Multi‐Legged Magnetic Soft Robot for On‐Demand Targeted Delivery in Pulmonary Environments

ABSTRACT

Intrabronchial drug delivery is an important method for treating respiratory diseases. Precise and selective deep‐lung delivery of drugs is still challenging through inhalation administration and bronchoscopes, due to the passive transport and insufficient maneuverability in narrow and tortuous distal airways. Herein, a magnetic‐driven soft robot with a multi‐legged structure is developed for minimally invasive and precise targeted drug delivery in the lung. The robot exhibits multimodal and adaptive locomotion in complex airways, relying on its soft nature. A magnetothermal‐triggered and on‐demand liquid drug release mechanism is integrated onto the robot. Rolling, stick‐slip and crawling motion modes can be achieved, and the mode transition is realized to enhance the adaptivity. Precise navigation in complex environments and the distal bronchial tree of a phantom is realized. The magnetothermal‐induced drug release is also demonstrated. Ex vivo distal navigation, targeted drug delivery and retrieval of the robot in a porcine lung are validated.

This work presents a magnetically actuated multi‐legged soft robot enabling minimally invasive drug delivery in distal bronchial airways. Through self‐adaptive rolling, stick‐slip and crawling locomotion, the robot can navigate complex bronchial environments. The robot enables drug loading, magnetothermal‐triggered drug release at target sites and robot retrieval, offering a promising strategy for precision pulmonary therapy.

1. Introduction

Untethered magnetic soft robots have emerged as a promising new strategy in the field of minimally invasive delivery, owing to their advantages of remote controllability and precise navigation [1, 2, 3, 4]. These soft robots utilize the elastomer matrix embedded with magnetic particles to achieve high controllability, ensuring safe interaction with biological tissues [5, 6]. Developments of the magnetic soft robot have transitioned from uniform magnetization profile to complex and spatially programmed magnetization profile, enabling more intricate locomotion and deformation. By precisely encoding the orientation of magnetic domains within the soft robot body, complex deformation behaviors, such as bending, folding and twisting, can be pre‐designed and triggered by the external magnetic field [7, 8, 9, 10]. By dynamically adjusting the orientation, position and frequency of the actuating magnetic field, the robot can exhibit diverse locomotion modes, including crawling, swimming, jumping and tumbling [11, 12, 13], to adapt to complex environments. The synergy between programmed magnetization profiles and external magnetic control provides the robot with superior controllability and motion dexterity.

Magnetic soft robots have shown great potential in biomedical applications for various physiological environments, such as blood vessels [14, 15, 16, 17], gastrointestinal tracts [18, 19, 20, 21, 22, 23], brains [24, 25] and reproductive tracts [26]. Driven by rotating magnetic fields, a magnetic milli‐spinner can navigate high‐flow vasculature to perform clot removal, drug delivery and aneurysm treatment [27]. Magnetic multi‐layer soft robots can perform on‐demand targeted adhesion to multiple gastric ulcer sites [28]. Magnetic soft microfiberbots can achieve robotic embolization within submillimeter blood vessels owing to high steerability and multimodal shape reconfigurability [29]. For respiratory diseases, systemic administration often leads to off‐target drug accumulation, and intrabronchial administration has emerged as a promising strategy to minimize side effects [30]. However, it is still challenging to achieve precise navigation and targeted therapy in the respiratory tract due to the complex lung structure and the presence of viscoelastic mucus [31, 32]. Soft miniature robots serving as controllable active platforms that can traverse intricate bronchial tree and achieve on‐demand drug delivery in distal bronchi are highly demanded for precise therapy in the lung.

In this work, we propose a magnetically actuated soft robot designed for targeted pulmonary drug delivery. The robot features a multi‐legged structure with a central cavity for drug loading (Figure 1A). The multi‐legged structure design of the robot mitigates adhesion between the mucus layer and the robot body. Both ends of the cavity are sealed with a phase change material (PCM) to prevent off‐target release of the drug. To navigate the robot in the complex bronchial tree, the robot exhibits self‐adaptive multimodal locomotion, that can switch between rolling, stick‐slip and crawling locomotion modes to effectively traverse airways with different diameters under an external rotating magnetic field (Figure 1B). Drug delivery in the lung can also be achieved using the robot with an on‐demand release mechanism (Figure 1C). A high‐frequency alternating magnetic field is applied to melt the seals through magnetothermal effect, releasing the loaded drug precisely at the target site. Finally, the robot can be navigated back to an endotracheal tube for retrieval. Ex vivo trials are realized in a porcine lung. This work presents a drug delivery method capable of overcoming environmental barriers to access distal lung lesions, offering a promising strategy for precision pulmonary therapy.

FIGURE 1.

Mult‐legged magnetic soft robot for targeted drug delivery in the lung. (A) Schematic and optical photograph of the multi‐legged magnetic soft robot. (B) Schematic of the self‐adaptive switching among three locomotion modes in the bronchial tree with different environmental constraints under an external rotating magnetic field. (C) Precise navigation, on‐demand drug release triggered by the magnetothermal effect, and retrieval of the robot by an endotracheal tube.

2. Result and Discussion

2.1. Robot Design and Structure Optimization

By following the magnetization process described in Figure 2A(i), the robot can be programmed to have a single‐wavelength harmonic magnetization profile m along its body (indicated by the blue arrow). Under an upward external magnetic field B applied in the z‐axis, the robot experiences a magnetic torque that induces reversible elastic deformation, bending into a concave shape with an upward net magnetic moment M _net (indicated by the black dashed arrow in Figure 2A(ii)). Under a rotating magnetic field, the robot exhibits transitions between different locomotion modes in response to different environmental constraints (Figure 2B and Movie S1). To further clarify the locomotion mechanisms, step‐by‐step gait diagrams over one complete magnetic rotation cycle for the rolling, stick‐slip and crawling modes are provided in Figure S1. The net magnetic moment M _net of the robot tends to align with the external magnetic field driven by the magnetic torque. When the diameter of the channel (D) is greater than the length of the robot (L), the robot rotates along its short axis to minimize the potential energy barrier required for rolling motion. As the channel narrows into a moderately constrained region (D < L), the rolling of the robot is restricted, forcing a transition to a stick‐slip mode. This motion mode can be described in the stick stage and the slip stage. During the stick stage, the applied magnetic torque is insufficient to overcome the maximum static friction between the robot and the channel wall. Consequently, the robot maintains a fixed position, storing elastic potential energy. In the subsequent slip stage, as the magnetic field orientation continues to rotate, the accumulated actuating force exceeds the frictional resistance, leading to a rapid release of the stored energy, resulting in an instantaneous slide that generates a net forward displacement. When entering a highly constrained narrow region (D << L), physical constraints force the robot to fit inside the channel, performing a crawling motion with a metachronal wave [33].

FIGURE 2.

Multimodal locomotion mechanism and structural optimization of the soft robot. (A) Schematic of the magnetization profile and the deformation behavior under an upward magnetic field. (B) Experimental demonstration of the self‐adaptive mode‐switching in a channel with different inner diameters. The inner diameters of the three segments of the tubular environments are 8, 5 and 3 mm, respectively. Scale bar, 10 mm. (C) Robots with different leg‐leg distances along the axial direction. Optical images of three robots fabricated with different leg‐leg distances are shown (d₁ = 0.6 mm, d₂ = 1.0 mm, and d₃ = 1.4 mm). Scale bar, 3 mm. (D) Quantitative velocity analysis of three robots under varying magnetic actuation frequencies for rolling, stick‐slip and crawling modes.

To optimize the locomotion efficiency, the influence of the leg‐leg distance (d) on the velocity of the robot is investigated. Three robots with varying leg‐leg distances along axial direction (d₁ = 0.6 mm, d₂ = 1 mm, d₃ = 1.4 mm) are fabricated (Figure 2C). The locomotion performance of these three robots is systematically characterized across three locomotion modes (Figure 2D). In the rolling mode, the velocities of all the robots have minor difference when actuated by a rotating magnetic field with the same frequency, indicating that the velocity of the rolling mode is almost independent from the variations in leg‐leg distance. This is because rolling is primarily governed by the overall magnetic torque exerted on the robot body rather than the interaction between the legs and the channel wall. When the locomotion mode switches to the stick‐slip mode, the geometric advantages of the leg structure begin to emerge. In the stick‐slip mode, the robot with the smallest leg‐leg distance (d₁) has the highest translational speed, reaching a velocity of 6.9 mm/s at 2 Hz. In the crawling mode, the robot with the densest legs (d₁) reaches a maximum speed of approximately 20 mm/s at 20 Hz, which is significantly higher than the velocity of robots with larger leg‐leg distances (d₂ and d₃) by 47.3% and 90.6% at the frequency, respectively. Consequently, owing to its higher locomotion efficiency in confined spaces, the design with the minimum leg‐leg distance (d₁) is adopted for further evaluation. The key structural parameters of the robot are summarized in Table S1.

2.2. Efficient Translational Locomotion of the Robot

To validate the navigation capability of the robot in complex environments, a three‐dimensional airway phantom mimicking bronchial structure is designed, and a rotating magnet is positioned above the channel to actuate the robot (Figure 3A). The robot firstly navigates to the branching point 1 with a stick‐slip motion, and then selectively enters the upper branch (0 – 32 s). The robot exhibits efficient locomotion in the tilted branch. At branch point 2, the robot is steered into a side branch by tuning the direction angle of the rotating magnetic field, demonstrating the directional selectivity in the bifurcated channel (Figure 3A and Movie S2). Furthermore, to evaluate the effective locomotion of the robot in extremely torturous paths, a channel with five turns with different angles (30°, 60°, 90°, 120°, 150°) is designed (Figure 3B). By adjusting the position and orientation of the rotating magnet along the designed trajectory, as indicated by the black dashed line in Figure 3B, the robot can adapt to the curvature of the channel. Even in the bending state, the robot still performs effective locomotion, crawling through the curved segment and reaching the end point of this channel at t = 108.8 s (Figure 3B and Movie S2).

FIGURE 3.

Locomotion of the robot in complex environments. (A) Schematic and experimental demonstration of the robot navigation in a three‐dimensional bifurcated channel. (B) Schematic and experimental demonstration of the robot locomotion in a tortuous channel with different turning angles. (C) Schematic and experimental demonstration of the robot locomotion on a simulated tracheal model covered with simulated mucus. (D) Experimental demonstration of the robot navigation in a bifurcated channel partially filled with simulated mucus. (E) Schematic and experimental demonstration of the robot locomotion in a dynamically deformable channel mimicking physiological respiration. Scale bars, 10 mm.

To validate the locomotion capability of the robot in a physiologically relevant environment mimicking the airway with cartilages and mucus layer, the robot is deployed in a trachea model with a periodic wavy structure covered with hyaluronic acid (HA) solution, simulating mucus (Figure 3C). Despite the uneven terrain and the viscous resistance induced by the simulated mucus, the robot performs stable rolling motion to cross the model actuated by a rotating magnetic field (Figure 3C and Movie S3, 0 – 54.4 s). Furthermore, the locomotion capability of the robot in fluid‐filling environments is validated. A Y‐shaped bifurcation channel with an inner diameter of 3 mm is employed, with one branch filled with simulated mucus, which simulates the mucus plugs in the bronchi. Actuated by a rotating magnetic field, despite the existence of the air‐fluid interface, the robot rapidly crosses the air‐simulated mucus interface and achieves stable locomotion within the simulated mucus‐filled branch with a crawling motion (Figure 3D and Movie S3). This capability to navigate and traverse viscous environments validates the potential of the robot to penetrate mucus plugs in obstructed airways and reach deep lesion sites for targeted therapy. Moreover, the locomotion performances of the robot are tested in simulated mucus with different viscosities, including 1580, 8890, 26690, and 44154 cP at a shear rate of 1 s^{− 1}. As shown in Figure S2A, with increased HA concentration and viscosity, the locomotion velocity of the robot decreases due to the increased drag resistance. It is noted that, even in simulated mucus with a viscosity of 44154 cP, the robot can still maintain effective locomotion with a velocity of approximately 0.2 mm/s (Figure S2B).

To evaluate the locomotion of the robot in dynamic bronchial environments, a soft stretchable channel is prepared (Figure 3E). By cyclically stretching the model, the inner diameter undergoes periodic variations, alternating between d₁ = 4 mm (initial state) and d₂ = 2.8 mm (stretched state), resembling the physiological bronchial expansion‐contraction rhythm (Figure 3E and Movie S4). Under the actuation of a rotating magnetic field, the robot achieves a net displacement (D_net1) of 4.1 mm per cycle. After 11 cycles, the cumulative net displacement (D_net11) reaches 60.6 mm. To further evaluate the influence of respiratory motion and tissue depth to the effectiveness of magnetic actuation, experiments under varying magnet‐robot distances are conducted. The experimental setup mimics the change of magnet‐robot distance with the respiratory motion of the thoracic geometry (Figure S3A). The results indicate that the robot actuated by the proposed magnetic field can still be navigated through the channel effectively, despite the dynamically changed magnet‐robot distance (Figure S3B). It is noted that, in crawling mode, the robot could still be actuated at 12 cm with a velocity of 0.12 mm/s (Figure S3C). The result demonstrates the environmental adaptability of the robot to the dynamic environment, showing the potential of the robot for application in the real lung. To further evaluate the feasibility of multi‐robot deployment, experiments including simultaneous actuation, sequential deployment and independent actuation are conducted. As shown in Figure S4A, three robots are simultaneously actuated and move from left to right from 0 to 54.6 s. The robots show similar locomotion behaviors, indicating that relatively homogenous robot locomotion can be achieved under the same magnetic actuation field and similar geometrical constraints. The result in Figure S4B shows that sequential deployment and geometrical constraints of the environment can be used to actuate multiple robots to different target regions. Figure S4C shows that heterogeneous multi‐robot locomotion can also be achieved by designing specific magnetic fields using multiple magnets. Two robots in parallel channels are simultaneously driven in opposite directions using two rotating magnets.

2.3. On‐Demand Magnetothermal‐Triggered Drug Release of the Robot

To precisely deliver drugs to the lesion site and ensure therapeutic efficacy, both ends of the cavity filled with drugs are sealed to prevent leakage during robot navigation. The phase change material (PCM) composed of hexadecanol, oleic acid, and iron oxide (Fe₃O₄) particles is selected as the sealing material (Figure 4A). By mixing hexadecanol with oleic acid, the phase transition temperature of the PCM can be adjusted to a range that is safe for the living tissue. Meanwhile, the Fe₃O₄ microparticles can heat the PCM under a high‐frequency alternating magnetic field, triggering the melting of the seals for on‐demand drug release. The phase transition temperature of the seals is investigated through small‐amplitude oscillatory shear tests on a rheometer. Figure 4B shows that as the temperature increases from 25°C to 45°C, both the storage modules (G′) and loss modules (G″) decrease, and remain nearly constant above 45°C. During heating, G″ exceeds G′ at 42.3°C, indicating that the phase transition from solid to the liquid state occurs at 42.3°C. The seals can remain solid state at body temperature (37°C) and transition to liquid state at around 43°C (Figure S5). Relying on the thermoplastic property of the PCM, the seals can be easily remolded into various shapes, such as hexagram, diamond star, and cross shapes (Figure 4C). The overall heating process can be precisely controlled by adjusting the Fe₃O₄ concentration and heating duration. The magnetothermal heating efficiency is evaluated with different Fe₃O₄ concentrations. As shown in Figure 4D, a higher Fe₃O₄ concentration leads to a more rapid temperature increase of the PCM. The PCM containing 5 wt.% Fe₃O₄ reaches approximately 42.3°C within 10 s, whereas the PCM containing 1 wt.% Fe₃O₄ remains around 33°C, which is below the phase transition temperature of the PCM, during the same period.

FIGURE 4.

On‐demand magnetothermal‐triggered drug delivery system. (A) Schematic of the drug‐loaded robot sealed with the PCM. (B) Storage modulus and loss modulus of the PCM are recorded in oscillatory temperature sweeps. (C) The seals can be remolded to various shapes at 42.3°C, such as hexagram, diamond star and cross shapes. Scale bar, 5 mm. (D) Magnetothermal heating profiles of PCM composites with different Fe₃O₄ concentrations under a high‐frequency alternating magnetic field. (E) Comparison of drug protection performance between unsealed and sealed robots. Scale bars, 5 mm. (F) Corresponding concentration‐time curves quantifying the leakage prevention capability of the seal. (G) Optical photographs of the dye release process triggered by a high‐frequency alternating magnetic field. Scale bar, 5 mm. (H) Magnetothermal‐triggered payload release profile. (I) Optical photographs demonstrating the controllable locomotion in a maze and magnetothermal‐triggered dye release at the targeted region. Scale bars, 10 mm.

To verify the sealing efficiency, the drug protection capability of the robot is evaluated using Rhodamine B as the model dye. The group without seals exhibits rapid leakage due to the open cavity, with significant dye diffusion observed within the first 10 min (Figure 4E). In contrast, the robot equipped with seals demonstrates superior protection performance (Figure 4E). After a 30‐min observation, no significant leakage of the dye in the surrounding medium is observed. This observation is further corroborated by quantitative measurements using a microplate reader. Experimental results demonstrate that in the group without seals, the dye release concentration in the external medium increases rapidly to approximately 6 µg/ml within the first 10 min, then stabilizes at approximately 8 µg/ml. The robot with seals demonstrates superior drug‐protection capabilities. The dye release concentration remains consistently below 2 µg/ml within 30 min, which is significantly lower than the release levels observed in the group without seals at the corresponding time (Figure 4F).

The magnetothermal‐triggered dye release process is illustrated in Figure 4G. Initially, the surrounding medium remains clear, indicating negligible dye leakage at room temperature due to the encapsulation of the seals. By applying a high‐frequency alternating magnetic field, the embedded Fe₃O₄ particles heat the seals to their phase transition temperature, and the seals are melted to perform dye release within 2 s. As the time increased from 2 to 60 s, a significant diffusion of the dye is observed, confirming that the dye release can be triggered via the magnetothermal effect. Quantitative analysis in Figure 4H shows sustained release of both Rhodamine B and doxorubicin (DOX) over 30 min after the magnetothermal triggering. Rhodamine B reaches a cumulative concentration of approximately 8 µg/mL, while DOX reaches approximately 7 µg/mL. Multiple robots loaded with DOX solution are prepared, and the delivery process is repeated one by one in sequence. After one robot completes the delivery, it is retrieved, and the next robot continues for the next round of drug delivery. The results in Figure S6 show that the cumulative released DOX concentration reaches approximately 13 µg/mL after 3 deliveries, indicating that the total delivered amount can be increased through repeated deployment and release. A maze with bifurcated structures and turns with different angles is designed to mimic intricate channels in the human lung. The central area of the maze (enclosed by blue dashed lines) is designed as the navigation endpoint and the dye release region. Under a rotating magnetic field, the robot loaded with dyes effectively navigates and successfully reaches the endpoint of the maze, followed by dye release triggered by a high‐frequency alternating magnetic field (Figure 4I and Movie S5).

2.4. Robot Navigation in a Phantom of the Bronchial Tree

To evaluate the navigation capability of the robot in a simulated physiological environment, a three‐dimensional phantom of human bronchial tree is employed. The phantom and complete pre‐planned trajectory is illustrated in Figure 5A and Figure S7. In Region 1, the robot is almost unconstrained by the phantom and shows rolling locomotion mode, which exhibits the highest velocity. Under the actuation of an external rotating magnetic field, the robot follows the pre‐planned path (highlighted by the blue area), passing by the branches highlighted by the red area, and traversing more than half of the whole path within 17.1 s (Figure 5B and Movie S6). When the robot makes locomotion in Region 2, the spatial confinement from the phantom walls triggers a transition from the rolling mode to the stick‐slip mode. Under magnetic control, the robot is accurately steered into the branch highlighted by the blue area (Figure 5B and Movie S6). As the robot moves into Region 3, the environment becomes highly constrained, triggering the mode switching from stick‐slip mode to crawling mode (Figure 5B and Movie S6).

FIGURE 5.

Multimodal navigation in a phantom of the bronchial tree. (A) Overview of the phantom and pre‐planned navigation trajectory. (B) Optical photographs showing multimodal locomotion of the robot in the phantom, including rolling, stick‐slip, and crawling modes. Scale bars, 10 mm.

2.5. Robot Navigation, Targeted Drug Delivery, and Robot Retrieval in an Ex Vivo Lung

The navigation, payload delivery and robust retrieval of the robots are performed in an ex vivo porcine lung, as shown in Figure 6. The schematic of the ex vivo experiments is shown in Figure 6A. The robot is first deployed to the main bronchus through a bronchoscope, and the robot is then magnetically actuated in the bronchial tree to track the pre‐planned trajectory across region 1‐3, as shown in Figure 6A. Bronchoscope is applied for real‐time observation. In a segment of the main bronchus (Region 1), the robot applies a rolling gait to achieve efficient locomotion (t = 0 – 12.3 s). The precise navigation of the robot avoids it from entering branches A1 ‐ A4 (Figure 6B and Movie S7). As the robot enters Region 2 with a smaller diameter, the constrained environment enables the robot to actively switch from rolling mode to stick‐slip mode, passing by branches B1, B2, and B3 (Figure 6C and Movie S7). After reaching the confined distal bronchi (Region 3), the robot switches to a stable crawling gait with metachronal wave, and finally reaches the target area (Figure 6D and Movie S7).

FIGURE 6.

Robot navigation, targeted payload delivery, and robot retrieval in an ex vivo lung. (A) Schematic of the experimental setup utilizing a bronchoscope to visualize the operation of the robot within an ex vivo lung. (B) Schematic and experimental demonstration of robot locomotion in Region I. Scale bar, 5 mm. (C) Schematic and experimental demonstration of robot locomotion in Region II. Scale bar, 5 mm. (D) Schematic and experimental demonstration of robot locomotion in Region III. Scale bar, 5 mm. (E) Bronchoscopic view showing the magnetothermal‐triggered dye release at the target region. Scale bar, 10 mm. (F) Merged images of H&E staining and fluorescence images of the lung tissues collected at three cross sections after the targeted dye delivery. Cross section I: scale bar, 5 mm. Cross section II: scale bar, 5 mm. Cross section III: scale bar, 1 mm. (G) Total fluorescence intensity of the lung tissues at cross sections I to III. (H). Schematic and experimental demonstration of the robot retrieval with an endotracheal tube after the dye release. Scale bar, 5 mm.

To evaluate the payload delivery performance of the robot, the cavity of the robot is loaded with fluorescent dyes (Dil, red fluorescence) and blue dyes. Once the robot reaches the target area, the on‐demand payload releasing mechanism is activated through melting the seals using magnetothermal effects. It can be observed from Figure 6E that there is no dye release before heating. After the dye delivery process, it can be observed that the blue dye is attached on the bronchial wall. The lung tissues along the pathway (cross sections I and II) and target site (cross section III) are subsequently collected. By comparing the merged Hematoxylin and eosin (H&E) staining fluorescence image of the tissues from these three cross sections, the fluorescence intensity of cross sections III is significantly higher than those of cross section I and II. The results indicate that the successful dye release at the target site is achieved with high precision, and negligible dye leakage during the navigation is observed (Figure 6F and G, and Figure S8). To further validate the delivery capability using a representative therapeutic drug, targeted DOX delivery experiments in an ex vivo porcine lung are conducted. During navigation in the ex vivo lung, the robot shows rolling, stick‐slip and crawling modes in different airway regions (Figure S9A). Bronchoscopic images before and after the release process are recorded. Visible red DOX deposition is observed on the bronchial wall after magnetothermal triggering, indicating the successful DOX release (Figure S9B). Cross sections resected from the navigation path outside the target release region show negligible DOX fluorescence signal (Figure S10A,B). Fluorescence analysis showing DOX distribution of the tissue resected from the drug‐release region is presented in Figure S10C, further demonstrating the tissue‐level DOX delivery.

To ensure safety, the retrieval capability of the robot is verified. After dye release, an endotracheal tube is used to retrieve the robot (Figure 6H and Movie S7). The yellow dashed line in Figure 6H outlines the distal opening of the endotracheal tube. Under the control of a rotating magnetic field, the robot successfully enters the endotracheal tube (0 ‐ 6.6 s), demonstrating that the robot can be safely removed from the lung after completing the targeted payload delivery, eliminating the risk of long‐term retention. To further assess the thermal safety of the magnetothermal‐triggered payload release process, simulation of local temperature distribution, cell viability after thermal exposure and tissue‐level histological analysis are performed. The simulation results show that the temperature increase is mainly limited to the PCM seal region, with minor heat diffusion to the surrounding tissue (Figure S11). Human bronchial epithelial cells are incubated at 37°C and 42.3°C for 10 min, and live/dead staining is subsequently conducted. These results suggest that short‐term exposure at 42.3°C had no obvious effect on bronchial epithelial cell viability, indicating cellular‐level safety of the proposed strategy (Figure S12). Hematoxylin and eosin (H&E) staining of tissue resected from the payload release region shows preserved tissue morphology without obvious structural damage, validating the safety of the magnetothermal‐triggered payload release process (Figure S10C).

3. Conclusion

In this work, we design a soft robot with high adaptability and motion dexterity, which is capable of traversing bronchial environments with mucus. The robot can reach distal bronchi and achieve targeted drug delivery. The robot features a programmable harmonic magnetization distribution and an optimized leg array design. It has a self‐adaptive mode‐switching mechanism, which enables it to autonomously switch between rolling, stick‐slip, and crawling gaits according to different environmental constraints, thereby maximizing the locomotion efficiency. The robot can also maintain locomotion effectiveness under dynamic deformation and motion of external environments, mimicking the expanding and contracting motion of the lung caused by inhalation and exhalation. The robot can carry desired drug inside its body cavity, and through deploying temperature‐sensitive seals, the loaded drug can be protected, avoiding off‐target distribution of the drug. By applying a high‐frequency alternating magnetic field, the seals are melted, and on‐demand drug release can be achieved. The locomotion capability of the robot is validated in an ex vivo lung model, that the robot can reach distal regions via self‐adaptive multimodal locomotion. Drug delivery and retrieval are also demonstrated.

The key features of the robot may help reduce the relevant risks caused by ventilation and perfusion. For ventilation, the soft‐bodied robot has a small width of approximately 2.8 mm, which reduces the possibility of airway blockage during navigation. For perfusion, the soft body of the robot can better adapt to local airway geometry and distribute contact forces during locomotion. Therefore, the robot is expected to minimize local tissue deformation, reduce the risk of compressing nearby pulmonary microvasculature, and avoid interference with efficient gas exchange during short‐term navigation and delivery. Moreover, the robot is designed to be retrieved through the endotracheal tube after payload release, which reduces the risk of long‐term retention in the airway. In summary, the proposed soft robot provides a minimally invasive strategy for accessing hard‐to‐reach lung lesions, offering a new paradigm for precision treatment within the complex human respiratory tract.

4. Experimental Section

4.1. Fabrication of the Magnetic Multi‐Legged Robot

As illustrated in Figure S13, the master mold with a multi‐leg structure and the inner mold with a cylindrical convex are fabricated by 3D printing technology. Then, the polydimethylsiloxane (PDMS) is poured onto the master mold, and the outer mold with micro‐leg grooves is obtained after curing and demolding. To fabricate the robot body, NdFeB microparticles and PDMS are mixed thoroughly at a mass ratio of 1:1. The mixture is placed in a vacuum chamber for 30 min to remove bubbles, and then injected into the external mold. To ensure that the mixture completely fills the fine leg structures, it is vacuumed again for 30 min. The inner mold is then assembled to the outer mold, and the mold assembly is placed in an oven and heated at 80°C for 1 h to cure. After curing, the robot body is removed from the mold. The remanence and coercivity of the magnetic robot body are almost 49.3 emu/g and 8.6 kOe (Figure S14). To form the cavity, Ecoflex 00‐30 is spin‐coated on a glass slide using a spin coater. The demolded robot body is inverted and tightly attached to the uncured Ecoflex layer, and heated again at 80°C for 1 h to achieve bonding and curing. Then, the excess film is trimmed, and the sample is peeled off from the glass slide to obtain a multi‐legged robot with a central cavity. Finally, it is rolled into a cylindrical configuration and magnetized in a pulsed magnetic field to complete the preparation of the robot.

4.2. Fabrication of Sealing Materials

The magnetic phase change material for sealing is composed of superparamagnetic Fe₃O₄ nanoparticles and a phase change matrix composed of oleic acid and hexadecanol. To prepare the phase change material (PCM) with a melting point of 42.3°C, 100 mg of oleic acid and 250 mg of 1‐hexadecanol are mixed thoroughly. In the preparation of magnetic sealant, the synthesized Fe₃O₄ particles are uniformly dispersed into the molten PCM matrix, giving the material magnetothermal response characteristics.

4.3. Magnetic Actuation Setup

The actuation system utilizes a robot arm (CR16, Dobot Robotics) equipped with a motor‐driven cylindrical permanent magnet (height: 50 mm, diameter: 70 mm). The setup of this magnetic actuation system is illustrated in Figure S15.

4.4. Magnetothermal Experiment Setup

The magnetothermal‐triggered payload release is achieved by a high frequency heating equipment (SPG‐10A‐II, Shenzhen Shuangping, China). The robot is placed inside the center of the copper coil (6‐turn copper coil with an 8 cm outer diameter). The temperature variations of the sealing materials are recorded by an infrared thermography camera (FLIR, E76).

4.5. Rheological Test

The phase transition temperature of the sealing materials is measured using a rheometer (Anton Paar MCR302e, Austria). Small amplitude oscillatory shear measurements are used on the sample using 1% strain amplitude and an angular frequency of 1 Hz. During the phase transition investigations, the thermal profile is programmed to ramp between 25 and 55°C, applying a steady increment of 1°C per minute.

4.6. Ex Vivo Experiment

The porcine lungs are obtained from a local slaughterhouse within 48 h. During transportation, the organs are preserved with ice (approximately 2°C) to retard tissue changes. The lungs are gently and simply cleaned with water to reduce damage to the mucus layer. The movement of the robot is observed using a bronchoscope when the targeted delivery experiment is performed in the ex vivo porcine lungs. After the payload release is completed, the robot is recovered using an endotracheal tube.

4.7. Cell viability Test

The viability of airway epithelial cells is investigated at the phase‐transition temperature of PCM seals. Human bronchial epithelial cells (HBE135‐E6E7) are incubated at 37°C and 42.3°C for 10 min. Subsequently, cells are stained with calcein acetoxymethyl ester (calcein AM) and propidium iodide (PI) for 15 min. Fluorescence images of the stained cells are captured using an inverted fluorescent microscope (ECLIPSE Ti2, Nikon Corporation) and then analyzed using ImageJ.

4.8. Fabrication of Simulated Mucus

Hyaluronic acid (HA) solutions with different concentrations are prepared as simulated mucus by slowly adding HA powder into deionized water, followed by continuous stirring for 24 h at room temperature to obtain HA solutions. The viscosities of HA solutions are measured using a rheometer (Anton Paar MCR302e, Austria).

Funding

This work was financially supported by the National Key R&D Program of China under Project No.2022YFA1207100, the Shenzhen Science and Technology Program under Projects No. JCYJ20241202124015021 and JCYJ20250604141050027, the Shenzhen Institute of Artificial Intelligence and Robotics for Society under Project No. BN00202312037‐1D, the China Merchants Group funding under Project No. BN00202312037, the Longgang District Shenzhen's “Ten Action Plan” under Project No. LGKCSDPT 2024002 and 2024003, the National Natural Science Foundation of China (No. 52305610), the Guangdong Basic and Applied Basic Research Foundation (No. 2024A1515010107), and the Guangdong S&T program 2024A0505040014.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File 1: adma73957‐sup‐0001‐SuppMat.docx.

Supporting File 2: adma73957‐sup‐0002‐MovieS1.mp4.

Supporting File 3: adma73957‐sup‐0003‐MovieS2.mp4.

Supporting File 4: adma73957‐sup‐0004‐MovieS3.mp4.

Supporting File 5: adma73957‐sup‐0005‐MovieS4.mp4.

Supporting File 6: adma73957‐sup‐0006‐MovieS5.mp4.

Supporting File 7: adma73957‐sup‐0007‐MovieS6.mp4.

Supporting File 8: adma73957‐sup‐0008‐MovieS7.mp4.

Data Availability Statement

The data that supports the findings of this study are available in the supplementary material of this article.