Spatially-Twisted Finger TUMs

• Spatially-twisted finger TUMs are innovative robotic gripper architectures that convert a uniaxial input into coupled spatial contraction and twisting motions.
• They leverage diverse actuation methods—mechanical, tendon-driven, and pneumatic—to achieve stable pinch grasps, bidirectional rotation, and adaptable in-hand manipulation.
• Key design parameters such as helix angle, compliance of elastic elements, and passive friction locks are fine-tuned to optimize performance metrics like force-closure and dynamic maneuver speed.

Spatially-twisted finger Twisted Underactuated Mechanisms (TUMs) are a class of robotic finger and gripper architectures distinguished by their ability to produce non-coplanar, multi-degree-of-freedom (DoF) fingertip motions—typically combining axial contraction with spatial rotation or bending—using mechanisms intrinsically structured in three dimensions. These mechanisms provide dexterous manipulation capabilities, including stable grasping and in-hand object rotation or reorientation, often with reduced actuation and sensing complexity. Recent implementations have achieved spatial untwisting, bidirectional spinning, adaptive wrapping, and high-speed dynamic maneuvers, using distinct actuation strategies (mechanical, tendon-driven, or pneumatic), geometry manipulations, and passively encoded transmission logics (Jang et al., 11 Jan 2026, Ishikawa et al., 28 Oct 2025, Guan et al., 1 Mar 2025).

1. Mechanistic Principles of Spatially-Twisted Finger TUMs

Spatially-twisted finger TUMs are characterized by their ability to transform a uniaxial, single-input actuation (either rotational, tensile, or pressure-driven) into a coupled spatial contraction and twisting of the finger modules. In canonical implementations, such as the SPINE gripper (Jang et al., 11 Jan 2026), TUMs deploy two parallel end plates linked by multiple elastic or compliant strips set at a helix angle α. Upon input shaft rotation by ±θ, the compliant strips undergo both torsional and axial deformation, resulting in a non-coplanar downward contraction ($X(\theta)$) and storage of elastic energy, with direction-invariant behavior ($X(\theta)$ is even in θ). This geometric encoding yields identical contraction profiles for clockwise and counterclockwise inputs, directly enabling bidirectional functionality.

For soft adaptive-twist modules (Ishikawa et al., 28 Oct 2025), spatial deformation is generated by a pneumatic chamber driving jointed exoskeletons, allowing both in-plane and out-of-plane finger wrapping. Similarly, the Offset-Trimmed Helicoids (OTH) architecture (Guan et al., 1 Mar 2025) achieves 3D, variable curvature by routing tendons helically along an elastomeric helix whose deformation center can be offset by design, thus supporting complex spatial bending and twist within a single compliant finger.

2. Kinematic, Elastic, and Static Models

The spatial kinematics of TUM-based fingers are determined by the interplay of helical strip geometries, compliant beam theory, and the topology of the spatial arrangement.

Twisted Beam TUMs:

In TUMs composed of $N$ strips of length $L$ connected between plates of radius $R$, the contraction $X(\theta)$ upon input shaft rotation is

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

with an associated Jacobian

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

Elastic force output from each strip, modeled as a bent beam, is

$$P(\theta) = \left[\frac{\pi}{L - X(\theta)}\right]^2 E I,$$

projected along the central axis by angle $\beta$ arising from the local strip geometry. Aggregate contraction force and the mapping to finger joint torque are then assembled via finger linkage Jacobians.

Soft and Modular TUMs:

In adaptive-twist soft mechanisms, the spatial kinematics are compounded by pressure-driven bulging of elastomeric backbones, mediated through articulated exoskeletons with subsequently locked axes. While no P→θ closed-form mapping is reported, the variable-stiffness response, as the pressure passes structure-defined locking thresholds, sharply transitions the joint response from highly compliant to rigid.

OTH TUMs:

Offset-trimmed helicoid models (Guan et al., 1 Mar 2025) apply piecewise constant curvature (PCC) methods to treat each helicoid finger as a chain of segments with architected compliance and curvature:

$X(\theta)$0

Multitendon actuation introduces additional control over segmental configuration, supporting both pure bend and spatial twist.

3. Frictional, Stiffness, and Mode-Transition Mechanisms

Spatially-twisted TUMs frequently leverage passively tuned frictional or structural thresholds to govern transition between grasp and in-hand manipulation.

Mechanical Friction Locks:

In the SPINE Gripper (Jang et al., 11 Jan 2026), preloaded nitrile-rubber O-rings define torque thresholds for passive mode switching: below the static friction torque, actuation is directed to contract the fingers; above this threshold, the mechanism passively "slips," freeing the entire finger module to spin about its axis. The relevant torque thresholds are

$X(\theta)$1

with empirical ranges set via mechanical spacer design, directly regulating achievable grasp force and rotational slip torque.

Variable Stiffness Locking:

The adaptive-twist finger (Ishikawa et al., 28 Oct 2025) introduces pressure-tunable modular joints wherein contact-normal forces between interlocking exoskeleton plates provide a variable locking moment:

$X(\theta)$2

where locking is designed to occur only beyond a pressure setpoint, thus switching the finger from insertion-compliant to load-bearing.

Material and Mounting Effects:

OTH-based modules (Guan et al., 1 Mar 2025) facilitate compliance and force transmission tuning by adjusting the lateral offset of the neutral axis and tailoring the orientation of base mounting, providing distributed compliance ellipsoids and expanded workspace as required for specific tasks.

4. Spatial Arrangement and Multifinger Configurations

Spatially-twisted finger TUMs are well-suited for both single- and multi-fingered gripper deployments due to their architected 3D deformation and mode-transition properties.

Symmetric 2-Finger Linkages:

The SPINE gripper arranges dual fingers in a plane orthogonal to the TUM contraction axis, connected by a single slider-crank to the central contraction plate (Jang et al., 11 Jan 2026). This ensures symmetric radial pinch and subsequently uncoupled rotational manipulation.

Soft Modular Arrays:

Adaptive-twist modules can be duplicated and concatenated to create multi-finger hands; all units share a common high-pressure supply with independently scalable exoskeleton locking (Ishikawa et al., 28 Oct 2025). Multi-DoF wrapping and grasping are thereby realized without additional actuators.

Tendon-Driven OTH Arrays:

Three-finger OTH grippers employ equiangular mounting with 6 independent tendon DOFs, achieving 6-DoF dexterity (global pinch, lateral shear, twisting, and object spinning) by coordinated actuation and by leveraging spatial offset to maximize overlapping compliance in the grasp frame (Guan et al., 1 Mar 2025).

5. Design Parameters, Tuning, and Performance Metrics

Critical TUM parameters, both geometric and operational, directly determine the grasp strength, mode-transition threshold, range of achievable spatial twist, and dynamic performance.

Architecture	Key Parameters (as reported)	Performance Metrics (prototype)
SPINE Gripper	R=20 mm, L=40 mm, N=4 strips, α≈8.6°, O-ring preload (by spacer), E≈1.2 GPa	Max contraction ≈40 mm, finger rotation 0°–120°, grasp force 1–5.3 N, bidirectional rotation >360°, actuation torque up to ≈700 N·mm
Adaptive-Twist	NBR tube ID=20 mm, wall=6 mm, 3-part PLA rings, joint gap d=2.5 mm, μ=0.4–0.7	Max in-plane bend 135°, out-of-plane 115°, max tip force 8.5 N/joint, grasp of up to 3 kg, response rise-time ~0.6 s
OTH Gripper	Helix r, pitch p, width w; offset Dc=0–10 mm; finger tilt α=45°	Workspace ≈125 mm (y), 143 mm (z), tip compliance tunable 3×, object spin 60°/15 ms, total 6 DoF, force-closure for diverse objects

Fine-tuning these design quantities—plate radius, elastic modulus, O-ring preload for frictional locks, exoskeleton gap for variable stiffness, and helicoid geometries for OTHs—allows adaptation to desired ranges of object size, grasp force, compliance profile, and required manipulation dynamics.

6. Applications and Demonstrated Capabilities

Spatially-twisted finger TUMs have been validated in diverse dexterous manipulation tasks.

• The SPINE Gripper achieves stable two-finger pinch grasps and continuous bidirectional in-hand rotation for tasks such as bolt manipulation and object reorientation, powered by wrist torque alone (Jang et al., 11 Jan 2026).
• Adaptive-twist modules enable grasping by deep wrapping around objects packed in dense environments and then stiffen upon command to securely pick up objects weighing up to 3 kg (Ishikawa et al., 28 Oct 2025).
• OTH-based three-finger grippers have demonstrated simultaneous in-plane and out-of-plane spatial manipulation, secure force-closure on geometric primitives, and high-speed dynamic in-hand spinning (60° in ~15 ms), emulating levels of dexterity comparable to the human hand for certain tasks (Guan et al., 1 Mar 2025).

A common implication is that by harnessing spatially-twisted kinematic paths and passive mode-transition logics, multifunctional manipulation can be realized robustly without recourse to complex control or sensing.

7. Research Context, Generalizations, and Prospects

Spatially-twisted finger TUMs embody two strategic advances in robotic manipulation: (1) the direct mechanical encoding of sequential or concurrent spatial motion primitives, and (2) the reduction of control and sensing overhead through passive or structurally intrinsic logic. Their realization across rigid, hybrid, and soft material platforms, with transmission architectures spanning mechanical, pneumatic, and tendon-driven systems, attests to their generality and adaptability.

Ongoing lines of inquiry include scaling of spatial TUMs to larger multi-fingered hands, optimizing trade-offs between compliance and force output through informed geometry-material choices, and integrating additional spatial DoF (e.g., twist about the Z-axis in adaptive modules (Ishikawa et al., 28 Oct 2025), or multi-segment OTH stacking (Guan et al., 1 Mar 2025)). Practical limitations noted include the need for precise geometric and friction tuning, as well as robustness to material fatigue—highlighting areas for continued experimental and theoretical exploration.