Light‐Driven Amphibious Mini Soft Robot Mimicking the Locomotion Gait of Inchworm and Water Strider

Amphibious robots, which are expected to move agilely in both land and water environments with huge differences in medium density, have always been a hot spot in the field of robotics. However, limited by the simple structure and drive system, the existing mini‐robots have a relatively single‐movement mode, which limits their ability to move in complex environments. This article fuses two different biomimetic morphological features to create an amphibious mini‐robot: mimicking the body shape and gait of a terrestrial inchworm, and mimicking the superhydrophobic leg shape and gait of an aquatic water strider. The light‐driven approach also brings the advantage of remote wireless manipulation. The mass of the robot is only 8 mg, and it has fast land movement speed (≈0.05 times body length s−1), water surface movement speed (≈0.5 times body length s−1), and some obstacle‐crossing ability (over obstacles of ≈64% of the robot's height). Moreover, the robot can quickly switch from land to water locomotion, which is expected to facilitate emerging applications in various industrial and biomedical settings.


Introduction
Soft robots show good potential for applications in medical, military, and detection fields due to their strong impact resistance adaptability and biocompatibility.Currently, the trend of soft robots is intelligent, miniaturization, and multifunction, and wireless remote manipulation of robots has become the preferred choice.Light, as a clean energy source, not only enables wireless remote manipulation, but also effectively protects the environment, making it the most ideal driving method. [1,2]ompared with other robots, amphibious robots have more advantages in complex terrain and narrow spaces, and can work in various extreme conditions or areas that are difficult for humans to reach. [3,4]urrently, most amphibious robots are equipped with two sets of motion systems, namely, a land motion system and a water surface (or underwater) motion system.Such design of two motion systems increases the complexity of the robot's structure, thus limiting the miniaturization of the robot.Moreover, the robot needs to switch the motion system manually when moving across environments, which increases the cumbersome operation and reduces the moving efficiency. [5]everal researchers have proposed the concept of bionics to solve the problems posed by the two locomotion systems, i.e., to mimic the appearance and movement modes of amphibians in nature to achieve cross-environmental mobility.Most of the existing bionic amphibian robots are electrically driven because of their ability to provide powerful driving forces and diverse control modes. [6,7]owever, the electric drive requires circuitry designed inside the robot, which is not conducive to the miniaturization of the robot.Moreover, the electric drive method with a built-in battery has to consider the balance between battery life and battery volume.The light-driven approach is more conducive to the miniaturization and remote manipulation of the robot, and is less dependent on power supply.[14][15][16] The bionic design and multifunctional characteristics of light-driven amphibious soft robots are still to be studied urgently.
In the animal kingdom, some insects are able to employ relatively simple driving mechanisms and structures to achieve diverse locomotion patterns.In this article, inspired by the terrestrial crawling of inchworms [17][18][19][20] (Figure 1a top inset) and the rapid movement on water surface of water strider [21][22][23][24][25] (Figure 1a bottom inset), we designed a light-driven soft robot with carbon nanotubes (CNTs) and MXene (Ti 3 C 2 T x ) as the main light-absorbing materials.As shown in Figure 1b, the soft robot can not only crawl directionally on the ground like an inchworm, but also stand and glide rapidly on water surface like a water strider, and it also has a certain obstacle-crossing function.Besides, the robot can also realize the movement switch from the ground to the water surface without manual intervention, which simplifies the tedious operation of the amphibious robots and improves the efficiency.Finally, the four functions of land motion, water motion, obstacle crossing, and water-land motion switching have been integrated into a soft robot with a mass of only 8 mg through a bionic structural design.The experimental results are combined with simulation analysis to explore the structural design optimization method, performance influencing factors, and motion mechanism of each function.

Design and Fabrication of Amphibious Mini-Robot
The design is inspired by the movement patterns of inchworms and water striders.Among them, the inchworm can crawl through a series of actions: the front legs are fixed on the ground by suction cups, the body is flexed by muscle contraction, and then the hind legs move forward.Next, the hind leg suckers are immobilized, the body relaxes and returns to a flat shape, then the front legs move forward, and so on alternately.As a small aquatic insect commonly found in lakes, ponds, paddy fields, and wetlands, water striders can move quickly across the water surface or stay on the water surface without sinking.Water striders are able to stand on the surface of the water because of the special micro-nanostructure of their legs.Water striders have thousands of multilevel micro-nanosized bristles on their legs that are arranged in the same direction.Water striders use this special micro-nanostructure to efficiently adsorb air into the gaps of these oriented micron bristles and spiral nanogrooves, forming a stable air film on its surface and impeding the infiltration of water droplets.Thus, it macroscopically exhibits the superhydrophobic property of the water strider's legs. [23,26]igure 2a shows a schematic structural diagram of a soft robot, which consists of three parts: body, legs, and feet.The body of the soft robot is an end-to-tail asymmetric bilayer membrane structure.The passive layer is a polycarbonate (PC) film, and the response layer is a film obtained by pumping a mixture of MXene, polyvinyl alcohol (PVA), and nanocellulose.As a light-driven key layer in this soft robot, the bilayer film consists of a passive layer (hydrophobic layer-PVA layer) and a response layer (hydrophilic layer-MXene/PVA/nanocellulose composite layer) with strong interface bond strength.The light response principle of the bilayer is based on the different responsiveness of the connected passive layer and response layer to the photothermal conversion effect, as well as the directional bending phenomenon triggered by small asymmetric deformations of the bilayer.For the response layer, it is easy to absorb moisture in the air and maintains a high and saturated water content in the absence of near-infrared (NIR) light, while the passive layer does not absorb surrounding moisture and has a lower water content (with or without NIR irradiation).
In the absence of infrared light, the volume of both the response layer with high water content and the passive layer with low water content does not change, so the bilayer structure remains flat.Under infrared irradiation, the temperature of the response layer increases due to the photothermal conversion effect.And the increase in temperature leads to a decrease in its own water content, i.e., volume reduction, so that the responsive layer exhibits a contracted state.However, the passive layer is essentially unaffected by the photothermal effect, with minimal changes in moisture.At this time, neither the volume nor the shape of the passive layer changes.As a result, the entire bilayer structure bends to the side of the response layer due to the generation of asymmetric deformation.When the infrared irradiation disappears, the response layer is no longer affected by the photothermal effect and will continue to absorb moisture from the surrounding environment until the water content of the response layer returns to the original state and its volume returns to the initial state accordingly.At this point, the response layer and passive layer return to equilibrium, and the bilayer structure becomes flat.As a result, the bilayer structure will do periodic bend-return action with regular NIR irradiation switching.
Under light illumination, the bilayer film first bends, and the asymmetrical body structure brings asymmetrical force on the legs, which provides power for terrestrial travel.Therefore, the dimensions of L 1 , L 2 , L 3 , L 4 , and L 5 should be optimally designed.Among them, the length of the head L 4 and the length of the feet L 5 are important parameters that determine the ability to overcome obstacles on land.The detailed analysis is shown in the next section.The material of the legs is paper, which is obtained by cutting A4 paper.The feet are fabricated by using A4 paper as a substrate and then spraying a mixed solution consisting of CNTs, thermoplastic elastomer (TPE), and iron powder.The mixed solution on both sides of the foot is dried in a magnetic field environment to form a multilevel micro-nanoporous structure with a superhydrophobic effect similar to that of a water strider leg.Under light illumination, the photothermal conversion of CNT combined with the light trapping of the micro-nanostructure causes the temperature of the water near the feet to rise rapidly.A water surface tension gradient is generated between the irradiated area and the nonirradiated area, which provides an energy source for the robot's water movement.Details of the fabrication process of each part of the soft robot can be found in the Experimental Section.

Ground Directional Movement Mechanism and Impact Factors Analysis
First, the bilayer membrane material composition was optimized.For the response layer, it should be hydrophilic, easily absorbable, and sensitive to photothermal effect, while the passive layer should be hydrophobic, difficult to absorb, and insensitive to photothermal effect.The material selection for the photoresponsive layer was inspired by previous work, [27] and a composite of MXene, nanofiber, and PVA was used.The selection of MXene and nanocellulose as important constituent materials of the response layer stems from the fact that the hydrophilic functional groups (-OH and -O) on their surfaces will confer hydrophilic and hygroscopic properties to the response layer.The dynamic hydrogen bonds formed between adsorbed water molecules and MXene and nanocellulose readily bind or dissociate with small humidity fluctuations.Besides, MXene also has a good photothermal effect, which absorbs infrared light and converts it into heat deformation, providing the power to body bend of soft robot.But a single MXene membrane is not tough enough.The introduction of nanocellulose cellulose enhances the strength of the response layer and increases the water absorption of the response layer.Moreover, the presence of nanocellulose can also make the adjacent MXene film connect more firmly and change their shapes quickly under photothermal effect.And a PVA component is also added to the response layer, due to its good cohesiveness.The introduction of PVA not only helps to firmly bind the ternary heterogeneous materials in the response layer, but also effectively enhances the interfacial bonding strength between the response layer and the passive layer.This improves the failure problems such as interfacial peeling and detachment of the bilayer structure and effectively improves the stability and reliability of the bilayer structure.As shown in Figure S1, Supporting Information, the addition of PVA has almost no effect on the bending performance of the composite bilayer film, but it improves the stability of the bilayer film in repeated tests.The hydrophobic layer is made of PC because it is hydrophobic and presents inertness to water molecules and has a homogeneous porous structure (%220 nm).These properties contribute to the rapid transportation of water molecules and accelerate the light-driven response.
As shown in the inset of Figure 2b, the prepared bilayer is irradiated with NIR light and the bilayer bends toward the side of the photoresponsive layer.MXene has excellent photothermal conversion capability, while the PC film presents inert to light stimulation.When the NIR light is irradiated, the moisture in the photoresponsive layer desorbs rapidly at high temperature, resulting in a significant shrinkage of the photoresist layer.When the NIR light is turned off, the temperature returns to room temperature of 25 °C, and the photoresponsive layer absorbs moisture from the air to expand and recover.In contrast, the PC film expands when heated with a thermal expansion coefficient of about 6.5 Â 10 À5 K À1 and shrinks when the temperature decreases.According to the bending principle of the bilayer film, we can control the magnitude of the film bending angle θ by changing the light intensity (Figure 2b).When the light intensity is less than 40 mW cm À2 , the angle θ decreases rapidly with the increase of light intensity; when the light intensity is greater than 40 mW cm À2 , the angle θ tends to stabilize gradually.Besides, changing the light intensity can not only control the magnitude of the angle θ, but also control the response speed of the bilayer film.As shown in Figure 2c, the light response speed experiments at light intensities of 56.39, 81.77, and 145 mW cm À2 were conducted, respectively.Figure 2c shows the data results corresponding to the three cycles of the NIR lamp on and off.Although increasing the light intensity enhances the rate of bilayer bending recovery, it will not continue to reduce the angle θ.For the follow-up experiments without special instructions, the light intensity was chosen to be 145 mW cm À2 .
As the contraction and recovery of the response layer in the bilayer structure are affected by infrared illumination and moisture absorption, respectively, the relative humidity (RH) in the surrounding environment will have a large impact on the recovery of the response layer.Therefore, the response and recovery behavior of the double-layer structure was tested in various humidity environments.Four kinds of saturated salt solution environmental chambers, including LiCl (RH 12%), K 2 CO 3 (RH 44%), NaCl (RH 76%), and K 2 SO 4 (RH 97%), were used as humidity control devices.Three cycles of response and recovery experiments were performed by subjecting the bilayer structure into four humidity environments under the control of a regular infrared illumination switch, and the results are shown in Figure 2e.It can be clearly seen that changes in humidity have a significant effect on the recovery of the response layer.Nevertheless, changes in ambient humidity had little effect on the bending angle.The recovery of the response layer in the bilayer structure in high humidity environments (RH 76% and RH 97%) is significantly faster than in low humidity environments (RH 12% and RH 44%).The response and recovery of the bilayer in a single cycle tend to slow down with decreasing humidity: RH 97% %9 s>RH 76% %10.5 s>RH 44% %16 s>RH 12% %18.5 s.As a result, in high humidity environments, the responsive layer can absorb sufficient moisture from the surrounding air faster and easier, realizing a quick transition from a contracted state to a flat state.
The results of the soft robot's service life experiments are shown in Figure 2f.The soft robot was tested for over 1000 bending-recovery motions under periodic infrared light irradiation switches.The bending angle and kinematic performance of the soft robot were very stable during this long-cycle life test, and no degradation was observed.As shown in the inserted images, after 1000 tests, the device remained intact and no separation of the body and legs occurred.The above results prove that our soft robot has stable performance and good service life.
In order to observe the crawling movement of the soft robot on the ground, the robot was placed under an NIR lamp, and the switch was continuously opened and closed to make the robot's body continuously bend and recover.As shown in Figure 2d, the soft robot moves alternately with front and rear feet like an inchworm.In order to better optimize the land motion performance of the soft robot, the force analysis of its crawling process was performed, as shown in Figure 3a.Before bending, the soft robot is mainly subjected to the gravitational force G of the earth and the supporting force N 1 and N 2 of the two feet on the ground, that is, At this point, a force balance is formed.And the soft robot still maintains moment balance at the center of gravity, that is, where r 1 and r 2 are the horizontal distances from the front and hind legs to the center of gravity, respectively, and N 1 and N 2 are the supporting forces of the front and hind legs, respectively.Due to the asymmetric structure of the head and tail of the soft robot, r 2 is greater than r 1 , so N 1 is greater than N 2 .The differential leg supporting force causes the maximum static friction of the front leg to be greater than that of the rear leg.When the NIR light is on, the passive layer expands and the responsive layer contracts.In the horizontal direction, the front leg receives a horizontal force backward, and the rear leg receives a horizontal force forward.These two forces are equal in magnitude and opposite in direction, and we call them the contraction force F. When F gradually increases to be greater than the maximum static friction of the rear leg, F is still less than the maximum static friction of the front leg.Therefore, the front leg remains stationary and the rear leg moves forward.When the body of the soft robot bends to the maximum amplitude, the rear legs stop moving.Turning off the light, the passive layer starts to shrink and the responsive layer absorbs water molecules the air to gradually expand and recover.The forces generated by the recovery of the bilayer film act on the front and rear legs of the soft robot, respectively, as the recovery forces F 1 and F 2 .
The force situation of the soft robot is shown in Figure 3a, and the front and rear legs maintain a force balance in the vertical direction, that is, where α and β are the angles of the restoring forces F 1 and F 2 with respect to the vertical direction, respectively.Due to the asymmetric structure, the light-driven deformation of the robot's head is much larger than that of the tail, and the equivalent recovery force F 1 is presumed to be much larger than F 2 after the light is turned off.Therefore, the restoring force F 1 acting on the front leg in the horizontal direction is greater than its maximum static friction f 1 , and the front leg slides forward.The restoring force F 2 acting on the rear leg in the horizontal direction is less than its maximum static friction, and the rear leg remains stationary.We recorded the displacement of the rear leg of the amphibious soft robot with time using camera equipment (Figure S2, Supporting Information), and found that its motion trajectory was basically consistent with the theoretical analysis.
To sum up, the soft robot's directional crawling benefits from its head-to-tail asymmetrical structural design.When the NIR light is turned on and off, the two legs of the soft robot move accordingly, and we refer to the opening and closing of the light as one cycle.The soft robot moves its front and rear legs alternately once in a cycle.As the light keeps turning on and off, the soft robot keeps repeating the alternating movements, thus realizing continuous crawling on land (Video S1, Supporting Information) and transitioning from horizontal ground to downward crawling on slopes on its own (Video S2, Supporting Information).
To further verify the importance of the asymmetric structure, we conducted a control experiment, as shown in Figure 3b.The left picture in Figure 3b is the control group, which adopts a symmetrical structure design.The body of the robot is almost repeatedly bent on the spot without moving forward.The right picture in Figure 3b is the experimental group, which adopts a head-to-tail asymmetric structure design, and the robot can continue to move in a directional manner.The ability of soft robots to crawl on land is indeed due to the design of the asymmetrical structure.In order to improve the crawling speed of soft robots on land, experiments were conducted to investigate the influence of asymmetry (i.e., the value of L 1 /L 2 ) on crawling speed.The crawling speed is described in terms of how many times its body length moves per second (BL s À1 ).Five groups of soft robots with different values of L 1 /L 2 were designed for performance testing, as shown in Figure 3c.When L 1 /L 2 = 1/1, the motion speed of the soft robot is 0, that is, the head-to-tail symmetrical structure cannot make the soft robot crawl on land.When L 1 /L 2 = 1/10, the motion speed of the soft robot is also 0. The reason for this is that the center of gravity is too close to the front leg and too far from the rear leg.According to the experimental results combined with the force analysis in Figure 3a, it can be seen that the maximum static friction force of the front leg is always greater than the contraction force and recovery force in one cycle, and the front leg remains motionless.However, the maximum static friction force of the rear leg is always smaller than the contraction force and recovery force, and the rear leg moves back and forth.Thus, the soft robot as a whole behaves as stationary in place.When L 1 /L 2 = 1/5, the movement speed of the soft robot is the best, which is about 0.05 BL s À1 .

Water Movement Mechanism and Impact Factors Analysis
The ability of a water strider to stand smoothly on water and move quickly without getting wet is due to the superhydrophobic properties exhibited by the micro-and nanostructures on the surface of its feet.Therefore, in order to allow the soft robot to have the ability to glide on water, the two sides of its feet are treated with a superhydrophobic coating.A mixture of CNT, TPE, and iron powder was used as the superhydrophobic coating due to the demand for light-driven capability and photothermal conversion capability.CNT, TPE, and iron powder were uniformly dispersed in cyclohexane and sprayed onto both sides of the A4 paper substrate, and a magnetic field was applied during the drying process to form a superhydrophobic coating with rich micro-nanopore structure.The specific experimental operation is described in the Experimental Section.
Figure 4a displays the hydrophobicity comparison between the feet after superhydrophobic treatment and the untreated control sample.Apparently, A4 paper is a hydrophilic material, but after spraying the coating, it switches to hydrophobic.But whether the feet of soft robots can achieve superhydrophobicity, the contact angle must be measured, as shown in Figure 4b.It can be seen from the illustration in the figure that the contact angle of the hydrophobic layer is about 157°, which is greater than 150°, realizing superhydrophobicity.According to Figure 4c, the surface of the feet of the soft robot forms a multilevel micro-nanostructure similar to that of a water strider.In order to investigate the effect of CNT, TPE, and iron powder on the contact angle, we sprayed different material combinations onto A4 paper and measured their contact angles, as shown in Figure 4d.The contact angle of ordinary A4 paper is 99°, which is hydrophilic.The contact angles of coatings sprayed with CNT and CNT þ TPE þ Fe are 152°and 157°, respectively, while the contact angle of coating sprayed with CNT þ TPE is 142°.It shows that CNT and Fe have a positive effect on the hydrophobicity of the coating, while TPE has a negative effect.Considering that TPE improves the reversible elastic deformation and stability of the coating, the CNT þ TPE þ Fe composites were standardized in the subsequent experiments.
To verify that the superhydrophobic coating can improve the velocity of the soft robot moving on the water surface, a control experiment was carried out as shown in Figure 4e.The soft robot without hydrophobic treatment was used as the control group, and the soft robot with superhydrophobic treatment was used as the experimental group.It is evident from Figure 4e that the superhydrophobic coating does enhance the movement speed by nearly 1 time, about 10 mm s À1 (Video S3, Supporting Information), which corresponds to 0.5 times the BL s À1 , faster than the reported speed of a similar light-driven surface moving robot (about 0.25 BL s À1 ). [43]e drew a simple maze at the bottom of a 120 Â 120 mm petri dish to simulate the realistic and complex aquatic environment.As shown in Figure 4f and Video S4, Supporting Information, the soft robot was placed in the water-filled petri dish, and a laser was used to illuminate the various positions of the soft robot's feet to enable it to move forward and steer, and eventually exit the maze.[46][47] The surface of water has surface tension, the magnitude of which decreases with increasing temperature.When the laser spot acts on the middle of the rear edge of the soft robot's rear foot, the laser is absorbed by the CNT and converted into heat transfer to the water surface forming an uneven temperature gradient, as shown in Figure 4g.The figure shows the temperature distribution of the soft robot's foot under laser irradiation obtained by using COMSOL multiphysics field simulation, and the temperature changes at four points at the same distance from the center of gravity of the foot are measured, as shown in Figure 4h.Combining Figure 4g,h, the temperature changes at the left and right points are exactly the same, and the temperature gradient in the vicinity is completely symmetrical, i.e., the surface tension of water forms a force balance in this direction.However, the temperature difference between the front and rear points of the foot has been formed since the laser irradiation.The temperature of the front point is less than the temperature of the rear point, that is, the surface tension of the water in front of the foot is greater than the surface tension of the water behind the foot, so the tension gradient will push the whole soft robot to move forward.Similarly, when the laser spot acts on the lower left or lower right corner of the foot, the soft robot will turn to the right or left.By constantly adjusting the position of the laser spot acting on the soft robot's foot, it can glide quickly on the water surface arbitrarily.

Soft Robot Crosses Obstacles and Switches from Ground to Water Motion
Overcoming obstacles is an indispensable capability for soft robots to cope with complex real-world environments.Inspired by pole-vaulting (Figure 5a), a small section of PC film protrudes from the body of the soft robot at both the front and rear leg positions as a "spreader bar" with a length of L 4 , as shown in Figure 2a.After the soft robot's body is bent to a certain angle under the illumination of the NIR light, the spreader bar will prop up the two legs of the soft robot off the ground, thus crossing the tiny obstacles.Figure 5b shows the obstacle-crossing schematic of the soft robot, the upper figure shows the front leg crossing the obstacle, and the lower figure shows the rear leg crossing the obstacle.Combining the mechanism of obstacle crossing, it is speculated that the length L 4 of the spreader bar and the robot's leg length h (Figure 2a) are the keys to the soft robot's obstacle crossing.
In order to investigate the effect of L 4 on the obstacle-crossing ability of the soft robot, we made five groups of soft robots with different L 4 to test the obstacle-crossing ability.As shown in Figure 5c, either too long or too short L 4 is not conducive to obstacle crossing.The soft robot does not have a spreader bar when L 4 = 0 mm, losing the ability to cross the obstacle.When L 4 = 10 mm, the stiffness of the spreader bar (PC film) is not enough to support the body of the soft robot, and it cannot cross the obstacle.Only when L 4 = 5 mm, the spreader bar of the soft robot has a suitable length while its stiffness can support the body of the soft robot, so it has a good ability to cross the barrier.In order to clarify the obstacle-crossing conditions of the soft robot, we performed the following theoretical calculations: where H denotes the height of the obstacle that can be crossed, and L 4 and x denote the length of the spreader bar and the length of the robot's foot, respectively, as shown in Figure 5b.From this schematic diagram, we can know that the theoretical condition for the robot to cross the obstacle is that the height of the obstacle object H 0 cannot exceed H, that is, As we have determined the dimensions of the robot's foot to be a rectangle of 5 mm Â 7.5 mm, the value of x can be determined to be 5 mm.We measured the diameter of the obstacle (toothpick) by vernier caliper to be 1.6 mm, and then at this time H 0 = 1.6 mm.Then there is At this point, the value of L min is 4.1 mm, that is, This is one of the conditions that the robot needs to meet in order to be able to overcome obstacles in theory.Combining the experimental results in Figure 5c, we can know that only the soft robots of the two groups of L 4 = 5 and 10 mm meet the conditions.However, a value of L 4 that is too large will result in a spreader bar that is too long to support the weight of the robot.Therefore, the final size of L 4 = 5 mm is determined to be more suitable.Consequently, H = 2.5 mm is the maximum height over which the soft robot can cross the obstacle.After determining the appropriate length of the spreader bar L 4 , the soft robot easily stepped over the obstacle with a height of 1.6 mm (about 64% of the robot's height), as shown in Figure 5d and Video S5, Supporting Information.
After a theoretical analysis combined with the schematic diagram in Figure 5b, we found that the length h of the robot's legs also affects the robot's ability to cross the obstacle.Figure S3, Supporting Information, shows a partial schematic of the robot's front leg before crossing the obstacle.Combining the theoretical analysis and experimental validation of the soft robot crossing an obstacle, it is concluded that the length of the robot's legs should be higher than the height of the obstacle (Figure S3, Supporting Information).Thus, one of the conditions for the robot to cross an obstacle is obtained: where h is the length of the robot's leg and H 0 is the height of the obstacle.The height of the obstacle measured earlier is 1.6 mm, and then the length of the robot's leg should be at least greater than 1.6 mm.In order to cope with the realistic and complex water surface environment, the water surface obstacle-crossing function of soft robots is also necessary.Combined with the characteristics that NIR light can make the body of the soft robot bend and the laser can make the soft robot glide quickly on the water surface, the soft robot has a certain ability to overcome obstacles on the water surface under the combined action of the two.As shown in Figure 5e and Video S6, Supporting Information, the soft robot passed a wooden stick about 5 mm out of the water under the combined effect of laser and NIR light.In addition, the soft robot has demonstrated the ability to crawl from the ground to the water surface on its own.In this experiment, a slope of about 15°was set up for the soft robot to crawl down from the ground to the water surface (Figure 5f, Video S7, Supporting Information).Under the dynamic adaptive switching of the two working modes, the soft robot successfully moved from the land to the water surface, showing good application potential.

Conclusions
Inspired by the land locomotion mode of the inchworm and the superfast water gliding mode of the water strider, we have created a light-driven amphibious mini soft robot.It can move from the land environment to the water environment and also has the ability to overcome obstacles.Through the asymmetric structural design of the head and tail, the soft robot can convert the simple repeated bending motion of the body into directional crawling motion.In addition, the feet of the soft robot are hydrophobically treated with a multilevel micro-nanoporous structure, enabling it to stand firmly on the water surface and glide quickly under laser irradiation.The introduction of the unique "spreader bar" structure can make the soft robot have a certain ability to overcome obstacles on land.The combined use of infrared light and laser light can allow soft robots to pass through floating obstacles on the water.In addition, under the dynamic adaptive switching of the two motion modes, the soft robot also demonstrates the ability to crawl from the ground to the water surface by itself.This light-driven amphibious soft robot is much smaller in size compared to existing electrically driven amphibious soft robots, which allows it to work in some narrow workplaces.The advantages of remote manipulation, amphibious movement, and obstacle crossing make the mini soft robot show great application potential in the real environment.

Experimental Section
Fabrication of a Bionic Amphibious Soft Robot: As shown in Figure S4, Supporting Information, to prepare the bilayer film formed by MXCC-PVA film and PC film, we first mixed MXene powder, nanofibrillated cellulose (NFC), and PVA powder in deionized water in the ratio of 10:1:1.Specifically, 10 mg MXene powder, 40 mg NFC solution (nanocellulose, 2.5% concentration), 1 mg PVA powder, and 10 mL deionized water.The obtained mixed solution was sonicated for 1 h to form a suspension, and then the suspension was poured into a vacuum filtration apparatus and vacuum filtered with a PC filter membrane.A rectangular layer of 30 mm in length and 15 mm in width was formed on the PC filter membrane using a mask during the extraction process, and finally the bilayer membrane was obtained by natural drying in air for 2 h (see Figure S5, Supporting Information, for details).The obtained bilayers were cut into long strips, which underwent directional bending driven by NIR light, as shown in Figure S6, Supporting Information.
As shown in Figure S7, Supporting Information, to fabricate the feet of the soft robot, we first mixed 30 mg CNT in 100 mL cyclohexane and stirred it with an ultrasonic crusher for 1.5 h.Then, 30 mg TPE was added and stirred magnetically for 30 min, and finally, 30 mg iron powder was added into the ultrasonic crusher for 1.5 h to form a suspension.As shown in Figure S8, Supporting Information, the obtained suspension was injected into a spray gun and sprayed on the surface of A4 paper with the gun at a distance of about 10 cm from the substrate.A permanent magnet was used to continuously adsorb the magnetic particles on the surface while being placed in a drying oven to dry.The paper was then turned over to the other side of the A4 paper and the above operation was repeated so that both sides of the A4 paper were sprayed with CNTs, submerged in anhydrous ethanol for about 5 min, and left to dry in air.This resulted in a multilayer film structure with A4 paper as the substrate, covered with a CNT-TPE-Fe film layer on both upper and lower surfaces, possessing hydrophobicity (Figure S9, Supporting Information).The multilayer film was cut into 5 Â 7.5 mm rectangles for the legs of the soft robot, and then the A4 paper was cut into 2.5 Â 5 mm rectangles for the feet of the soft robot.
Finally, the body, legs, and feet prepared in front were assembled.The body and legs were attached with double-sided tape, while the legs and feet were attached with glue.A complete soft robot was finally obtained, as shown in Figure S10, Supporting Information.For subsequent optimization of various parameters of our amphibious soft robot, we repeated the above steps to prepare a variety of amphibious robots of different sizes as a backup, as shown in Figure S11, Supporting Information.
Characterization of a Bionic Amphibious Soft Robot: In order to better understand the microgeometry of each part of the soft robot, we tested and analyzed the surface and cross section of the robot's carcass and the surface and cross section of the robot's foot, respectively, using field emission scanning electron microscopy (SEM) (JSM 7800F Prime) with an accelerating voltage of 5 kV.Two small pieces of composite bilayer films were cut and put into the field emission SEM after the gold spray treatment to test the surface and cross section of the composite bilayer films, respectively.The test results are shown in Figure S12a, Supporting Information, where the MXene-NFC-PVA composite uniformly covers the upper surface of the PC filter membrane.We can see from the SEM image that the surface is not smooth and flat, and the magnification reveals that there are many lumpy bumps on the surface, whose composition is MXene powder.These powders, which are uniformly dispersed in the film, help to improve the absorption of infrared light in the film.To further understand the structure and bonding of the composite bilayer film, we need to observe the SEM image (Figure S12b, Supporting Information) of its cross section, and it can be seen that the uneven upper surface of the composite layer increases the surface area, which helps to improve the efficiency of absorbing infrared light.Two small pieces of paper matrix composite films covered with CNT composites were intercepted and put into the field emission SEM after gold spraying treatment to test the surface and cross section of the films, respectively.The test results are shown in Figure S13, Supporting Information.The surface has longitudinal and horizontal stripes and dense holes, and the magnification shows that there are tiny bumps and micropores on the surface.These bumps and micropores play a key role in the superhydrophobicity of the film.To further understand the microstructure of the film, we also need to observe the SEM image (Figure S14, Supporting Information) of its cross section, and we can see that the top and bottom surfaces of the A4 paper are covered with CNT composites, and both surfaces are bumpy, and these microstructures increase the surface area and thus improve the efficiency of light absorption.At the same time, these multilevel micro-nanostructures give the surface of the robot's foot good hydrophobicity, thus enhancing the motion performance of our amphibious soft robot on the water surface.
COMSOL Simulation of Robot's Foot Temperature Change under Laser Irradiation: The temperature change of the robot's foot surface under laser irradiation is simulated by the solid heat transfer module of COMSOL multiphysics field simulation software.The dimension of the robot's foot was set to be a rectangle of 7.5 mm Â 5 mm Â 0.1 mm.The laser power is set to 12 W and the laser spot radius is set to 0.5 mm.The ambient temperature is set to 25 °C.Set the relevant material properties of the robot's foot to the following parameters: constant pressure heat capacity of 690 J(kg K À1 ), density of 2100 kg m À3 , and thermal conductivity of 6600 W (m K) À1 .The material properties of the water were adopted from the data of liquid water in the COMSOL built-in material library.In order to give a clearer picture of the temperature gradient formed at the water surface, the thermal conductivity of the water was adjusted to 2000 W (m K) À1 .The 100 mm Â 100 mm Â 0.1 mm rectangle is set as a fluid and forms a union with the 7.5 mm Â 5 mm Â 0.1 mm rectangle.Four symmetrically positioned probe points were set around the robot's foot for detecting the temperature change at that point (Figure 4h).Then a finer mesh was automatically generated and finally the simulation results were calculated, as shown in Figure 4g.

Figure 1 .
Figure 1.Amphibious soft robot: a) schematic diagram of mini-robot's structure; b) functions of mini-robot.

Figure 2 .
Figure 2. Ground motion of the mini-robot: a) schematic diagram of the structure of the mini-robot; b) variation of the bending angle of the composite double-layer film with light intensity; c) variation of the bending angle of the composite double-layer film with time under different light intensities; d) optical photograph of the ground motion of the mini-robot; e) variation of the bending angle of the composite double-layer film with time under different humidity environments; and f ) the service life experiment of soft robot.

Figure 3 .
Figure 3. Ground motion mechanism and performance of the mini-robot: a) force analysis of the mini-robot; b) comparison of the performance of the mini-robot with symmetrical and asymmetrical head and tail; c) effect of different values of L 1 /L 2 on the ground motion performance of the mini-robot; and d) ground motion performance level of the mini-robot; the five-pointed star represents our work (for detailed data, see TableS1, Supporting Information).

Figure 4 .
Figure 4. Water motion of the mini-robot: a) hydrophobicity difference before and after A4 paper surface treatment; b) water contact angle of the paperbased composite film; c) SEM image of the paper-based composite film surface; d) water contact angle of the film surfaces with different material surfaces; e) comparison of the motion performance of the robot with superhydrophobic surface and the robot with hydrophilic surface on the water surface; f ) motion of the mini-robot in the water maze; g) temperature change simulation for robot's feet under laser irradiation; and h) temperature change around the robot's feet.

Figure 5 .
Figure 5. Mini-robot over obstacles and its autonomous switching from ground to water: a) pole-vaulting; b) schematic diagram of robot over obstacles; c) effect of different pole lengths on robot over obstacles; d) robot over obstacles on land; e) robot over obstacles on water; and f ) robot's motion switching from ground to water.