Magnetic Tentacle Robot (MTR) in Endoluminal Therapy

• Magnetic Tentacle Robot (MTR) is a soft, continuum manipulator with embedded magnetic particles for curvilinear navigation in human cavities.
• They are fabricated from elastomeric composites and actuated by external magnets to achieve bending, torsion, and mode switching for precise interventions.
• Recent studies demonstrate MTRs' applicability in targeted pancreatic duct drug delivery, using ultrasound triggering to control payload release.

Searching arXiv for the cited MTR-related papers to ground the article in current literature. arXiv search query: magnetic tentacle robot pancreatic duct ultrasound (Moorcroft et al., 4 Sep 2025) A magnetic tentacle robot (MTR) is a magnetically actuated soft continuum manipulator characterized in the literature as “thin, soft, scalable, and can be shaped into curvilinear morphologies under magnetic influence” for navigation within human cavities. In the current arXiv literature, the term is used explicitly for an endoluminal soft magnetic carrier that transports a stimulus-responsive therapeutic coating into the pancreatic duct, holds position during external triggering, and is then retracted so that no foreign body is retained (Moorcroft et al., 4 Sep 2025). Closely related work often uses the designation “magnetic continuum robot” for mechanically similar devices: slender, flexible, catheter-like bodies with distributed compliance, remote magnetic actuation, and local task execution in constrained lumens (Zhang et al., 2 Oct 2025).

1. Terminology, scope, and conceptual identity

The most explicit recent definition appears in the pancreatic-duct drug-delivery study, which presents MTRs as soft, continuum, tentacle-like robots intended for curvilinear deformation and endoluminal navigation. In that work, the robot is not itself the drug reservoir; it is the navigated carrier for an ultrasound-responsive fabric wrapped around the distal end. The distinction is central: navigation and positioning are assigned to the soft magnetic body, while payload retention and release are assigned to an attached material system (Moorcroft et al., 4 Sep 2025).

A broader taxonomy emerges when this device is read alongside adjacent magnetic-soft-robot papers. The dual-mode magnetic continuum robot of 2025 is not named an MTR, but it shares the defining architectural features of the tentacle-robot class: a long, flexible, catheter-based body; distributed deformation rather than rigid-link articulation; wireless magnetic actuation at the distal region; and the capacity to navigate tortuous channels and then perform a local intervention (Zhang et al., 2 Oct 2025). By contrast, the helical micro-robot for filamentous microcargo is explicitly not a tentacle robot; it is a rigid screw-propelled swimmer whose relevance lies in non-contact transport physics rather than continuum tentacle mechanics (Ghadami et al., 2023). Likewise, the tri-leg silicone robot controlled by a vision-language-action model is morphologically distinct from a tentacle robot and is more directly relevant to autonomy and voltage-level magnetic control than to tentacle embodiment (Tang et al., 28 Feb 2026).

A recurrent misconception is to equate all magnetic medical robots with untethered swimmers or rigid capsules. The recent MTR literature instead centers a different regime: soft or continuum bodies, often delivered through conventional endoscopic channels, steered by external fields, and shaped by compliance rather than by rigid-body propulsion. In this sense, “tentacle” refers less to a specific geometry than to a continuum mode of deformation and interaction in anatomically constrained spaces (Moorcroft et al., 4 Sep 2025).

2. Embodiment, materials, and magnetic actuation

The pancreatic-delivery MTR provides the clearest concrete embodiment. It is fabricated by blending Dragon Skin 30 with $5\,\mu\text{m}$ neodymium-iron-boron microparticles in a $1{:}1$ mass ratio, yielding a soft elastomeric composite rod with cross-sectional diameter $1.82 \pm 0.04\,\text{mm}$ and length $60 \pm 1\,\text{mm}$. The finished body is saturated in a uniform magnetic field of $4.644\,\text{T}$ along its axis, producing an axially magnetized soft continuum manipulator rather than a rigid catheter (Moorcroft et al., 4 Sep 2025).

In that system, actuation is provided by a KUKA LBR iiwa 14 manipulating one cylindrical external permanent magnet with diameter and length $101.6\,\text{mm}$ and magnetic moment $970.1\,\text{A}\cdot\text{m}^2$ (Grade N52). The external magnet is positioned to apply a pulling magnetic field on the fabric-coated MTR, allowing it to travel into the pancreatic duct. The paper is careful on modeling scope: it demonstrates magnetic manipulation experimentally but does not present explicit magnetic force, torque, beam-bending, or continuum-kinematics equations for the MTR. Its only printed equations quantify gelatin functionalization chemistry rather than robot mechanics (Moorcroft et al., 4 Sep 2025).

The dual-mode magnetic continuum robot extends this actuation logic by replacing the usual axial magnet arrangement with three radially embedded permanent magnets near the distal tip. Under an external field, each internal magnet experiences torque according to

$\bm{m}=\boldsymbol{\mu}\times \mathbf{B},$

and the resultant torque can be directed either transverse to the catheter axis, producing bending, or along the axis, producing torsion. This architecture allows the same external permanent magnet, manipulated by a robotic arm, to switch between bending mode for navigation and torsional mode for local intervention (Zhang et al., 2 Oct 2025).

A different extension appears in the millimeter-scale soft robot with two tentacles and a reprogrammable module. That platform combines soft tentacle appendages, a multifunctional main body, and state-switchable magnetization in selected modules. It is actuated with relatively uniform and weak magnetic fields “at most 65 mT and 1.5 T/m,” while using shape-induced magnetic anisotropy and gradients to recover full six-degree-of-freedom motion, including rotation about its net magnetic moment—a degree of freedom that conventional five-degree-of-freedom magnetic robots lack (Ng et al., 19 Sep 2025).

Taken together, these embodiments define the MTR field less by a single geometry than by a design pattern: a compliant slender body, encoded or embedded magnetic moment, remote actuation by external magnets or coils, and task execution through deformation at or near the distal region.

3. Endoluminal drug delivery as the canonical recent MTR application

The most fully developed application to date is localized pancreatic-duct drug delivery. In this proof-of-concept system, the MTR serves as the navigated carrier for an electrospun poly($\epsilon$-caprolactone) fabric coated with poly(vinyl alcohol) and albumin-loaded vinylbenzyl-functionalized gelatin nanoparticles. The fabric is prepared by coating the electrospun substrate with a PVA solution containing albumin-loaded nanoparticles and wrapping it around the distal end of the MTR. Attachment is achieved by fixing one end of the strip on top of the tentacle with a $30\,\text{wt}\%$ aqueous PVA solution, wrapping the strip around the tentacle, fixing the other end to the underlying layer, and applying $1{:}1$0 of $1{:}1$1 PVA to each corner (Moorcroft et al., 4 Sep 2025).

The procedural workflow is a phantom endoluminal intervention. The MTR is deployed from the tool channel of a duodenoscope (JF-130, Olympus), guided adjacent to the major duodenal papilla, steered past the bile duct into the pancreatic duct of an anatomical ultrasoft phantom, held in place, exposed to external ultrasound, and then retracted. The ultrasound is applied directly toward the pancreatic duct for 3 minutes at $1{:}1$2, pulsed mode, $1{:}1$3 duty cycle, $1{:}1$4 repetition rate, and in many experiments $1{:}1$5, using a $1{:}1$6 collimating transducer in unfocused mode (Moorcroft et al., 4 Sep 2025).

The attached fabric does not interfere with motion out of the duodenoscope or magnetic navigation into the pancreatic duct, and all fabrics remain securely attached throughout the procedure. The navigation and retraction sequence is completed in under 90 seconds. Release is tunable at several levels. Passive nanoparticle leakage decreases as PVA concentration is increased up to $1{:}1$7, while pulsed ultrasound at $1{:}1$8 triggers an additional $1{:}1$9 nanoparticle release from the fabric. For albumin-loaded systems, ultrasound after hydration produces a burst release, and MTR-mounted fabrics show statistically significant increases in albumin release within the phantom pancreatic duct compared with incubation without ultrasound. A $1.82 \pm 0.04\,\text{mm}$0 MTR-tipped fabric releases about $1.82 \pm 0.04\,\text{mm}$1 of albumin under ultrasound agitation, while a $1.82 \pm 0.04\,\text{mm}$2 strip yields nearly double that release, though scaling is nonlinear because the mat coils over itself and shields lower layers (Moorcroft et al., 4 Sep 2025).

This system establishes the most concrete recent meaning of an MTR in biomedicine: a soft magnetic endoluminal manipulator that provides spatial precision, while an orthogonal trigger—in this case ultrasound—provides temporal control of payload release. Figure 6C in the paper is the primary in situ visualization of that role, showing MTR emergence from a duodenoscope, transit through the papilla, passage into the pancreatic duct, and retraction (Moorcroft et al., 4 Sep 2025).

4. Mechanics, control, and multifunctionality beyond bend-only operation

One major trajectory in MTR research is the move from bending-only steering to mode-selective or multifunctional deformation. The dual-mode magnetic continuum robot provides the clearest explicit mechanics. It models the distal section as a slender elastic continuum with flexural rigidity $1.82 \pm 0.04\,\text{mm}$3 in bending and torsional rigidity $1.82 \pm 0.04\,\text{mm}$4 in twist, and validates its actuation principle through physics-based formulation, finite-element analysis, and benchtop experiments. The reported maximum deformations are a bending angle of about $1.82 \pm 0.04\,\text{mm}$5 and a torsional angle of about $1.82 \pm 0.04\,\text{mm}$6. In phantom intervention, the robot follows $1.82 \pm 0.04\,\text{mm}$7-wide square channels, reaches selected branches, and completes 7/7 successful intervention–release trials. A representative run takes $1.82 \pm 0.04\,\text{mm}$8 from phantom entry to target arrival, with about $1.82 \pm 0.04\,\text{mm}$9 spent positioning and orienting the external permanent magnet (Zhang et al., 2 Oct 2025).

The same study is important conceptually because torsion is not treated as incidental. The distal release structure contains outer grooves and inner plates; magnetic torsion opens the outlet path, while additional proximal torsion induces helical deformation and payload expulsion. The result is a compact, cable-free continuum robot in which local function is encoded mechanically into the body rather than delegated to a separate motorized end-effector (Zhang et al., 2 Oct 2025).

The reprogrammable miniature soft robot pushes multifunctionality further. Its two soft tentacles are not a long hyperredundant backbone, but they behave as compliant continuum beams and are modeled with a large-deflection magnetic-beam relation. The platform can switch among drug-dispensing mode, cutting mode, gripping/storage mode, and remote heating by magnetizing or demagnetizing selected low-coercivity modules while leaving hard-magnetic tentacle and body components unchanged. In locomotion mode it rolls or performs two-anchor crawling; in function modes it deploys internal mechanisms. Reported capabilities include drug-dispensing, cutting through gelatin, gripping, sample storage, and remote heating, all on a millimeter-scale body of reported length $60 \pm 1\,\text{mm}$0 (Ng et al., 19 Sep 2025).

Control remains a limiting frontier. The pancreatic MTR paper claims automated manipulation only in the restricted sense that a robotic arm moves the external permanent magnet; it does not present closed-loop image guidance, explicit path planning, or model-based continuum control (Moorcroft et al., 4 Sep 2025). The tri-leg TMR-VLA paper, although not an MTR paper in morphology, is relevant as a control architecture: it maps visual history and language instructions directly to low-level coil voltages through a calibrated field map $60 \pm 1\,\text{mm}$1, and reaches a $60 \pm 1\,\text{mm}$2 average success rate across five soft-robot primitives. A plausible implication is that future MTRs may similarly exploit vision-conditioned, actuator-side magnetic control when analytical state estimation is inadequate (Tang et al., 28 Feb 2026).

5. Fabrication, simulation, and design infrastructure

MTR research increasingly depends on fabrication methods that can encode magnetization spatially rather than merely embed magnetic particles uniformly. An automated stereolithography-based printer for magnetic soft microrobots addresses precisely this need by reorienting magnetic microparticles with a three-degree-of-freedom rotating permanent magnet and then curing the local voxel with ultraviolet light. The process is explicitly voxel-by-voxel and layer-by-layer, with each voxel assigned a magnetization vector $60 \pm 1\,\text{mm}$3. The paper reports a previous circular spot size of $60 \pm 1\,\text{mm}$4 and an updated optical design that, in simulation, achieves a square spot size of $60 \pm 1\,\text{mm}$5 using a 405 nm laser, a flat-top beam shaper, a dielectric mirror, and a focusing lens (Clancy et al., 8 Apr 2025).

This fabrication logic is directly relevant to tentacle robots because distributed magnetization governs distributed torque. The paper states the standard relation

$60 \pm 1\,\text{mm}$6

and evaluates “worm,” “gripper,” and “zipper” designs as magneto-mechanical test cases. The “worm” is particularly germane to MTRs because it is a beam-like body with uniform magnetization undergoing field-driven bending. The work stops short of an experimentally validated tentacle robot, but it establishes a manufacturing route for segment-wise or graded magnetization along slender compliant bodies (Clancy et al., 8 Apr 2025).

On the simulation side, MagRobot provides the first universal open-source simulation platform for magnetically navigated robots and is especially relevant to catheter-like or tentacle-like MTRs that can be approximated as elastic rods with tip magnetic actuation. It models continuum robots as soft rods using a one-dimensional finite element method based on Kirchhoff rod theory, includes collision and friction in deformable anatomies, and supports permanent magnets, electromagnets, and Helmholtz-Maxwell coils. Magnetic torque is expressed as

$60 \pm 1\,\text{mm}$7

with force and gradient relations also provided. The continuum-robot validation reports an RMS Euclidean position error of $60 \pm 1\,\text{mm}$8 and RMS average orientation error of $60 \pm 1\,\text{mm}$9 against phantom experiments of a magnetically steered continuum robot (Wang et al., 6 Mar 2026).

The fit between these tools and MTRs is strong but not complete. MagRobot is directly suitable when the robot is a tip-steered continuum rod; it is less directly suitable for fully soft, distributed-magnetization tentacles with strong nonlinearities, hysteresis, or fluid–structure interaction. Conversely, the automated fabrication platform supports distributed magnetization in principle but has not yet demonstrated a complete high-aspect-ratio tentacle with experimentally validated $4.644\,\text{T}$0-class features. This suggests that MTR development remains a co-design problem spanning material programming, magneto-mechanical simulation, and anatomy-aware control (Clancy et al., 8 Apr 2025, Wang et al., 6 Mar 2026).

6. Biomedical significance, limitations, and unresolved questions

The biomedical significance of MTRs follows from anatomy. The pancreatic duct, branching lumens, and other constrained cavities demand access tools that are small, compliant, and capable of curvilinear navigation. The pancreatic-delivery study explicitly motivates MTR use by the difficulty of systemic chemotherapy in pancreatic cancer and by the need for localized, controllable endoluminal drug delivery. It also places MTRs in the context of prior potential in deep brain tissue, peripheral nodules in the lungs, and the complex architecture of the pancreas and gallbladder (Moorcroft et al., 4 Sep 2025).

Several misconceptions are corrected by the recent literature. First, an MTR need not be the therapeutic depot; in the pancreatic system it is a retractable carrier for a separate responsive coating (Moorcroft et al., 4 Sep 2025). Second, MTRs are not intrinsically limited to bending; radial magnet layouts can add torsion, and reprogrammable magnetic modules can add tool-state switching and six-degree-of-freedom locomotion (Zhang et al., 2 Oct 2025, Ng et al., 19 Sep 2025). Third, the existence of experimental demonstrations does not imply mature closed-loop control or complete physical modeling. The pancreatic MTR paper lacks explicit magneto-mechanical equations, the dual-mode continuum robot provides no formal repeatability statistics, and the simulator literature remains strongest for lumped-dipole, catheter-like models rather than for distributed-magnetization tentacles (Moorcroft et al., 4 Sep 2025, Zhang et al., 2 Oct 2025, Wang et al., 6 Mar 2026).

The main limitations are consistent across papers. The pancreatic delivery platform is validated only in vitro and in phantom, uses albumin as a surrogate rather than an anticancer drug, and exhibits passive leakage, PVA-dependent release tradeoffs, and nonlinear dose scaling because wrapped fabrics self-overlap (Moorcroft et al., 4 Sep 2025). The dual-mode continuum robot demonstrates release qualitatively with ink rather than quantitative dose metering and requires combined magnetic and proximal torsional input for strong expulsion (Zhang et al., 2 Oct 2025). The reprogrammable miniature soft robot demonstrates impressive multifunctionality but is not a classical long tentacle and still depends on specialized magnetic hardware and state switching (Ng et al., 19 Sep 2025). The fabrication paper leaves experimental characterization of cure time, actual voxel geometry, and smallest realizable spot size as future work (Clancy et al., 8 Apr 2025).

A plausible synthesis is that the MTR field is transitioning from proof-of-concept navigation toward integrated intervention. The recent literature already shows three convergent directions: decoupled navigation and triggered drug release in endoluminal therapy, field-orientation-based mode switching between bending and torsion, and magnetically reprogrammable functional states in miniature soft robots. The unresolved challenge is to unify these capabilities in systems that are simultaneously anatomically navigable, quantitatively modelable, precisely controllable, and manufacturable at clinically relevant scales (Moorcroft et al., 4 Sep 2025, Zhang et al., 2 Oct 2025, Ng et al., 19 Sep 2025).