Dual-Segment Continuum Robot

• Dual segment continuum robots are flexible, multi-section manipulators with continuously compliant segments enabling smooth bending and extension.
• They use tendon-driven, pneumatic, or hybrid actuation, allowing independent or coordinated control for expanded workspace and dexterity.
• Advanced kinematic modeling and control techniques, including piecewise constant curvature and deep-learning methods, boost performance in medical, inspection, and industrial tasks.

A dual segment continuum robot is a serial, multi-section robotic manipulator in which each segment is a spatially continuous, compliant structure with bending and/or extension degrees of freedom (DoFs), enabling large, smooth deformations. The dual-segment architecture enables independent or coordinated shape change in two serially connected continuum elements, granting substantially enhanced workspace, dexterity, and tip orientation capabilities over single-segment systems. Dual-segment continuum robots underlie significant advances in medical robotics, inspection in constrained environments, nuclear decontamination, and fine manipulation, with embodiments that span soft pneumatic arms, tendon-driven mechanisms, and hybrid eversion-steering architectures. The following sections synthesize the main modeling, mechanical, control, and design principles from the most recent peer-reviewed accounts.

1. Mechanical Architectures

Dual-segment continuum robots exist in both soft and hybrid-soft modalities. Representative categories include:

• Tendon-driven dual-segment arms: Each segment comprises a backbone (e.g., NiTi rod, polymer shaft) with spacer disks to route antagonistic tendon pairs, producing planar or spatial constant-curvature arcs. Modular couplings enable reconfiguration. Key parameters: segment length 0.10–0.50 m, disk diameter 20–60 mm, up to four tendons per segment, and discrete or pseudo-continuous DoF (Ouyang, 2019, Wang et al., 2019, Walker et al., 16 Mar 2025, Shao et al., 3 Feb 2026).
• Pneumatically actuated soft (cylindrical) segments: Segments are fabricated from silicone rubber with fiber reinforcement. Intersecting sets of pneumatic chambers drive extension and multi-axis bending. Segment decoupling is achieved via internal partitioning, e.g., a variable-stiffness granular-jamming spine (Wang et al., 2024, Doroudchi et al., 2022).
• Hybrid continuum-eversion robots: A long, soft, pneumatically everting “vine” provides gross deployment; an actively steerable continuum tip (disks + springs + tendon actuators) provides distal precision (Al-Dubooni et al., 2024).
• Mixed-modality (tendon + magnetic/telescoping) robots: Proximal continuum segment actuated by tendons; distal telescoping segment consists of magnetic spheres actuated by an external field, producing a highly dexterous tip (Pittiglio et al., 2024).

Down-selection of segment lengths, diameters, materials, and DoF follows the constraints of required workspace, dexterity, stiffness/payload, environmental compliance, and miniaturization.

2. Kinematics and Modeling

The dominant framework is the piecewise constant curvature (PCC) model, treating each segment as a parameterized arc. This admits analytical or semi-analytical forward kinematics:

• For segment i (i=1,2), bending of length $l_i$, curvature $\kappa_i$, and plane angle $\phi_i$:

$$\theta_i = \kappa_i l_i, \qquad T_i = \text{exp}([\hat{u}_i]\,\theta_i)$$

where $T_i$ is the homogeneous transformation, $u_i$ is the rotation axis.

• Tip frame:

$T_\text{tip} = T_1\,T_2$

or, for planar case, compound two planar rotations and translations as detailed in (Walker et al., 16 Mar 2025, Ma et al., 19 Mar 2025, Wang et al., 2021).

Closed-form inverse kinematics (IK) are available for two-segment, inextensible PCC robots, yielding fast convergence and analytic workspace boundaries (Wang et al., 2021). For continuum robots with extensible, variable-stiffness, or continuum-eversion structures, modeling must account for nonuniform Young’s modulus, jamming transitions, and compliant environmental contacts (Wang et al., 2024, Doroudchi et al., 2022, Al-Dubooni et al., 2024).

For more intricate modeling, Cosserat rod and Kirchhoff rod theories capture bending, torsion, shear, and extension—solved by shooting methods, real-time finite difference, or Koopman operator reduction (Doroudchi et al., 2022, Ristich et al., 15 Sep 2025).

3. Actuation, Sensing, and Control

• Tendon-driven actuation: Tendon displacement $\Delta l$ linearly or nonlinearly maps to curvature: $\kappa = \Delta l/dl_i$ (for tendon offset $d$). Dual-segment designs employ two independent sets for spatial control; coupling terms may require compensation (Ouyang, 2019, Shao et al., 3 Feb 2026).
• Pneumatic actuation: Pressure in multi-chambered elastomeric segments, mapped to moments and curvature via $M = pA d$ for chamber spacing $d$. Variable segment stiffness via granular or layer jamming enables underactuated modes (Wang et al., 2024).
• Hybrid/eversion actuation: Bulk body growth by pneumatic eversion, steering by tendon-actuated tip or magnetic fields. Decoupled gross and fine control (Al-Dubooni et al., 2024, Pittiglio et al., 2024).
• Sensing: Distributed IMUs, tip-mounted cameras, cable tension cells, and (in soft arms) motion capture or embedded curvature sensors (Walker et al., 16 Mar 2025, Doroudchi et al., 2022).
• Feedback control: Joint-space and task-space controllers, often via resolved-rate algorithms, Jacobian-based inverse kinematics, or, for high-fidelity tracking under nonlinearity and hysteresis, deep recurrent neural networks (GRU, LSTM) (Shao et al., 3 Feb 2026). Koopman operator–based linear MPC allows real-time whole-shape tracking (Ristich et al., 15 Sep 2025).

4. Stiffness Modulation and Variable Compliance

Several dual-segment platforms implement tunable or region-specific stiffness:

• Layer jamming: Segments constructed with PET flaps, latex sleeves, and vacuum-bag overlays shift between compliant and stiffened modes via negative pressure; measured tip-load stiffness ratio between jammed and unjammed up to 17.5 (Ouyang, 2019).
• Granular jamming spine: A fabric "growing" spine, loaded with hollow glass bubble granules, can be jammed pneumatically to set the Young's modulus of any prefix of the manipulator, realizing continuous spatial stiffness profiles (Wang et al., 2024).
• Integrated rigid-compliant joints: Alternating NiTi rods and rigid disks with mechanical interlocks for selective shape-locking and payload support (Wang et al., 2019).
• Decoupling for fine manipulation: Modular mid-links or low-friction interfaces to minimize crosstalk between proximal and distal segments, yielding independent segment motion (Pittiglio et al., 2024, Shao et al., 3 Feb 2026).

These strategies enable S-shaped curves, adaptive grasping, payload support, or, in nuclear and surgical contexts, the requisite precision-force tradeoff.

5. Workspace, Dexterity, and Task Performance

Dual-segment architectures substantively expand reachable workspace and achievable local orientation at the tip:

• Workspace quantification: Direct, analytical formulation of reachable positions and dexterous orientation sets, including explicit boundaries for inextensible dual-PCC robots. The dual-segment system supports full orientation in a larger subset of positional workspace than any single segment of similar length (Wang et al., 2021, Pittiglio et al., 2024).
• Dexterous workspace: For hybrid tendon–ball-chain robots, the combined architecture allows any target in the “fully dexterous zone” to be approached from any orientation, validated experimentally with ≤7% positional error over 24 configurations (Pittiglio et al., 2024).
• Precision and repeatability: Reported RMSEs for dual-segment systems with advanced controllers are in the sub-millimeter (e.g., 0.14 mm) and sub-degree (0.7°) range even in biological tissue interaction (Shao et al., 3 Feb 2026). For spraying, coverage was 97% with ±3.7% standard deviation (Al-Dubooni et al., 2024).
• Disturbance rejection: Closed-loop dual-segment controllers maintain <5% spatial and <5° angular error under 300 g distal loads; configuration recovery within ≈3 s after perturbation (Walker et al., 16 Mar 2025).

6. Inverse Kinematics, Planning, and Computational Efficiency

Efficient dual-segment inverse kinematics and planning algorithms are a focal area:

• Variable-separation IK: Dual-segment, constant-length continuum robots admit a geometry-based separation, reducing IK to a one-variable nonlinear root solve with provable 100% convergence in simulation, and offering 96% computational time savings over Jacobian-DLS (Wang et al., 2021).
• Geometric iterative methods: Two-layer solvers combining per-segment FABRIKc with outer yaw-mode compensation achieve convergence in typically ≈2 iterations, yielding 4 mm position and ≤1° orientation errors (experimental) (Ma et al., 19 Mar 2025).
• Follow-the-leader planning: Dual-segment arms with floating bases and collision-aware planners navigate confined, occluded, or obstacle-rich environments, leveraging conical/spherical sampling for constraint satisfaction (Ma et al., 19 Mar 2025).
• Data-driven, control-affine modeling: Koopman operator projections per segment enable real-time (≤20 ms update) tracking of shape targets, reducing MSE by orders of magnitude versus nonprojected control. These methods are applicable to dual-segment shape controllers (Ristich et al., 15 Sep 2025).

7. Application Domains and Future Trends

Dual-segment continuum robots have demonstrated impact in:

• Medical/surgical: Endoscopic submucosal dissection, peg transfer, with workspace >2× single segment, decoupled 6-DoF tip control, and soft-tissue compliance (Shao et al., 3 Feb 2026).
• Hazardous/nuclear inspection: Growth via eversion for pipe-traversal and remote decontamination; precision delivery of liquids, aerosols with >95% accuracy (Al-Dubooni et al., 2024).
• Industrial/aeroengine: In-situ inspection and repair employing ultra-slender dual-stage arms with c- and C-shape form-locking, selective stiffness, and payload handling up to 125 g (Wang et al., 2019).
• Marine/subsea: Underwater operation with precise configuration and disturbance restoration (Walker et al., 16 Mar 2025).

Emergent directions include miniaturization (to sub-10 mm diameters), extension to more segments, integrated module-swapping for task versatility, and closed-loop autonomous path following with embedded sensor networks (Al-Dubooni et al., 2024, Wang et al., 2024). Advances in model-based, Koopman-based, and deep-learning control architectures are driving further increases in closed-loop performance, robustness to nonlinearity, and computational scalability.

---

References:

• (Al-Dubooni et al., 2024, Wang et al., 2024, Ouyang, 2019, Walker et al., 16 Mar 2025, Doroudchi et al., 2022, Wang et al., 2019, Pittiglio et al., 2024, Ma et al., 19 Mar 2025, Wang et al., 2021, Muhmann et al., 26 Mar 2025, Ristich et al., 15 Sep 2025, Shao et al., 3 Feb 2026)