Complex Three-Dimensional Terrains Traversal of Insect-Scale Soft Robot

Abstract

This article proposes a piezoelectric-driven insect-scale soft robot with ring-like curved legs, enabling it to traverse complex three-dimensional (3D) terrain only by body-terrain mechanical action. Relying on the repeated deformation of the main body's n and u shapes, the robot's leg-ground mechanical action produces an “elastic gait” to move. Regarding the detailed design, first, a theoretical curve of the front leg with a fixed angle of attack of 75° is designed by finite element simulation and comparative experiments. It ensures no increase in drag and no decrease in the lift when climbing steps. Second, a ring-like leg structure with 100% closed degree is proposed to ensure a smooth pass through small-sized uneven terrain without getting stuck. Then, the design of the overall asymmetrical structure of the robot can improve the conversion ratio of vibration to forward force. The shape of curved legs is controlled by pulling the flexible leg structure with two metal wires working as spokes. The semirigid leg structure made of fully flexible materials has shape stability and structural robustness. Compared with the plane-legged robot, the curved-legged robot can smoothly traverse different rugged 3D terrains and cross the terrain covering obstacles 0.36 times body height (BH) at a speed of >4 body lengths per second. Moreover, the curved-legged robot shows 100% and 64% chances of climbing steps with 1.2- and 1.9-times BH, respectively.

Introduction

In nature, insects live in complex terrain and need to cross uneven land, such as grass, rocks, and roots of their body size.¹ We expect microrobots to be able to leap over obstacles such as insects and complete tasks in the real world on rugged terrain.^2–5 People have developed small robots that can navigate obstacles in different methods.^6,7 Jumping robots can climb over large obstacles or jump out of traps, such as magnetoelectric soft robots,⁸ continuous jumping robot with explosive and pneumatic legs,⁹ electrostatic-controlled jumping robot,¹⁰ origami robot that jumps by turning somersaults,¹¹ and soft robots powered by artificial muscles.¹² Guided by sensors or external signals, the gait of a multilegged robot can be precisely controlled to span gaps, climb stairs, or crawl on ceilings, such as the pneumatic quadruped robot of soft lithography,¹³ the parameterized tripped gait robot HAMR-E,¹⁴ and the bionic stick insect robot.¹⁵

Robots resembling caterpillars or inchworms can climb hills, drill crevices, cross rough terrain, or overcome large obstacles by peristalsis, such as the twin anchor robot,^16,17 the light-driven soft robot,^18,19 and the earthworm-like robot with flexible mesh materials.²⁰ The robot with large body deformation can squeeze to drill, crawl, or climb over obstacles, such as the light-driven arc robot,¹⁹ the electromagnetic driven robot,²¹ and the multifoot magnetically controlled soft robot.²² Rolling or rotating can also help robots climb hills or leap over obstacles, such as the pneumatic soft robots²³ and the soft magnetic origami robots.²⁴ These microrobots can navigate complex three-dimensional (3D) terrain, but cannot cross obstacles at high speed because they do not take advantage of the passive mechanical effects of structures.

In the high-speed escape behavior of evading predators, the response time of insects is insufficient and even reaches the limit of neural control. There is little time to adjust the body posture, and the probability of collision with obstacles is very high due to desperate running. Insects such as cockroaches^1,25 have evolved a reasonably compliant body structure to adapt to the terrain, using only mechanical mediation (passive dynamics involving inertia^26,27) to lift the body and then dynamically climb over obstacles. Jayaram et al²⁵ improved the body structure of DASH,²⁸ a small bionic cockroach robot. Colliding head-on with the wall will generate elevation in the pitch direction and transit from the floor to the vertical wall in 75 ms. Gart and Li designed a hexapod robot that could traverse large steps²⁹ and gaps³⁰ through “body-obstacle” mechanical action.

Zhang et al^31,32 proposed the bionic sand lizard robot, which can run fast on granular media such as sand or glass balls by leg–ground interaction. Li et al^33,34 built a streamlined cockroach robot rolling to adapt to the terrain and quickly crossing gaps narrower than its shell. Although the microrobots mentioned above can climb over obstacles in running by the mechanical action of the body and terrain, they are all composed of rigid structures.

The rigid structures used by the micro-obstacle climbing robots often fail under impact or compression, limiting their use in extreme environments such as postearthquake debris search and rescue or hazardous area exploration.¹⁰ However, the yielding soft robot body is simple to manufacture and light in weight, which has additional degrees of freedom¹⁸ and can adapt to large strains/stresses, showing better flexibility, robustness, and inherent terrain adaptability. The polyvinylidene fluoride (PVDF)-driven insect-scale soft robot proposed by Wu et al^35,36 can achieve a motion speed of 20 body lengths per second (BL/s). However, the flat-shaped legs are easily stuck in the gullies and can only move on flat ground.

Liang et al³⁷ added electrostatic footpads on that robot's legs and developed a high agility robot with trajectory control. In terms of the performance of crossing obstacles, this robot can only climb 2-mm steps. Therefore, the leg shape of the microsoft robot has an important influence on its all-terrain traversing ability.

This article proposes an insect-scale soft robot with ring-like curved legs based on PVDF-driven mode (Fig. 1). The robot has four ring-like curved legs and a plane-shaped tail attached to the body. Two silver wires control the shape of each curved leg, creating a semirigid leg structure using a fully flexible material. Thanks to the curved leg design, the robot can use the mechanical action of the body-obstacle to achieve dynamic crossing of complex terrain. The robot can maintain a relative speed of >4 BL/s on terrain covered by small obstacles with 0.36 times body height (BH). It also performs a 100% probability of passing through steps with 1.2 BH and a 64% probability for steps with 1.9 BH. In addition, the robot can also cross uneven terrains such as ladders, stairs, and plum flower piles.

FIG. 1.

Optical images and main structure of the curved-leg soft robot. PVDF, polyvinylidene fluoride.

Materials and Methods

Overall structure and driving principle

The robot has four ring-like curved legs and a planar tail. Two narrow front legs and two narrow hind legs can reduce the body's weight and increase its flexibility, reducing its impact on speed attenuation. In the initial state, both legs and tail are in contact with the ground. The detailed working principle of the piezoelectric soft robot for continuous motions could be found in our previous work³⁶ and also be explained by Supplementary Section S1, Supplementary Figures S1 and S2. When the robot moves on a flat surface at resonance frequencies, the robot's front end and back end vibrate up and down around the body's center. This effect will cause the robot to produce an “elastic gait” (Supplementary Fig. S14), mimicking the movement of a cockroach.^36,37

The planar and curved leg robots with different foreleg shapes are fabricated using the paper-cutting technique (Supplementary Figs. S3 and S4). The fabrication of the soft robot is shown in Supplementary Section S2 in the Supplementary Materials. To test their passing performance on a diverse terrain, we build a motion tracking platform that is composed of the signal generator, amplifier, high-speed camera, lamp, computer, and the robot test surface (Supplementary Fig. S18). The output signal of the electrical drive system (Supplementary Fig. S5) is DC bias voltages from −250 to 250 V, and its frequency is the structure's resonance frequency (200 Hz for planar leg robot, 210 Hz for Design 1, 207 Hz for Design 2, and 205 Hz for Design 3).

The overall symmetry of the robot affects the conversion rate of vibration-forward force. Without sticking the tail, an asymmetrical body structure creates asymmetrical force,³⁸ and we want to design asymmetrical legs to provide forward direction force. If the leg's curve is self-symmetric, or the front and back legs are axisymmetric concerning the axis of the body, the robot will stay in place (Supplementary Fig. S15). Only when the curve of the robot's legs is asymmetrical, and the arrangement of the front and rear legs is asymmetrical can it move forward under the driving signal.

The motion state of the robot is divided into aerial posture and touching ground posture. The robot gains forward force only when it touches the ground. The touching state can be roughly divided into seven conditions: front legs touching, rear legs touching, tail touching, front legs-tail touching, front legs-rear legs touching, rear legs-tail touching, and all three touching (Supplementary Fig. S16). The function of the planar tail is similar to the hind leg of the biped robot mentioned in Wu's et al article,³⁶ which can effectively increase the aerial time and running speed. Therefore, we simplify the analysis by neglecting the tail and only considering the front and rear legs hitting the ground. The posture is divided into front legs touching, rear legs touching, and both touching.

According to the drive signal, there are two modes of contracted or expanded tendency under each posture. Meanwhile, the robot's body shape has two kinds: “n-shape” and “u-shape.” Therefore, there are 12 cases in total (Fig. 2).

FIG. 2.

Force analysis of robot (without tail) in the cross-sectional view. (a–c) n-shape state. (d–f) u-shape state.

The solid red line indicates the current position, and the dashed gray line indicates the position at the next moment. The robot has a tendency to slide relative to the ground, as well as a tendency to roll, and the legs of the robot are flexible curved surfaces that can be slightly deformed. Therefore, the ground reaction force on the robot includes static friction force and support force on the deformation surface. We assume that the robot contacts the ground with 12 postures from the stationary state, and the driving signal generates internal forces that contract or expand. The direction of the static friction force is opposite to the motion trend, and the direction of the support force on the deformation surface is along the normal direction of the deformation part. G, F_f, and F_b represent gravity, static friction force, and support force on the deformation surface.

The magnitude and direction of the resultant (blue dotted arrow) of the F_f and F_b will vary with the robot's posture and shape. Overall, the effect of both the hind and front legs is to move the robot forward. The most efficient configuration is n-shape, front-touching, and contract when the components of the horizontal forward direction and vertical take-off direction are the largest.

The ring-like legs also help the robot traverse terrain with a continuous distribution of small bumps. This shape is similar to the smooth and constant body of snakes. The smooth body shape of limbless athletes can produce a bridging effect for gaps³⁹ to avoid getting stuck in small concave and convex to overcome the complex 3D terrain with kinetic energy and contact force.^40–43 The parameter “closed degree c” is proposed to quantify the open or closed state of the shape of the part of the robot leg that is in contact with the ground. A large degree of closure c will make the robot smooth when it contacts the ground (Supplementary Section S3). $\begin{matrix} c = \frac{N u m b e r o f c l o s e d r i n g - l i k e p a r t s t o u c h i n g t h e g r o u n d}{T o t a l n u m b e r o f p a r t s t o u c h i n g t h e g r o u n d} \end{matrix}$ (1)

There are no sharp edges when the robot's ring-shaped legs contact the terrain, and the robot will not get stuck in bumps or gullies. We analyze the movement of planar leg robots and curved leg robots on small concave and convex ground, such as boards with rectangular, circular, and lozenge obstacles (Fig. 3). Ignoring the effect of friction, we only consider the supporting force. The blue arrow indicates the magnitude and direction of the supporting force.

FIG. 3.

Robot motion analysis in small concave and convex terrains in the cross-sectional view. (a–c) Motion analysis of planar leg robot. (d–f) Motion analysis of curved leg robot.

In the expanded state, both the front and rear legs tend to swing upward, similar to weightlessness, and so, the supporting force and friction are smaller than those in the contracted state. The planar-legged robot encounters resistance in the expanded state. In the contracted state, the legs will be trapped by the backward and forward forces of the gap so that they cannot jump out and advance. Curved legs can provide forward force when they touch small bumps at a single point, similar to flat ground. When two points contact, the direction of the supporting force is the normal direction of the curve.

In the contracted state, the component of the supporting force is forward. In the expanded state, the backward supporting force is tiny. When the tail of the curved-legged robot touches the bump, it slaps the ground to lift the robot, as on a flat surface. When the tail touches the gap, four legs will pull it out with little drag, and there is little impact on the robot's traversal performance.

Theoretical curve of foreleg with a constant attack angle

Many insects rely on their legs to alternately grasp the ground when crawling slowly, but when they run fast and encounter obstacles, they have little time to adjust their legs. The cockroach's body has evolved a reasonable structure to climb over an obstacle in two steps.²⁵ (1) The cockroach's body collides head-on with the obstacle, and its head is lifted by mechanical action. (2) Alternating leg climbing provides continuous forward and upward force until it climbs over the obstacle.

Robots also need to navigate the terrain with body-size steps. There are also two steps for the robot to climb the steps. (1) The front curve of the robot's front leg collides with the steps, and the front part of the body is lifted due to bouncing from collision. (2) The robot body continues to vibrate under the driving signal. The front leg slaps the steps while the tail touches and slaps the ground, providing a continuous forward and upward force, helping to lift the robot's rear body and eventually climb over the steps.

When the robot climbs steps, its front legs first contact with obstacles. There is no doubt that the shape of the front legs has a significant influence on the climbing effect. If the reaction force of the steps to the front legs can lift the robot's head and reduce the forward resistance, it is easier to increase the height of the steps that the robot can cross and improve the climbing speed. According to the analysis above, when moving on flat ground, the rear legs and tail contribute a lot to the forward movement of the robot, and so, the slight adjustment of the front leg curve has a limited influence on the movement speed. Therefore, we optimize the curve shape of the foreleg's front part to make the robot perform better when crossing body-size large obstacles.

The angle between the tangent of the front leg and the horizontal direction is one of the most significant factors affecting the terrain passing ability, which is called the attack angle α. We analyze the change of attack angle and forces when the robot climbs the step (Fig. 4). $α_{1}$ , $α_{2}$ , $α_{3}$ is the attack angle between the front leg and the horizontal direction, and F is the reaction force of the step on the leg. When the robot climbs, the pitch angle will change. If the front leg of the robot is planar, in the process of lifting, the attack angle between the front leg and the horizontal direction gradually increases. The foreleg force can be divided into vertical and horizontal forces, and so, the horizontal resistance increases and lifting force decreases.

FIG. 4.

(a) The direction change of the foreleg tangent and the reaction force during climbing in the cross section view. The attack angle of the planar leg robot increases, while that of the curved leg robot remains unchanged. (b) Coordinate system. (c) The equivalent problem of calculation of the curve equation.

In extreme cases, the attack angle approaches 90°, with all the force reduced to horizontal resistance and no lift, making it difficult to climb steps. To solve the problem of attack angle variation, we propose a gradient curve leg instead of a planar leg (plane-legged robot, attack angle becomes larger when climbing the steps). Curved leg can maintain a constant attack angle during climbing, providing a stable upward lift without increasing drag (Fig. 4a).

We define the coordinate system (Fig. 4b) and calculate the gradient curve equation of the foreleg with a constant attack angle in the climbing process. As the robot climbs the steps to the right, its body rotates counterclockwise as it lifts. The robot body is flexible and has large deformation freedom in vibration movement. To simplify the calculation, we set the center of rotation as the robot's center of mass. Then, the problem is equivalent to that when the robot body is fixed, and the coordinate origin is the rotation center, the step moves to the left, and the height drops gradually and rotates clockwise (Fig. 4c). At this time, the calculated curve is only the front part of the foreleg, not the whole theoretical curve of the ring-like leg.

During movement, the angle between the tangent of the foreleg curve and the wall is δ. Since the wall is perpendicular to the radial path, the angle between the curve tangent and the radial path is α = 90° − δ. The problem is equivalent to the attack angle at any curve point and is always α. Suppose that the included angle between the radial path (x, y) and the positive x-axis is A, so tan A = y/x. We let the positive angle between the tangent line and the x-axis be B, so tan B = $y'$ . Since the angle between the tangent line and the radial diameter is α, so B = A − α or B = A + α. Then y′ = tan B = tan(A ± α). The ordinary differential equation is as follows: $\begin{matrix} y' = \frac{tan A \pm tan α}{1 \mp tan A \cdot tan α} = \frac{\frac{y}{x} \pm tan α}{1 \mp \frac{y}{x} \cdot tan α} = \frac{y \pm x \cdot tan α}{x \mp y \cdot tan α} . \end{matrix}$ (2)

The implicit equation for the solution of this “first-order nonlinear ordinary differential equation” (C is a constant) is as follows: $\begin{matrix} \frac{1}{2} ln (\frac{y^{2}}{x^{2}} + 1) \mp \frac{1}{tan α} \cdot {tan}^{- 1} (\frac{y}{x}) + ln |x| = C . \end{matrix}$ (3)

The initial condition is that the half-length of the robot body is l, so the curve crossing point (l,0). The curve equation is as follows: $\begin{matrix} \frac{1}{2} ln (\frac{y^{2}}{x^{2}} + 1) \mp \frac{1}{tan α} \cdot {tan}^{- 1} (\frac{y}{x}) + ln |x| = ln l . \end{matrix}$ (4)

We set l = 15 and draw the theoretical curves with different attack angles (Fig. 5a). Polylactic acid materials are melted to make 3D-printed plane-legged and curved-legged robots with different angles (Fig. 5b). The silver wire is attached to the head of the robot and pulled horizontally at a constant slow speed with a tension meter so that the robot can collide and climb over steps of different heights. According to the two-force principle, when the soft robot is in equilibrium, the measured pulling force on the silver wire equals the horizontal resistance of the robot (Fig. 5c).

FIG. 5.

(a) Theoretical curves at different attack angles. (b) Schematic diagram of 3D printing of planar and curved leg robots. (c) Maximum horizontal resistance when the 3D printing plane-legged and curved-legged robots cross steps of different heights. (d) Sizes of three robots with different forelegs (mm). (e) Three foreleg curves and a theoretical curve. 3D, three dimensional.

Regardless of the shape of the leg, the maximum resistance to climbing increases with the height of the step. The gradient curve leg can significantly reduce climb resistance compared with the flat leg at the same attack angle α. In addition, the climbing difficulty increases with the increasing attack angle, as shown in Figure 5c. We finally choose the curve when α = 75° for the following three reasons: (1) There is a remarkable phenomenon that in the case of limited space for the placement of front legs, the smaller the angle α, the lower the height of the front legs, which is not conducive to climbing higher steps. (2) From Figure 5c, we can see that the curve of 75°attack angle can climb the highest step height, and the resistance is not much greater than the curve of 60° attack angle. (3) In the experiment that follows, when the angle of the front leg gradually becomes smaller, the robot runs backward on the flat ground. Therefore, considering the height of the steps the robot can climb, the speed of forwarding motion, and the resistance of the steps, the angle should not be too small, and the curve of 75° attack angle is selected as the optimal curve: $\begin{matrix} \frac{1}{2} ln (\frac{y^{2}}{x^{2}} + 1) \mp (2 - \sqrt{3}) \cdot {tan}^{- 1} (\frac{y}{x}) + ln |x| = ln 15 . \end{matrix}$ (5)

The theoretical curve above satisfies the optimal attack angle, but it needs further improvement into a ring-like foreleg. We design three kinds of ring-like legs with different shapes, but the same total length for comparison (Fig. 5d). One robot's front leg curve is the theoretical curve of α = 75°. We draw the three robots' dimensions and compare the three front leg curves with the theoretical curve (Fig. 5e), expecting Design 2 to have the best climbing ability.

Finite element simulation for better curve

The qualitative mechanical effects when three robots collide with steps are analyzed. We establish finite element models of robots in software (quasistatic mechanical analysis, a certain momentary force during the movement process is analyzed), and let them hit the steps at the same initial speed (Supplementary Fig. S17). The model's materials and load parameters are shown in Supplementary Table S4, which are not precisely the same as the actual parameter.

Picking one of the fixed points on the step always in contact with the front leg, we draw the reaction force curve on the front legs exerted by steps during climbing (Fig. 6a). The reaction forces include x-direction resistance and the y-direction driving force. The area between the curve of the contact force and the horizontal time axis represents the value of the force impulse. The y-direction force of Design 3 is always zero, indicating that the robot lifts due to body deformation and elastic potential energy increasing. This way to climb steps is very inefficient. However, Design 2 has a small x-direction resistance impulse and a large y-direction dynamic impulse and is most conducive to climbing steps.

FIG. 6.

(a) The y-direction and x-direction force ratio versus x-direction displacement. (b) The displacement of the robot's center of mass in x-direction over time. (c) The velocity of the robot's center of mass in x-direction over time.

It is reasonable to assume that the robot's center of mass is in the body's geometric center. Plot the displacement and velocity curves of the center of mass in the horizontal direction with time. The velocity attenuation of Design 2 in the x-direction (blue curve) is the smallest (Fig. 6c). The x-direction displacement of the robot's center of mass reaches 15 mm, indicating it crosses the step. The results showed that Design 2 (blue curve) had the shortest climbing time and the best performance (Fig. 6b). We also calculated the external force in the experiments considering the whole-body kinematics of the robot to confirm the feasibility of simulation (Supplementary Section S4).

Results and Discussion

First, we measure the speed on flat ground as a reference. Two parallel glass panels are set up to form a movement channel. The relative velocities of the plane-legged and the curved-legged robot Design 1, Design 2, and Design 3 are 8.98, 7.19, 6.31, and 3.79 BL/s, respectively.

Terrain with small-size obstacle traversing

Taking three boards with small rectangular bumps as an example, test three robots' passing speed on small-scale concave and convex terrains (Fig. 7b). The experiment of each robot moving on each terrain was repeated 10 times and averaged. Experiments show that the planar leg robot could not pass the small rectangular obstacle of 0.5 mm. So we only draw the relative velocity (BL/s) of three curved-legged robots on the terrain of small rectangular blocks with the height of 0.5, 1, and 1.5 mm (Fig. 7c–f). Regardless of the foreleg shape, the bigger size of the continuous small obstacles leads to a lower crossing speed. Design 3 has the lowest relative rate among the three curved-legged robots (Supplementary Movie S1), at about 3 BL/s, which still exceeds the plane-legged robot.

FIG. 7.

(a) The robot moves on the boards with small rectangular, diamond-shaped, and round bumps taken by a high-speed camera. (b) Size of three boards with small rectangular bumps. (c–e) Relative velocity (BL/s) of three curved-leg robots passing through 0.5, 1, and 1.5 mm boards with small rectangular bumps. (f) Comparison of the relative speeds of the three robots. BL/s, body lengths per second.

Design 1 and Design 2 (Supplementary Movie S2) have roughly the same relative speed on three terrains, and both achieve good performance of above 4 BL/s. Therefore, the crossing speed of Design 2 in small uneven terrain with obstacles of 0.36 BH is >4 BL/s.

For the terrain with the small diamond-shaped and round bumps taken by a high-speed camera (Fig. 7a), the three curved-legged robots also performed better than the plane-legged robots (Supplementary Movie S3). Plane-legged robots will get stuck by small bumps, while the other three curved-legged robots will pass smoothly.

Terrain with body-size obstacle traversing

Taking climbing steps as an example, test the robot's passing performance on large concave and convex terrains.

The plane-legged robot in the control group could only cross 1 mm steps. The motion of the curved-legged robot was slowed 20 times by a high-speed camera, and we see the robot successfully climb steps in two stages: front end rises and back end rises (Supplementary Fig. S19 and Supplementary Movie S4). The experiment across each height step of 1–8 mm (Supplementary Movie S5) is repeated 50 times to obtain the climbing time and passing probability.

We set the time from the robot body touching the step to the tail passing the edge of the step as the climbing time. If the robot tilts over or climbs for a long time (according to previous work,²⁹ for 5–8 mm steps, this time is set to 8 s, while for 1–4 mm steps, this time is set to 6 s), it is judged as a failed attempt. The climbing time, passing probability of the three robots (Fig. 8a–c), and the comparison result (Fig. 8d) are drawn. In the experimental results, the standard deviation of climbing time is large because the robot posture can be divided into three situations: head-up collision, head-on collision, and head-down collision (Supplementary Section S5). Design 2 has the highest probability of the optimal head-up collision.

FIG. 8.

Transit time and passing probability of robots when climbing steps. (a–c) Specific transit time, average time, and passing possibility for robots, Design 1, Design 2, and Design 3, to climb steps 1–8 mm in height. (d) Comparison diagram of the transit time and passing probability of three robots.

In Figure 8, the success probability of Design 1 and Design 3 drops sharply, while Design 2 has a relatively smooth transition. It shows that when facing the lower steps, the robot can quickly cross over by relying on the aerial attitude or head-up collision. However, when faced with a high step and cannot be directly crossed, the front legs of the robot will continuously collide with the steps to adjust the pitch angle. Design 2's front legs result in stable lift and little forward drag, which makes it easier for the robot to turn into a head-up collision and climb higher steps. It just shows that the design of the front legs with a fixed angle of attack has stable performance for climbing steps.

For steps of the same height, Design 2 has the shortest climbing time among the three robots (Supplementary Movie S6). The possibility of crossing steps with 1.2 and 1.9 BH is 100% and 64%, respectively. The probability of Design 2 crossing steps of 1.2 BH is 2.78 times and 2.94 times higher than Design 1 and Design 3. The maximum steps height it can reach over is eight times higher than the planar-legged robot.

Application demonstrations

The previous experiments verify that Design 2 is the most efficient design. In the demonstration of traversing the complex terrain (Fig. 9 and Supplementary Movie S7), this robot passes through stairs, ladders, slopes covered with gauze paper, boards with diamond-shaped and rounded bumps, and plum flower piles (refers to the top of a series of piles, such as a scattered plum blossom). In only 10 s, the robot traverses the complex 3D terrain 17.3 times its body length without getting stuck in the bumps or gaps. So, the ring-like curved leg robot Design 2 has good all-terrain traversal capability. Statistical tests to show significant differences between the three designs are supplied in Supplementary Section S6.

FIG. 9.

Robot Design 2 traverses complex terrains.

Considering a grand challenge in soft robots is how to achieve untethered control within limited volume and weight, we also add a high-efficiency untethered power autonomous system in the robot. The weight ratio of robot to the battery system is 8.69. Powered by a battery (3.7 V, 40 mAh; HuiXinLi Inc.), the robot can run continuously for 34 min. More details are available in Supplementary Section S7.

After lowering the center of mass, robot Design 2 with a height of 4.152 mm and a body width of 15 mm can shuttle in a gap (Supplementary Section S8) that is narrower than half the body width (0.33 times the body width).

Conclusions

This study improved the tiny soft robot's legs into a ring-like surface and used two silver wires such as spokes to control the leg's shape to enhance its all-terrain passing ability. The shape of the robot's curved legs ensures a smooth body and a fixed attack angle against the steps. The closed shape of the ring legs ensures the robot to cross small uneven terrain. The asymmetric structure can convert the vibration into forward power. This semirigid leg structure has both the robustness of soft material and the shape stability of the rigid frame. The robot uses “body–terrain” mechanical interaction to overcome obstacles in fast running and realizes dynamic crossing on the uneven terrain with small bumps or large steps by passive dynamics. The crossing speed of small concave and convex terrains with obstacles of 0.36 BH is >4 BL/s.

The possibility of crossing steps with 1.2 and 1.9 BH is 100% and 64%, respectively. The movement is only conducive to the robot's structural and mechanical effects and does not involve neural regulation so that sensory regulation and neural processing can be put in more valuable places. To enable locomotion on more complex terrain (such as sand or grass), future studies need to increase the rolling friction force and breakaway ability of the robot. Possible future work also includes further improving the terrain passability and achieving closed-loop control of the robot by attitude control from the feedback of added sensors.

Footnotes

Author Disclosure Statement

No competing financial interests exist.

Funding Information

This work is supported by the Natural Science Foundation of Guangdong Province (2020A1515010647) and the Shenzhen Fundamental Research Funds (WDZC20200817152115001).

Supplementary Material

References

Grimaldi

, Engel

. Evolution of the Insects. J Insect Conserv, 2007; 11:211–212; doi: 10.1007/s10841-006-9003-3

Bruzzone

, Quaglia

. Review article: Locomotion systems for ground mobile robots in unstructured environments. Mechanical Sciences, 2012; 3(2):49–62; doi: 10.5194/ms-3-49-2012

Cully

, Clune

, Tarapore

, et al. Robots that can adapt like animals. Nature, 2015; 521(7553):503–507; doi: 10.1038/nature14422

Wallin

, Pikul

, Shepherd

. 3D printing of soft robotic systems. Nat Rev Mater, 2018; 3(6):84–100; doi: 10.1038/s41578-018-0002-2

Yang

, Bellingham

, Dupont

, et al. The grand challenges of Science Robotics. Sci Robot, 2018; 3(14):14; doi: 10.1126/scirobotics.aar7650

Aguilar

, Zhang

, Qian

, et al. A review on locomotion robophysics: The study of movement at the intersection of robotics, soft matter and dynamical systems. Rep Prog Phys, 2016; 79(11); doi: 10.1088/0034-4885/79/11/110001

St. Pierre R, Bergbreiter S. Toward autonomy in sub-gram terrestrial robots. Ann Rev Control Robot Auton Syst, 2019; 2(1):231–252; doi: 10.1146/annurev-control-053018-023814

, Lum

, Mastrangeli

, et al. Small-scale soft-bodied robot with multimodal locomotion. Nature, 2018; 554(7690):81–85; doi: 10.1038/nature25443

Bartlett

, Tolley

, Overvelde

JTB

, et al. A 3D-printed, functionally graded soft robot powered by combustion. Science, 2015; 349(6244):161–165; doi: 10.1126/science.aab0129

10.

Chen

, Liang

, Miao

, et al. Elevation control of a soft jumping robot. 2021 IEEE International Conference on Robotics and Automation (ICRA), 2021; 11843–11849; doi: 10.1109/ICRA48506.2021.9561676

11.

Zhakypov

, Mori

, Hosoda

, et al. Designing minimal and scalable insect-inspired multi-locomotion millirobots. Nature, 2019; 571(7765):381–386; doi: 10.1038/s41586-019-1388-8

12.

Zhao

, Zhang

, McCOul

, et al. Soft and fast hopping–running robot with speed of six times its body length per second. 2019;6(6):713–721. doi: 10.1089/soro.2018.0098.

13.

Shepherd

, Ilievski

, Choi

, et al. Multigait soft robot. Proc Natl Acad Sci U S A, 2011; 108(51):20400–20403; doi: 10.1073/pnas.1116564108

14.

de Rivaz

, Goldberg

, Doshi

, et al. Inverted and vertical climbing of a quadrupedal microrobot using electroadhesion. Sci Robot, 2018; 3(25); doi: ARTN eaau3038 10.1126/scirobotics.aau3038

15.

Blaesing

, Cruse

. Stick insect locomotion in a complex environment: Climbing over large gaps. J Exp Biol, 2004; 207(8):1273–1286; doi: 10.1242/jeb.00888

16.

Chang

, Chae

, Yoo

, et al. Loco-sheet: Morphing Inchworm Robot Across Rough-terrain. 2019 2nd IEEE International Conference on Soft Robotics (RoboSoft), 2019; 808–813; doi: 10.1109/ROBOSOFT.2019.8722724

17.

Wang

, Yang

, Tan

, et al. Bioinspired footed soft robot with unidirectional all-terrain mobility. Mater Today, 2020; 35:42–49; doi: 10.1016/j.mattod.2019.12.028

18.

Rogoz

, Zeng

, Xuan

, et al. Light-driven soft robot mimics caterpillar locomotion in natural scale. Adv Optical Mater, 2016; 4(11):1689–1694; doi: 10.1002/adom.201600503

19.

Ahn

, Liang

, Cai

. Bioinspired design of light-powered crawling, squeezing, and jumping untethered soft robot. Adv Mater Technol, 2019; 4(7):1900185; doi: 10.1002/admt.201900185

20.

Seok

, Onal

, Cho

, et al. Meshworm: A peristaltic soft robot with antagonistic nickel titanium coil actuators. IEEE ASME Trans Mechatron, 2013; 18(5):1485–1497; doi: 10.1109/tmech.2012.2204070

21.

Wang

, Mao

, Ge

, et al. Untethered and ultrafast soft-bodied robots. Commun Mater, 2020; 1(1):67; doi: 10.1038/s43246-020-00067-1

22.

, Zhang

, Yang

, et al. A bioinspired multilegged soft millirobot that functions in both dry and wet conditions. Nat Commun, 2018; 9:3944; doi: 10.1038/s41467-018-06491-9

23.

Shah

, Powers

, Tilton

, et al. A soft robot that adapts to environments through shape change. Nat Mach Intell, 2020; doi: 10.1038/s42256-020-00263-1

24.

Zhang

, Ke

, Jiang

, et al. Programmable and reprocessable multifunctional elastomeric sheets for soft origami robots. Sci Robot, 2021; 6:eabd6107; doi: 10.1126/scirobotics.abd6107

25.

Jayaram

, Mongeau

, Mohapatra

, et al. Transition by head-on collision: Mechanically mediated manoeuvres in cockroaches and small robots. J R Soc Interface, 2018; 15(139); doi: 10.1098/rsif.2017.0664

26.

Scarfogliero

, Stefanini

, Dario

, et al. Bioinspired jumping robot with elastic actuators and passive forelegs. In: 2006 1st Ieee Ras-Embs International Conference on Biomedical Robotics and Biomechatronics, Vol. 1–3. 2006; pp. 343–+.

27.

Schiebel

, Rieser

, Hubbard

, et al. Mechanical diffraction reveals the role of passive dynamics in a slithering snake. Proc Natl Acad Sci U S A, 2019; 116(11):4798–4803; doi: 10.1073/pnas.1808675116

28.

Birkmeyer

, Peterson

, Fearing

. DASH: A dynamic 16g hexapedal robot. 2009 IEEE/RSJ International Conference on Intelligent Robots and Systems, 2009; 2683–2689; doi: 10.1109/IROS.2009.5354561

29.

Gart

, Li

. Body-terrain interaction affects large bump traversal of insects and legged robots. Bioinspir Biomim, 2018; 13(2):026005; doi: ARTN 026005 10.1088/1748-3190/aaa2d0

30.

Gart

, Yan

, Othayoth

, et al. Dynamic traversal of large gaps by insects and legged robots reveals a template. Bioinspir Biomim, 2018; 13(2):026006; doi: ARTN 026006 10.1088/1748-3190/aaa2cd

31.

, Zhang

, Goldman

. A terradynamics of legged locomotion on granular media. Science, 2013; 339(6126):1408–1412; doi: 10.1126/science.1229163

32.

Zhang

, Qian

, Li

, et al. Ground fluidization promotes rapid running of a lightweight robot. Int J Robot Res, 2013; 32(7):859–869; doi: 10.1177/0278364913481690

33.

, Pullin

, Haldane

, et al. Terradynamically streamlined shapes in animals and robots enhance traversability through densely cluttered terrain. Bioinspir Biomim, 2015; 10(4):046003; doi: 10.1088/1748–3190/10/4/046003

34.

, Wang

, Li

. Omni-Roach: A legged robot capable of traversing multiple types of large obstacles and self-righting. 2022 International Conference on Robotics and Automation (ICRA), 2022; 235–242; doi: 10.1109/icra46639.2022.9811372

35.

, Ho

, Kariya

, et al. Pre-curved PVDF/PI unimorph structures for biomimic soft crawling actuators. In: 2018 IEEE Micro Electro Mechanical Systems. Belfast, Northern Ireland, 2018; pp. 581–584; doi: 10.1109/MEMSYS.2018.8346620

36.

, Yim

, Liang

, et al. Insect-scale fast moving and ultrarobust soft robot. Sci Robot, 2019; 4(32):9; doi: 10.1126/scirobotics.aax1594

37.

Liang

, Wu

YCA

, Yim

, et al. Electrostatic footpads enable agile insect-scale soft robots with trajectory control. Sci Robot, 2021; 6(55):11; doi: 10.1126/scirobotics.abe7906

38.

Liang

, Wu

, Shao

, et al. Manipulating the moving trajectory of insect-scale piezoelectric soft robots by frequency. In: 2019 IEEE 32nd International Conference on Micro Electro Mechanical Systems. Seoul, South Korea, 2019; pp. 1041–1044; doi: 10.1109/MEMSYS.2019.8870751

39.

Byrnes

, Jayne

. The effects of three-dimensional gap orientation on bridging performance and behavior of brown tree snakes (Boiga irregularis). J Exp Biol, 2012; 215(15):2611–2620; doi: 10.1242/jeb.064576

40.

Liljebäck

, Pettersen

,Stavdahl Ø, et al. Snake robot locomotion in environments with obstacles. IEEE/ASME Trans Mech, 2012; 17:1158–1169; doi: 10.1109/TMECH.2011.2159863

41.

, Li

. Robotic modelling of snake traversing large, smooth obstacles reveals stability benefits of body compliance. R Soc Open Sci, 2020; 7(2):14; doi: 10.1098/rsos.191192

42.

Ramesh

, Fu

, Li

CJA

. SenSnake: A snake robot with contact force sensing for studying locomotion in complex 3-D terrain. 2022 International Conference on Robotics and Automation (ICRA), 2022; 2068–2075; doi: 10.1109/icra46639.2022.9812159

43.

, Li

Snake robot traversing large obstacles using vertical bending with force feedback. ArXiv abs/2112.07815 2021.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

19.19 MB

4.61 MB

2.70 MB

5.58 MB

4.08 MB

4.69 MB

14.13 MB

3.24 MB

0.00 MB