Mechanical Intelligence

We are conducting research on mechanical intelligence system. By leveraging structurally embedded intelligence, we implement diverse functionalities without the need for active sensing or actuation. Our work involves both utilizing naturally embedded intelligence inherent in nature and developing mechanically programmable systems that design intelligence through various mechanical components.


1. Soft Yet Powerful Artificial Muscle – Twisted String Actuator (TSA)

Researchers: HyeonSeok Seong, JaeHyung Jang, Seung-Joon Baek, Euibin Shin, Juyeong Seo.

We introduce the Twisted String Actuator (TSA), a powerful and compact soft artificial muscle that mimics the naturally embedded intelligence of biological muscles. By twisting motor-driven strings, the TSA acts as a highly efficient linear transmission that generates immense contraction force. It perfectly serves as an artificial muscle by offering a high transmission ratio, high efficiency, low cost, and impact durability. Furthermore, its inherently flexible and backdrivable nature allows the system to safely absorb unexpected shocks, ensuring seamless human-robot interaction in collaborative environments.

1.1. Modeling/Control/Application of TSAs

Researchers: HyeonSeok Seong, JaeHyung Jang, Seungjoon Baek, Euibin Shin, Juyeong Seo.

Our lab pioneers comprehensive research encompassing the modeling, control, and application of TSAs. To unlock their full potential, we develop high-fidelity dynamic models to overcome inherent nonlinearities like varying stiffness, friction, and hysteresis. These models enable advanced control algorithms, including energy-preserving, adaptive tension, and pioneering sensorless strategies. In parallel, we enhance core mechanics through innovative designs. We enable single-motor bidirectional actuation using variable-radius pulleys, and we achieve transmission ratio (TR) linearization and stroke amplification using dual variable-radius drums. Broadening our scope in mechanical design, we also develop advanced passive and active continuously variable transmissions (CVTs). These integrated advancements in theory and mechanics drive exciting real-world applications, successfully resulting in high-force haptic displays and flexible wearable systems with ultimate precision and robustness.

Selected Papers

  • Gaponov, I., Popov, D., & Ryu, J. H. (2013). Twisted string actuation systems: A study of the mathematical model and a comparison of twisted strings. IEEE/ASME Transactions on mechatronics, 19(4), 1331-1342. Link
  • Jang, J., Song, Y. U., & Ryu, J. H. (2022). Active-type continuously variable transmission system based on a twisted string actuator. IEEE Robotics and Automation Letters, 7(2), 2605-2612. Link
  • Baek, S., & Ryu, J. H. (2023). Tension control of twisted string actuators in variation of stiffness and original length of strings. IEEE/ASME Transactions on Mechatronics, 28(5), 2483-2494. Link
  • Seong, H. S., Zhakypov, Z., Okamura, A., & Ryu, J. H. (2023, October). A TSA-based haptic display with a rhombus structure for rendering linearized high force. In 2023 23rd International Conference on Control, Automation and Systems (ICCAS) (pp. 792-795). IEEE. Link
  • Baek, S., Jang, J., & Ryu, J. H. (2024). Enhancing maximum stroke of twisted string actuators by adjusting twisting ratio. IEEE Robotics and Automation Letters, 9(6), 5887-5894. Link
  • Seo, J., Jang, J., Baek, S., & Ryu, J. H. (2026). Dual Variable-Radius Drum for Transmission Ratio Linearization and Stroke Enhancement in Twisted String Actuators. IEEE Robotics and Automation Letters. Link

1.2. Wearable Robots

Researchers: HyeonSeok Seong

A critical bottleneck in the real-world deployment of wearable robotics lies in the physical constraints of conventional actuators, which struggle to deliver high-force assistance without compromising user comfort and portability. We overcome this fundamental limitation by employing the TSA, a soft actuation technology that perfectly fulfills the demands of unobtrusive human assistance through its outstanding power-to-weight ratio and flexible routing. Driven by these core advantages, we have developed a comprehensive suite of wearable systems. We progressed from upper limb devices, creating an elbow exoskeleton for rehabilitation and a shoulder exosuit (Auxilio), to lower limb exoskeletons designed for hip and knee support in daily and industrial tasks. Currently, our research is expanding into active soft exosuits that maximize TSA’s flexibility, as well as quasi-passive exosuits that exploit the actuator’s intrinsic mechanical changes for highly energy-efficient operation. Across all platforms, our ultimate focus remains on optimizing actuator integration, transmission, and routing to achieve a maximized system-level power-to-weight ratio

Selected Papers

  • Popov, D., Gaponov, I., & Ryu, J. H. (2013, November). Bidirectional elbow exoskeleton based on twisted-string actuators. In 2013 IEEE/RSJ International Conference on Intelligent Robots and Systems (pp. 5853-5858). IEEE. Link
  • Popov, D., Gaponov, I., & Ryu, J. H. (2016). Portable exoskeleton glove with soft structure for hand assistance in activities of daily living. IEEE/ASME Transactions on Mechatronics, 22(2), 865-875. Link
  • Gaponov, I., Popov, D., Lee, S. J., & Ryu, J. H. (2017). Auxilio: A portable cable-driven exosuit for upper extremity assistance. International Journal of Control, Automation and Systems, 15(1), 73-84. Link
  • Seong, H. S., Kim, D. H., Gaponov, I., & Ryu, J. H. (2020, May). Development of a twisted string actuator-based exoskeleton for hip joint assistance in lifting tasks. In 2020 IEEE International Conference on Robotics and Automation (ICRA) (pp. 761-767). IEEE. Link

2. Mechanical Intelligence Embedded System for Rugged Construction Environments

Researchers: JaeHyung Jang, Juyeong Seo, Junhyung Park, JunKyu Han.

Harsh, unstructured environments—characterized by irregular terrain, variable loading conditions, dust, vibration, and frequent unexpected disturbances—are among the most demanding settings for robotic deployment. Operating reliably under such conditions requires robustness as a fundamental design requirement. However, conventional approaches to improving robot capability and performance tend to increase dependence on actuators, sensors, and complex control systems, which inherently reduces system robustness. This presents a key trade-off that is difficult to resolve through traditional design methods. We resolve this trade-off through mechanical intelligence. By embedding adaptive functionality, such as perception, decision-making, and actuation, directly into the mechanical structure, our systems autonomously respond to environmental inputs. This methodology minimizes the need for vulnerable active components, ensuring reliable, high-performance operation in diverse and demanding real-world conditions.

2.1. MORPH Wheel: Input-Responsive Transformable Wheel Embedding Mechanical Behavior Logic

Researchers: JaeHyung Jang, Juyeong Seo

The MORPH wheel is a fully passive variable-radius wheel that solves a fundamental challenge in wheeled robotics: conventional fixed-ratio wheels cannot adapt to changing loads and terrain, forcing a permanent trade-off between speed and torque. Rather than addressing this with additional motors, sensors, or control algorithms, the MORPH wheel embeds decision-making logic directly into its mechanical structure. At its core, a slider-crank torque-response coupler and spring-loaded connecting struts work together to sense input torque and autonomously switch between two modes: when output force is sufficient to overcome terrain resistance, the wheel maintains its full radius for efficient high-speed driving; when it is not, the structure contracts to a transformed radius, increasing the mechanical advantage to generate greater force. This threshold-based behavior is governed entirely by geometry and spring stiffness, requiring no electrical components whatsoever. Beyond its immediate application, the MORPH wheel exemplifies a broader paradigm of embedding context-dependent intelligence into physical hardware, with strong potential for extension to extreme environments where electronics are vulnerable, and to other robotic systems such as prosthetics, wearables, and humanoid platforms.

Associated Papers

  • Jang, J., Seo, J. Y., Lee, D. Y., & Ryu, J. H. (2026). MORPH wheel: A passive variable-radius wheel embedding mechanical behavior logic for input-responsive transformation. arXiv preprint arXiv:2602.06265. Link

2.2. SPINE Gripper: Passive Mode-Transition Gripper

Researchers: JaeHyung Jang, Junhyung Park

The SPINE gripper addresses a longstanding challenge in robotic manipulation: combining stable grasping and in-hand rotation typically requires multiple motors, sensors, and control software, adding weight, complexity, and failure risk. The SPINE gripper achieves both functions with a single actuator and no electronics, by embedding the decision logic for switching between these two modes directly into its mechanical structure. Mode switching is governed by a friction generator that sets a mechanical torque threshold: below it, torque drives finger closing; above it, the coupling slips and torque is redirected to rotate the gripper while maintaining the established grasp. This threshold, and therefore the maximum grasping force, is tuned simply by adjusting the compression of a rubber O-ring, with no sensors or feedback required. When mounted on a robotic arm and driven purely by wrist torque, the gripper requires no wiring at the tool tip, eliminating the cable fatigue and rotation constraints that burden conventional end effectors. Demonstrations spanning bolt tightening, object reorientation, and fruit harvesting confirm that mechanically encoded decision logic alone is sufficient to support a wide range of manipulation tasks that would otherwise demand fully actuated, sensor-equipped hardware.

Associated Papers

  • Jang, J., Park, J., Lee, J. K., & Ryu, J. H. (2026). SPINE Gripper: A Twisted Underactuated Mechanism-based Passive Mode-Transition Gripper. arXiv preprint arXiv:2601.06833. Link

Equipment

1. Torque Sensores and Load cells.

  • FUTEK, TRS605-FSH02052
    • non-contact rotary torque sensor
    • maximum capacity: 1Nm
    • resolution: 0.01Nm
  • Dacell, UU3
    • push-pull loadcell
    • maximum capacity: 5Kgf, 10Kgf, 20Kgf, 50Kgf, 100Kgf
    • resolution: 0.2% of maximum capacity
  • ATI industry, mini45
    • F/T sensor
    • maximum capacity: Fx,Fy=580N, Fz=1160N, Tx, Ty, Tz= 20Nm
    • resolution: 0.2% of maximum capacity

2. Motors and Motor Drivers

  • MAXON BLDC motor
  • MAXON motor driver
    • ESCON 70/10, ESCON 50/8

3. Strings

  • Dyneema series
    • 0.5mm~10mm

4. Encoders

  • HEDS-9740

5. 3D printers

  • Fromlabs, Form3
    • SLA Printer
  • Markforged, onyx pro, Mark two
    • High resolution FDM Printer
  • Stratasys, F170
    • FDM Printer