Twisted Underactuated Mechanism (TUM)

• Twisted Underactuated Mechanism (TUM) is a transmission architecture that uses geometric twist to produce multi-degree-of-freedom outputs from a single actuator input.
• It couples translational and rotational movements, demonstrated in soft robotic walking, spatially dexterous fingers, and passive mode-transition grippers.
• TUMs employ inherent passive mode switching via frictional thresholds or vibratory bifurcations, reducing the need for electronics and complex control systems.

A Twisted Underactuated Mechanism (TUM) is a mechanical transmission architecture that leverages geometric twist or non-coplanarity to generate multiple degrees of output motion and mechanical mode transitions from a single actuator input. TUMs produce non-coplanar kinematic responses, enable coupling between translational and rotational actions, and furnish frequency- or force-dependent functional switching without electronics or sensory feedback. They have been realized in soft robotics (vibratory walking), dexterous robotic fingers (spatial underactuation), and passive mode-transition grippers by exploiting anisotropic stiffness, closed-loop linkages, and geometric conversion of simple actuation inputs into complex outputs (Jiang et al., 2022, Hamon et al., 2021, Jang et al., 11 Jan 2026).

1. Fundamental Principles and Kinematic Configurations

Twisted Underactuated Mechanisms effect non-coplanar force/motion outputs by introducing geometric or material twist into their structure:

• Soft Twisted Beam TUMs: A compliant cantilever beam is fabricated with a uniform geometric twist angle $\phi$ about its axis. Driven at its base by a uniaxial input (often sinusoidal), the spatially-twisted cross-section rotates the principal axes of bending and torsion, generating frequency-dependent bending-bending coupling and enabling three-dimensional tip trajectories from a 1-DoF input. The mechanism is modeled as a chain of five pseudo-rigid links (three bending, two twisting joints) with torsional and bending compliance (Jiang et al., 2022).
• Spatially-Twisted Finger TUMs: In robotic fingers, twist is introduced at the joint-level architecture. Each finger possesses a 2-DoF spherical parallel base, with the second joint arranged orthogonally (twisted) to proximal axes, followed by sequential planar four-bar linkages. All joints are actuated via a single input, with motion and force transmission managed by a network of closed kinematic loops, resulting in spatially-dexterous outputs from one actuator (Hamon et al., 2021).
• Twisted-String/Strip TUMs in Grippers: Here, four compliant strips of length $L$ are wrapped between parallel plates around a radius $R$, forming a twisted-string actuator. Twisting the plates by an angle $\theta$ contracts the mechanism axially and drives symmetric non-coplanar finger motion. The contraction and elastic restoring force are even functions of $\theta$, enabling bidirectional symmetry in force transmission and actuation (Jang et al., 11 Jan 2026).

2. Mathematical Models and Transmission Characteristics

A core characteristic of TUMs is the coupling between their geometric twist and force/motion output. The analyses vary by architecture:

• Twisted Beam Model: For a pre-twisted beam, local axis bending ($k_{b_y}$, $k_{b_z}$) is coupled in the global frame by $$K_\text{bend, global} = R_z(\phi)\cdot \text{diag}(k_{b_y},k_{b_z})\cdot R_z(\phi)^T$$, introducing off-diagonal terms $k_{yz} = (k_{b_y}-k_{b_z})\sin\phi\cos\phi$ responsible for cross-axis energy transfer. The dynamics with ground contact are formulated via a set of coupled ODEs:

$$M(q)\,\ddot q + C(q,\dot q)\,\dot q + D\,\dot q + K\,q = B\,A\sin(2\pi f t) + J_c^T F_c(t)$$

where $q$ parameterizes bending and twisting states; $F_c(t)$ models ground impact and friction (Jiang et al., 2022).

• Spatial Finger Model: The mapping between input force and fingertip wrench is mediated by the actuation matrix $T$ and the Jacobian $J$:

$$\tau = T\, [F_{10}, \Gamma_1, N_1, \Gamma_2]^T, \quad \tau = J^T f$$

Contact forces are recovered as $f = J^{-T} T^{-T} t$ for a given actuator force vector $t$ (Hamon et al., 2021).

• Twisted-Strip Actuator Model: The axial contraction $\Delta L(\theta)$ and transmission Jacobian $J_{\rm TUM}$ as a function of twist angle are:

$\Delta L(\theta) = L - \sqrt{L^2 - (R\theta)^2}$

$$J_{\rm TUM}(\theta) = \frac{R^2\theta}{\sqrt{L^2-(R\theta)^2}}$$

Force transmission and equilibrium analysis yield:

$$F_c(\theta) = J_{\rm TUM}^{-1}(\theta)\tau_\mathrm{in} - F_s(\theta)$$

where $F_s$ is the strip elastic restorative force (Jang et al., 11 Jan 2026).

3. Passive Mode Switching and Transmission Logic

TUMs enable mechanically encoded mode transitions—e.g., from grasping to in-hand rotation—through nonlinear transmission logic governed by frictional thresholds or vibratory bifurcations:

• Dynamic Mode Selection (Walking TUM): By sweeping the actuation frequency, distinct locomotion gaits and travel directions emerge due to subharmonic transitions in ground contact: at critical frequencies, tip trajectories bifurcate from lines to figure-8 and double loops, and net motion switches direction. Small parameter adjustments (input amplitude, beam twist) alter gait without electronic or algorithmic control (Jiang et al., 2022).
• Friction-Based Mode Switching (Gripper TUM): In the SPINE Gripper, a passive friction generator (O-ring and compression spacers) sets static and kinetic torque thresholds $\tau_\mathrm{grasp}$ and $\tau_\mathrm{rotate}$. Input torque below $\tau_\mathrm{grasp}$ drives symmetric contraction (grasp mode), while torque above $\tau_\mathrm{grasp}$ overcomes friction and induces rotation of the entire gripper (rotation mode). This results in robust grasp-to-rotate transitions without sensors; mode selection is direction-invariant due to the even-function kinematics of the TUM (Jang et al., 11 Jan 2026).

4. Grasping, Stability, and Multifunctionality

Twisted underactuated architectures are exploited to achieve stable, adaptive, and reconfigurable grasping as well as spatial manipulation:

• Robotic Finger TUM: The kinematic design enables smooth variation between cylindrical and spherical grips via the twist coordinate $\theta_2$. Stability is analyzed via the positive-definiteness of the contact-force mapping $f = J^{-T} T^{-T} t$: if all resultant contact forces are nonnegative, the grasp is stable under one-input actuation over the full workspace. Load sharing between contacts is tuned by the internal geometry, ensuring grasp robustness in both coplanar and spatial configurations (Hamon et al., 2021).
• Twisted-Strip Gripper TUM: Grasp force is regulated solely by parameterized friction torques; empirical tests demonstrate reliable grasping and force maintenance over a continuously variable range (1–5 N), with failre only at intentionally underset static friction. The bidirectional nature permits forward and reverse in-hand rotation tasks (e.g., bolt loosening/tightening) using only input torque reversals, validated in uninterrupted pick-maintain-rotate demonstrations on robotic arms (Jang et al., 11 Jan 2026).

5. Design Guidelines and Parameterization

Effective TUM deployment depends on careful mechanism parameterization:

• Twisted Beam: Length $L$, twist angle $\phi$, width $w$, and thickness $t$ are selected to place mechanical resonances in the desired actuation frequency range (15–80 Hz). Maximal bending coupling is realized at $\phi\approx\pm90^\circ$. Rigid foot-length and friction coefficients of the ground interface modulate ratcheting behavior and efficiency (Jiang et al., 2022).
• Twisted-String/Strip: Strip length $L$, plate radius $R$, elastic modulus $E$, and moment of inertia $I$ are chosen such that the maximum actuation torque lies below available friction thresholds, and mechanical interference is avoided over the full range of $\theta$. Friction generator geometry and materials are tuned to predictably set static and kinetic friction for reliable mode transitions and grasp/rotate boundaries (Jang et al., 11 Jan 2026).
• Finger Mechanism: Geometric and kinematic parameters of the parallel and planar linkages, and the spatial twist, are determined to avoid singular configurations and ensure sufficient closure and load distribution in both cylindrical and spherical grip modes (Hamon et al., 2021).

6. Experimental Results and System Demonstrations

Extensive empirical validation across TUM platforms confirms predicted mechanical responses, transmission logic, and application-level functionality:

• Walking Robot with Twisted Soft Beams: Forward and backward gaits, subharmonic contact phenomena, and high walking speeds ($\sim$156 mm/s) are achieved solely by tuning input frequency; measured tip and contact force trajectories correspond to predicted bifurcation patterns (Jiang et al., 2022).
• SPINE Gripper: Laboratory tests establish quantitative correlation between friction generator tuning, grasp force range, and passive mode transitions. Practical demonstrations include bolt-handling and manipulator-integrated object reorientation without end-effector electronics or explicit control (Jang et al., 11 Jan 2026).
• Spatial Finger: Kinematic and kinetostatic analyses substantiate continuous force-closure during both planar and spherical gripping, confirmed by model-based mapping of contact force distributions across representative configurations (Hamon et al., 2021).

7. Applications and Broader Significance

TUMs enable underactuated soft locomotion, dexterous manipulation, and gripper mode elasticity, all with minimal sensors or active control:

• Robotics and Soft Machines: Their intrinsic coupling of spatial deformation, direction-invariant response, and mechanically encoded transitions supports simplified actuation strategies for mobile robots and grippers.
• Grasp Synthesis: By using internal twist to mediate transitions between grasp types, TUMs offer alternatives to multi-actuated, sensor-rich hands, particularly for compact or field-robust platforms.
• Mechanically Encoded Logic: The exploitation of geometric and frictional boundaries to trigger functional transitions highlights the potential of TUMs for on-device mode logic, reducing reliance on computation.

A plausible implication is that further research may extend TUM principles to broader classes of multifunctional soft and hard mechanisms, incorporating programmable compliance and logic through geometry and material selection rather than electronics.

Key references: (Jiang et al., 2022, Hamon et al., 2021, Jang et al., 11 Jan 2026)