Millimeter‐Scale Soft Continuum Robots for Large‐Angle and High‐Precision Manipulation by Hybrid Actuation - Zhang - - Advanced Intelligent Systems

2.1 Hybrid‐Actuation Millimeter‐Scale Soft Continuum Robot Design

To obtain a large bending angle and predict the posture accurately, a tendon‐driven mechanism is used as the actuation for the robot during the period of navigating to the pathological area. For the sake of achieving the steering capacity along the plane perpendicular to its central axis, two pairs of antagonistic tendons are applied. Four through‐holes are constructed within the wall to implement the tendons (a stainless steel wire with a diameter of 100 μm), placed at 90° between each other. To adapt to the circumstances within vessels in vivo, the outer diameter of the skeleton is designed as 3.0 mm, and a hollow structure (Figure S1a, Supporting Information) along the bending directions is designed to increase the compliance of the robot. Furthermore, the designed large inner lumen promotes the robot to explore versatile capacities via encapsulating varied surgical tools.

Here, the micro‐3D printing system (nanoArch P140, BMF Material Technology Inc.) with a printing precision of 10 μm is used to fabricate such a delicate robot skeleton (Figure 1c‐i). For obtaining enough strength and elasticity of the structure, a type of ductile photosensitive resin (HD, BMF Material Technology Inc.) with an elastic modulus of 3.6 GPa is utilized as the printing material. As shown in Table S1, Supporting Information, the material possesses the breaking elongation rate of 24.3%, leading to a low risk of breaking off for large deformation. The skeleton is printed with an outer/inner diameter of

 mm, hollow section of

 μm, and through‐holes of

 μm (Figure S1b, Supporting Information). As a result, an ultrathin wall of 300 μm is obtained, which is the thinnest wall among all reported continuum tendon‐driven robots.^[22] Moreover, the large inner‐to‐outer ratio, i.e., 0.8, grants the robot high superiority in encapsulating versatile small surgical tools while maintaining a tiny size (Figure 1b). Meanwhile, the hollow skeleton allows the robot to easily achieve a large bending motion while keeping enough stiffness.

To achieve high‐precise manipulation and maintain friendly interaction with the tissue, we decorate a thin layer of ferromagnetic composite elastomer skin (≈100–150 μm) on the surface of the catheter skeleton and use magnetic actuation as the second module of hybrid actuation for the robot. Here, the microiron particles (diameter: 5 μm) (Guangzhou Metallurgy Co., Ltd.) are chosen as the base material for magnetic response, and the silicone rubber, i.e., Ecoflex #20 (Beijing Angelcrete Art Landscaping Co., Ltd.), is chosen as the elastomer matrix of the composite. First, the nonmagnetized iron particles and uncured Ecoflex with a prescribed mass fraction are mixed to obtain the homogeneous ferromagnetic composite elastomer (Figure 1c‐iii). Second, the elastomer is brushed on the surface of the skeleton to form the skin via spin coating, avoiding the redundant elastomer permeating inside the skeleton (Figure 1c‐ii). After that, a uniform magnetic field is applied with the direction along the robot central axis to rearrange the iron particles, and the composite is cured in air (Figure 1c‐ii). As a result, a layer of ferromagnetic elastomer skin (thickness of ≈100–150 μm) is uniformly covered on the 3D‐printed skeleton. The optical image of the fabricated soft continuum robot is shown in Figure 1d. With an excellent breaking elongation rate of 620% and a hardness of shore of 20 A (Table S1, Supporting Information), Ecoflex grants the elastomer skin with enough softness and elasticity. In addition, the robot can obtain better actuation performance when it is subjected to an external magnetic field due to the rearrangement of the iron particles.

2.2 Large‐Angle Posture Control by Tendon‐Driven Robots

For soft continuum robots,^[18] due to only one bending degree on the catheter tip, they have to rotate the entire body to achieve full‐orientation bending motion. Such kind of manipulation would result in the risk of hurting the surface of the tissue. For more friendly interaction with the tissue surface, two pairs of antagonistic tendons are designed inside the tube wall in our proposed robot. With different actuation strategies of pull–release on each pair of tendons (i, ii, iii, and iv), the robot could achieve varied postures (Figure 2a), avoiding the rotation motion of the whole body.

The bending difference of catheter tip (without elastomer skin) between the actuation of the single pair (pulling i while releasing iii) and double pair (pulling i & ii while releasing iii & iv) of tendons is investigated first. The deflection against elongation is shown in Figure 2b. These two strategies perform a similar trend, which indicates that a circular trajectory perpendicular to the advancing direction could be reached through the cooperation of all four tendons with a proper pull–release strategy. In addition, to achieve the same deflection caused by the actuation of single‐pair tendons, less elongation is needed for the adjacent pairs of tendon‐driven strategies. Given two rotation angles of

and

corresponding to the actuation of the single and adjacent pairs of tendons, respectively, the theoretical elongation shall be

and

, where the consistency between the experimental results and the theoretical expression of Equation (1) is found. Furthermore, with the elastomer skin (embedded with different mass fractions of iron particles) coated on the skeleton, the similar difference between the actuation of single and adjacent pairs of tendons can also be noticed in the experimental results (Figure 2c).

For obtaining the workspace of the proposed soft continuum robot, we conduct actuation on a single pair of tendons on the robot without elastomer skin. The experimental measurements of deflection (Figure S2a, Supporting Information) and bending angle (up to about 100°) (Figure S2b, Supporting Information) of the catheter tip are obtained, indicating the excellent compliance of our robot. After the corresponding analytical model of the deflection is established, we can further predict the continuous surface in 3D space (Figure 2d) of the catheter tip utilizing the structural symmetry. Similarly, the same workspace can be achieved for robots with elastomer skin. Note that the soft continuum robot coated with elastomer skin leads to a higher shear modulus (see Figure S2c, Supporting Information for detail). The workspace proves a good steering performance of the proposed robot, suggesting that a complicated circumstance inside the human body can be adapted.

2.3 High‐Precision Positioning by Magnetic‐Driven Robots

With a varied mass fraction of iron particles embedded in the elastomer skin, the continuum robot would perform differently in both magnetic and mechanical properties. To obtain better actuation performance for later precise positioning under a magnetic field, an optimal design on the parameter of the mass fraction should be achieved.

As for the soft robot with ferromagnetic elastomer skin, both M and G would be affected by the mass fraction of iron particles. For the ferromagnetic particles embedded in elastomer skin, the relationship between the induced magnetization and mass fraction can be presented as

. According to Equation (3), the deflection caused by an external magnetic field is mainly affected by both magnitudes of the external magnetic field and its material, i.e.,

. For a certain mass fraction of iron particles, there should be a quadratic trend between the deflection and the magnitude of the external magnetic field, as magnetization could be approximately viewed as a linear function of the latter one before magnetic saturation.^[23] Given a specific external magnetic field B, there should be an optimal solution for the mass fraction at which a maximal deflection is obtained.

The experimental measurement and the corresponding curve fitting of deflection under a magnetic field are shown in Figure 3b. For each robot with a certain mass fraction of iron particles, the absolute value of deflection shows a similar quadratic relationship with the magnitude of the external magnetic field, in accordance with Equation (3) of the theoretical model. In addition, according to Figure 3b, the robot embedded with a higher mass fraction of iron particles generates a bigger deflection under the same magnitude of a magnetic field, especially when the mass fraction is larger than 40%. The relationship between deflection and particle mass fraction as proportional, i.e.,

, can be concluded. It further suggests the material‐related variable M/G presents the same changing trend with particle mass fraction, i.e.,

. Taking the relationship between the shear modulus and mass fraction of iron particles, i.e.,

, into consideration, it can be concluded that the increase of induced magnetization M of the robot is obviously faster than that of shear modulus G as the mass fraction of iron particles increases. Also, the actuation angle of the robot with a higher mass fraction experiences a faster uprising for a specific external magnetic field B (Figure 3c), indicating the proportional relationship

. It is consistent with the trend of deflection. Moreover, the composite elastomer with a high mass fraction of iron particles (≥70%) is difficult to be coated on the robot surface (see Figure S2d, Supporting Information for detail). Therefore, the optimal mass fraction of iron particles distributed in the elastomer skin could be viewed as 60% for the proposed soft continuum robot.

With the optimal mass fraction of iron particles embedded in elastomer skin, the merit of the high‐precision positioning of the proposed soft continuum robot (the prototype is shown in Figure S1, Supporting Information) is demonstrated under a finely controllable magnetic field. According to Figure 3d, the absolute value of deflection shows an approximately quadratic relationship with the magnitude of the external magnetic field. The static positioning accuracy of the catheter tip can be achieved up to ≈2 μm, under the guidance of both the magnitude and direction of the applied magnetic field.

2.4 Active Steering and Positioning for Searching the Pathological Area Within the Vessel

Hereafter, the primary capabilities of steering through the complex constrained environment (tendon driven) and high‐precision positioning (magnetic actuation) of the proposed soft continuum robot are demonstrated. Moreover, additional functions granted by the functional core assembled inside the lumen of the robot are proved.

The skeleton of the prototype is fabricated by micro‐3D printing technology (Figure 1c‐i). The ferromagnetic composite elastomer, composed of both iron particles and Ecoflex, is coated onto the robot surface to achieve a magnetic response (Figure 1c‐ii). The experimental process of steering through a 3D branched tunnel (see Figure S4a, Supporting Information for detail) of the prototype is shown in Figure 4a. Initially, the robot moves straight inside the tunnel under the push force applied at the proximal end. Upon approaching the first bifurcation, the robot is steered toward the north east by pulling both the upper and right tendons and releasing the rest of the two tendons. After accessing about 15 mm, it is retracted to the bifurcation point and keeps advancing to the second bifurcation point. The robot is actuated to access the left branch by pulling the left tendon and releasing the right one, posing an angle of 75° with the trunk. Later, the robot is retracted and performs navigation toward the right for 35° using a strategy opposite to the previous one. After another move back, the robot moves forward. Finally, the robot performs an upforward posture at about 25° by pulling the upper tendon and releasing the lower one.

The four main postures of the robot under the tendon‐driven mechanism during the steering process within a branched tunnel are shown in Figure 4b. Both the position and bending angle of the distal tip of the assembled prototype during the steering process are collected and shown in Figure 4c. According to the experimental measurement, we can easily find that there are four periods of noticeable variations of bending angles corresponding to the four postures. There are small increments along X‐axis and a relatively big increase along the Z/Y axis when the robot accesses the first two branches. A significant variation along both X and Y axes is experienced for the robot during the third branch. The incremental trend along the coordinate system agrees with the geometry contour of the experimental setup (Figure S4a, Supporting Information). The demonstration proves that our soft continuum robot has superior adaptability and navigation capability in a complex constrained environment.

The experimental process of high‐precision positioning of the proposed robot under magnetic actuation is shown in Figure 4d. During the advancing process, the robot is actuated by an external magnetic field to achieve micrometer‐scale navigation. The red dashed curve denotes the desired trajectory whereas the green curve represents the experimental trajectory of the tip. According to the two bow‐like trajectories, it can be concluded that the overall conformity is good. The schematic illustration of magnetic actuation for the robot is shown in Figure 4e, and the tip position data are shown in Figure 4f (Figure S3d, Supporting Information shown detailed magnetic field measurement). For the two experiments of positioning, the root mean square error (RMSE) for the actual trajectories relative to the desired one has been calculated as 10.2 and 9.7 μm, respectively. The tracking precision of around 10 μm grants the proposed robot’s capability of conducting micromanipulation in vivo.

2.5 High‐Precision Manipulation for Target Therapy and Nasopharyngeal Sampling

For further extending the application of the robot, we demonstrate the additional function of high‐precision micromanipulation within a vessel model (the dimension details are shown in Figure S4b, Supporting Information) by incorporating microtools inside. The experimental process of searching the pathological area (the purple block) within the vessel model is shown in Figure 5a. The proposed robot is first steered to the left vascular branch by pulling the left tendon and releasing the right one. As there is no target in this branch, the robot is retracted and actuated to the right, where the tendon‐driven strategy is opposite to the former one. After getting close to the pathological area, the microtools would protrude to conduct the corresponding micromanipulation.

The schematic illustration of the soft continuum robot that approached the pathology area within a vessel model is shown in Figure 5b. There are microscale tools, including microknife and needle, incorporated within the lumen of the robot. As a result of the magnetic response of the catheter robot, the microtools protruded will experience corresponding motion. Figure 5c shows the experimental process under magnetic actuation. The distal tip of the microknife has achieved a linear trajectory of about 300 μm under the actuation of an external magnetic field. As for the precise manipulation of the robot, it can be potentially applied to accomplish target therapy, such as target injection and target ablation (Figure 5d). The sphere‐like target experienced a visible color change after microinjection via the needle (Figure 5d‐i). Another seed‐like target has been cut into two pieces with the knife (Figure 5d‐ii). All the geometry parameters of tools, including sheath, needle, and knife, are shown in Table S2, Supporting Information.

Considering the current outbreak of COVID‐19, the potential application of nasopharyngeal sampling of the proposed soft continuum robot by incorporating a swab inside has been further demonstrated (Figure 5e). Compared with the oropharyngeal swab sampling, nasopharyngeal sampling would bring better tolerability for patients, i.e., not easily inducing coughing. It further results in a relatively long time for obtaining specimens and a lower risk of infection for medical staff.

As shown in Figure 5e, the green dashed curve passing through the middle nasal meatus represents the desired trajectory inside the nasal cavity (the dimension details are shown in Figure S4c, Supporting Information). The robot is first inserted through nostrils, along the direction parallel with the bridge of the nose. Once the swab tip reaches the middle nasal meatus, about 15 mm above the position of the snout, the robot is actuated by an external time‐varying magnetic field to achieve the probing trajectory according to the desired one. With a gentle push force applied at the proximal end of the soft continuum robot, the swab tip finally reaches the position of the nasopharynx. At this moment, the axis of the catheter tip poses about 91° with the bridge of the nose. After advancing and retracting slightly, the specimens adhered to the nasopharynx could be transferred to the surface of the swab tip. Finally, the soft continuum robot with the sampled swab can be retracted genteelly along the original inserting route. The enlarged detail of the swab tip during the sampling process is shown in Figure 5f, which demonstrates the good performance of adhering specimens.

2.6 Discussion

To implement surgeries within the complex constrained vessels, a small and soft body is the initial inherent requirement for continuum robots. A larger tender range‐steering ability is also needed for robots to achieve the ability to easily pass through during the period of searching for the pathological area. When conducting surgery manipulation, a higher precision and smaller dynamic error are beneficial for improving safety. Furthermore, a large inner lumen shall be granted for the robot to implement versatile manipulation by encapsulating varied surgical tools. However, the existing continuum robotic technologies encounter the difficulty of achieving the comprehensive performance of large‐angle steering and high‐precision manipulation while maintaining small‐scale size.

For conventional continuum robots, most actuation techniques own a contour dimension over 3.0 mm, e.g., tendon driven, or even with a centimeter scale, e.g., fluidic‐driven and dual‐mode actuation. Although the soft continuum robot actuated by magnetic field or smart material can be designed at a small scale, the small inner lumen less than 1.0 mm limits its functions. To obtain a larger bending angle, the fluid‐driven robots show their excellent performance (usually >150°), and the robots actuated by a magnetic field also can present the bending angle ranging from 90° to 180°. In addition, a steering angle up to 80° is reported as the bottleneck for soft robots actuated by the smart material. For positioning precision, the soft continuum robots actuated by fluid or smart material presents poor performance due to the complex control model and inherent hysteresis, respectively. Benefiting from the classic control model, many tendon‐driven robots can achieve a position precision of about 2.0 mm. The magnetic actuation is reported as the most precise actuation method for soft continuum robots and has achieved tracking error up to 0.3 mm under a controllable magnetic field.^[13] However, magnetic actuated robots show small resistance to an external force. To achieve comprehensive performance in vivo, tendon‐driven and magnetic actuation are combined in our proposed soft continuum robot. Thanks to 3D printing technology, the proposed robot here can be developed with an outer/inner diameter of 3.3/2.4 mm. Grant the potential of accessing smaller space and carrying varied surgical tools to implement different microinvasive surgeries. Under the actuation of tendons, the proposed robot can achieve a bending angle up to ≈100°, which is slightly improved, compared with many of the existing robots with smart material. In addition, the strong resistance capability of the tendon‐driven mechanism ensures the feasibility of navigation inside a constrained vessel for the proposed soft continuum robot. Furthermore, with the guidance of a controllable magnetic field, the demonstrated dynamic error (≈10 μm) for the proposed robot is 30‐fold improved than the state of art. Upon reaching near the pathological area, the active part can be actuated by a magnetic field to achieve highly precise positioning and manipulation.

Despite the existing error bar of static positioning ranging from 2 to 4 μm (1–2 pixels), given a more accurate measuring instrument, we believe that the capability of high‐precision manipulation would be demonstrated better. As for the demonstration process of active steering within the branched tunnel, the fluctuation shown in the measurement of tendon‐driven navigation mainly comes from the manual manipulation of pushing the robot forward. Both the instability of the cantilever (a long soft tube) and the vibration of the piezoelectric motor result in most of the dynamic tracking error (the linear tracking RMSE at about 6.3 μm without applying magnetic field, as shown in Figure S3f, Supporting Information) under magnetic actuation. As this study is the first step to demonstrate the steering and positioning capacities of the proposed soft continuum robot, a fully automated closed‐loop control system remains as the future work to improve the overall performance of the robot.