicrosurgerySurgical roboticsTeleoperationReal-time control
PRISM: Multi-Surgeon Collaborative Supermicrosurgery Robot
SupermicrosurgeryMulti-surgeon collaborationWorkspace-uniform precision
Supermicrosurgery — anastomosing vessels below 0.8 mm — demands precision beyond what unaided human motion can sustain, and the procedure often requires more than one surgeon acting on the same field. PRISM is a robotic platform built for that setting: multiple surgeons collaborate intuitively on a shared operative field, and the mechanism is designed so that precision stays uniform across the workspace rather than degrading toward its edges.
- Control System Design for Stable Teleoperation of Supermicrosurgical Robot
초미세수술 로봇의 안정적인 원격조작을 위한 제어시스템 설계
Journal of Korea Robotics Society, 19(2), 169–177, 2024 doi
Researchers
Young Min Lee
Woo Ryeol Lee
Sang Hyuk Choi
Minchul KimJunwon LeeGeonuk Kim
Kinematic Calibration for Tool-Changeable Surgical Robots
Kinematic calibrationTool changePose error compensation
A surgical robot whose instruments are exchanged during a procedure loses its calibration every time a tool is swapped, and recalibrating the whole system mid-operation is not practical. We developed a hybrid calibration framework that separates the arm’s persistent kinematic error from the tool-dependent component, so that a replaced instrument can be brought back into registration without a full-system recalibration.
- A Flexible Hybrid Framework for Pose Error Calibration in Tool-Changeable Surgical Robots
IEEE Robotics and Automation Letters (under review)
Researchers
Woo Ryeol LeeSumin Lee
Bimanual Workspace Drift Compensation
Bimanual teleoperationMotion scalingPosture mismatch
In bimanual teleoperation the two arms are often driven under different motion scalings, and the operator’s posture gradually drifts away from the robot’s — until commanded and actual configurations no longer correspond. We compensate this drift at the workspace level, reducing posture mismatch between master and robot without interrupting the operator.
- Pose Tracking of Supermicrosurgical Robot Towards Multi-User Teleoperation
Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), 2023 doi
Researchers
Sang Hyuk ChoiGeonuk Kim
Joint-Limit-Aware Teleoperation Control
Joint limit avoidanceRedundancy resolutionIntuitive teleoperation
During teleoperation a redundant arm can drift toward its joint limits, and the operator only discovers this when the arm stops following. We control a 7-DoF arm so that position and orientation tracking is preserved while joint limits are respected, biasing the commanded velocity away from limits before they are reached. The operator keeps an intuitive mapping without having to reason about the arm’s internal configuration.
Researchers
Young Min Lee
BiStiff: Stiffness-Switching Surgical Handpiece
Compliant mechanismStiffness transitionSupermicrosurgery
A supermicrosurgical handpiece must be compliant enough to protect tissue at the moment of contact, yet stiff enough to transmit precise motion once engaged. BiStiff is a handpiece built around two compliant mechanisms whose stiffness switches at contact — triggered by the contact event itself rather than by an external command.
ResearchersChaewon Kim
Young Min Lee
FBG-Sensorized Surgical Forceps
FBG sensingForce measurementSurgical forceps
Supermicrosurgery gives the surgeon almost no force feedback: the tissue is too delicate and the tool too small for conventional sensing. We build sensorized forceps that measure grasping and manipulation forces at the tip using fiber Bragg gratings (FBGs), so that tissue-handling forces in robot-assisted supermicrosurgery become measurable and controllable.
ResearchersJunwon Lee
Robotic Electrode Implantation Platform
Electrode implantationCoordinate registrationPeripheral nerve
Implanting electrodes into neural tissue demands positioning accuracy at a scale where hand tremor dominates. We are developing a robotic manipulation platform for electrode implantation with integrated coordinate registration, so that a target defined in imaging coordinates can be reached precisely in the surgical field.
Researchers
Woo Ryeol Lee
Sang Hyuk Choi
Grippers Research Area 3
We develop grippers and the sensing that makes them reliable. Our interest is less in gripping force alone than in knowing the state of a grasp — whether it is stable, about to slip, or misaligned — and in adapting the gripper to objects whose shape and size are not known in advance.
Adaptive graspingTactile sensingGrasp state estimation
Friction-Cone-Based Grasp State Estimation
Tactile sensingGrasp state estimationShared autonomy
Reliable manipulation requires knowing whether a grasp will hold before the object slips. Using an arrayed 3-axis tactile sensor, we estimate contact shape by singular value decomposition, compensate shear force at each taxel, and classify every contact against its friction cone — aggregating the result into a continuous Slip–Stick Ratio. The classifier reaches 94.5 % accuracy on both planar and curved objects, and grip-force control driven by that ratio arrests incipient slip under external disturbance, so operator and controller can hold a grasp together in shared-autonomy tele-manipulation.
- Friction-Cone-Based Grasp State Estimation Using Arrayed Tactile Sensors for Shared Autonomy in Tele-Manipulation
IFAC World Congress, Busan (to appear)
Researchers
Woo Ryeol Lee
Gilrok Do
Shape- and Pose-Adaptive Gripper
Adaptive gripperSlider-crank mechanismApproach adaptability
Which gripper topology to build is usually settled after fabrication, by argument. We settled it before, by simulation: twelve candidate topologies were swept in a large-scale grasp simulation with quality-weighted metrics, which showed that a reconfigurable palm consistently outperforms a fixed one — and that the margin is negligible for objects small relative to the palm radius but decisive for large ones. The winning 6-DOF configuration, combining a prismatic palm with wrist and finger dexterity, was realized as a radial slider-crank mechanism rated to 500 g and validated against six geometric primitives.
- Enhancing Approach Adaptability in Robotic Grasping: Simulation-to-Experiment Validation of a Radial Slider-Crank Gripper
Intelligent Service Robotics (under review)
ResearchersGeonbeom Lee
Young Min Lee