Power Autonomy and Agility Control of an Untethered Insect-Scale Soft Robot

Abstract

It is still challenging to achieve agility and trajectory control for untethered soft robots on an insect scale given their low mechanical impedance and compact structures. In this study, fast translational movements and swift turning motions are demonstrated on a power autonomous soft robot with a piezoelectric-thin-film-actuated body and electrostatic turning footpads. A high relative running speed of 2.5 body length per second compared with existing untethered robots is realized on a 24-mm-long untethered prototype integrated with power source, control, and wireless communication modules. An arc-shaped leg structure is adopted to self-regulate the frication forces on different footpads during turning by an inclination-induced redistribution of the payload gravity on legs and footpads. The trajectory maneuverability is demonstrated by navigating a 380 mg robot prototype with an 1810 mg payload to pass through a 58-cm-long S-shaped path with wireless control in 43.4 s. Due to the flexibility of the all-polymer body structure, the robustness of the untethered robot to large strain is demonstrated when compressed by 91 times the weight of the robot. A maximum travel distance of 58.6 m is achieved for the robot equipped with a 40 mA·h lithium battery, corresponding to the cost of transport of 261. This work provides a feasible solution to achieve high agility and advance the practicability of untethered soft robots on an insect scale.

Introduction

Power autonomy and trajectory control of the soft robots are of great value to practical applications. Relying on their advantages in small size, multiple degrees of freedom, high flexibility, and robustness, miniature soft robots driven by various actuators have been investigated extensively in recent years.^1–8 However, so far few soft robots in the insect scale can realize power autonomy due to the low driving force and power density of soft actuators.⁹ For small soft robots driven by the vibration of body or legs, a load of subgram can cause a significant change of the resonant frequency and therefore, deterioration of the robot mobility. It is still challenging to realize fast moving of untethered soft robots with power autonomy.^1,10,11

At present, most of the soft robots achieving untethered motion are mainly actuated by pneumatics,^12–14 shape memory alloys (SMA),^15–19 dielectric elastomer actuators (DEAs),^10,20–24 and ionic polymer metal composites (IPMCs).²⁵ Although soft robots driven by humidity^26,27 and magnetic force³ can also realize untethered movement, these robots require a specific external environment, which limits their practical applications. The untethered pneumatic robots need to be equipped with pumps and the large mass and volume of the pump lead to a relatively low running speed of the robot.^5,12 Must et al proposed a power autonomy robot actuated by IPMCs with a linear speed of 0.615 mm/s.²⁵ Huang et al developed an untethered robot driven by SMA, with various moving modes, including climbing, rolling, and jumping.¹⁹ The four-legged robot is 60 mm long with a moving speed of 3.2 cm/s (0.56 body length per second, BL/s). Ji et al used low-voltage stacked DEAs to propel an untethered soft robot to follow the printed paths.⁹ The moving speed of the robot is 12 mm/s and the trajectory cannot be controlled in real time.

To realize trajectory control, robots with strong steering capability are needed. Rigid-body Robots can realize turning motions by adjusting gaits.^28,29 While this could be challenging for soft robots with low structural stiffness.^9,25,30 Previously, we have proposed an insect-scale soft robot with relative turning centripetal acceleration of 28 body length per square second (BL/s²) enabled by front-foot electrostatic footpads.³⁰ A simplified analog control circuit is used to realize the movement along an S-shaped path. To achieve precise and multifunctional control, a microcontroller unit (MCU)-based embedded system is normally required.

Herein, we present an untethered soft robot actuated by an unimorph piezoelectric film structure with the integration of an MCU-based control system. Combined with an improved steering strategy and a suspension loading structure, the proposed soft robot allows for faster locomotion, better trajectory control, and payload capacity. Several key advancements are achieved in this article: (1) the highest relative running speed of 2.5 BL/s among reported untethered insect-scale soft robots; (2) remote trajectory control of the robot for the navigation through a 0.58-m-long path in a maze within 43.4 s; (3) a maximum travel distance of 58.6 m at an average speed of 1.88 BL/s (45 mm/s) is achieved with a 40 mA·h battery, corresponding to the lowest cost of transport (COT) of 261 among the published reports of untethered soft robots.

Design and Optimization

Robot structure design and optimization

A prototype 24 × 32 mm robot composed of a flexible piezoelectric unimorph body, a suspension framework for the assembly of the control circuit and battery, two arc-shaped legs, and two electrostatic footpads for steering is presented in Figure 1a and b. The piezo unimorph body consists of a 28 μm-thick polyvinylidene difluoride (PVDF) piezoelectric thin film with 6 μm silver (Ag) electrode films deposited on both sides and a 120 μm polyimide (PI) film attached on one side of the PVDF film as an inactive layer (Supplementary Fig. S1). The contraction and extension of the PVDF active layer under an alternating voltage lead to the vibration of the unimorph structure. Therefore, the legs on the unimorph body can strike the ground repeatedly and generate a forward driving force (Supplementary Section 1).⁸ A control system composed of power supply module, control module, and wireless communication module is fabricated on a double-sided flexible printed circuit board (FPC).

FIG. 1.

The structure and steering mechanism of the untethered soft robot. (a) The prototype of an untethered soft robot carrying power and control electronics. (b) The exploded illustration of the main structure of the untethered robot. (c–e) Illustration of the turning strategy of the untethered soft robot based on the combination effect of electrostatic footpads and arc-shaped legs.

The flexible control board, together with a lithium polymer battery and a radio frequency (RF) antenna are mounted on a suspended PI frame on the robot's back. The suspension structure leaves space for the free deformation of the robot body during locomotion and eliminates the influence of the payload on the robot's movement. The mass distribution on the suspension, especially the battery, MCU, and transformer chips, is optimized to adjust the center of gravity (COG) and ensure the balance of the robot.

Two footpads composed of a polyethylene terephthalate (PET) frame, an aluminum (Al) thin film electrode, and a PI substrate are attached on the two sides of the front leg. The detailed structure of the electrostatic footpads is described in our previous work.³⁰ A DC voltage as high as 500 V from the on-board high-voltage generating circuit is applied on the Al electrodes to generate electrostatic force between the footpads and ground.

When one of the footpads is attracted to the ground by electrostatic force, the body of a running robot will rotate around the fixed footpad due to the driving torque generated by the friction force on the footpad. Supposing the voltage is applied on the left pad, the driving friction force f_d is proportional to the electrostatic adsorption force F_e and the robot gravity force distributed on the left pad G_l as: $f_{d} = μ (G_{l} + F_{e})$ (1)

where μ is the friction coefficient. The electrostatic adsorption force F_e between the footpads and the ground can be calculated as: $F_{e} = \frac{1}{2} ε_{0} ε_{r} A E^{2}$ (2)

where ɛ₀ is the vacuum dielectric constant; ɛ_r the relative permittivity of the PI film under Al electrode; A the area of the footpad electrode and E the applied electrical field.

Meanwhile, the fraction of robot legs on the ground during rotation generates resistant torque M_r. Our initial tests show that the robots with rectangle-shaped legs are hard to make turns probably due to the large friction resistance. To solve this problem, an arc-shaped leg structure is adopted to reduce the friction force on the legs and besides, to redistribute the gravity forces on the footpads and supporting legs during turning. As shown in Figure 1c, the robot keeps level when moving straight ahead with no voltage applied on both footpads. As a DC voltage is added to the left pad, the electrostatic force generated on the pad leads to the increase of the friction between the footpad and the ground simultaneously. The friction force on the left pad limits the forward movement of the robot and leads to a turning motion together with the tilting of the robot body to the left by the assistance of the arc-shaped legs (Fig. 1d). The inclination of COG results in the suspension of right pad and the increase of G_l, which increases the driving friction force f_d and facilitates the turning motion. In this way, agile turning motion could also be realized on an untethered robot with a heavy payload.

A simplified 4-degree model, including translation, yaw, and roll rotation motions is established to further explain the turning mechanism. As shown in Figure 2a, the positions of the left and right footpads are simplified as point 1 and point 2, respectively. x and y represent the distance from the center of the footpad to the COG of the robot in the x-y coordinate system, respectively. With arc-shaped legs, the robot has both roll and yaw rotation when steering. Assuming the robot rotates counterclockwise with electrostatic adsorption force applied on left footpad, the right pad suspends from the ground due to the inclination of the robot body. To facilitate discussion, we translate the rotation center from the fixing footpad to the COG of the body. Also, force analysis is conducted by simplifying the left footpad and robot body into a spring, rod, and mass system (Fig. 2b). In this case, the steering torque generated by the driving friction force f_d is:

FIG. 2.

Optimization of the footpad location and COG height. (a) The simplified 4-degree model of the robot. (b) The simplified spring–rod–mass model for force analysis of a inclined robot. (c) Simulation results of the effects of footpad location and COG height on the net driving torque. COG, center of gravity.

M_{d} = f_{d} x_{1} = μ (F_{p} sin α + F_{e}) x_{1}

(3)

where x₁ is the horizontal distance from left pad to the inclined COG, F_P is the supporting force along with the left pad. Therefore, the net driving torque can be calculated as: $M = M_{d} - M_{r} = μ (F_{p} sin α + F_{e}) x_{1} - M_{r}$ (4)

The suspension loading structure is optimized by analyzing the influence of the COG height on the turning motion of the robot. According to Figure 2b: $\{\begin{matrix} F_{p} sin α + F_{b} sin θ = m g \\ F_{p} cos α + F_{I} = F_{b} cos θ \end{matrix}$ (5)

where F_b is the supporting force along the leg, m the mass of the robot. F_I the centrifugal force of the robot during turning, which can be described as: $F_{I} = m \frac{v^{2}}{x_{1}}$ (6)

where v is the linear velocity of the robot. According to Equation (5) and the triangle similarity theorem, F_P can be solved as: $F_{p} (x, h) = \frac{l_{p}}{x h} [m g (x - \sqrt{l_{p}^{2} - h^{2}}) - \frac{m v^{2} h}{\sqrt{l_{p}^{2} - h^{2}}}]$ (7)

where h is the height of COG, l_p the length of the left footpad after being squeezed, which can be described by: $l_{p} = l_{p 0} - \frac{F_{p} (x, h)}{k}$ (8)

where l_p0 is the initial length of the left footpad and k the spring constant. Combining Equations (4) and (7), the relationship between the net driving torque M with pad distance x and COG height h is obtained as: $M = μ x_{1} [m g (x - \sqrt{l_{p}^{2} - h^{2}}) - \frac{m v^{2} h}{\sqrt{l_{p}^{2} - h^{2}}} + x F_{e}] - M_{r}$ (9)

The simulation result for the net driving torque under various combinations of x and h values is obtained according to Equation (9) using MATLAB (Fig. 2c). It is found that the combination of large x and relatively small h values leads to high driving torque. Increasing x value gives a larger arm of the friction force f_d and results in a higher driving torque value. While a higher h value first increases the gravity force distributed on the left pad and therefore increases f_d when dynamic friction occurs. However, when further increasing h, f_d will keep constant after exceeding a threshold where the friction between the footpad and ground changes from dynamic friction to static friction. On the contrary, a higher h results in the decrease of the arm length x₁ of f_d due to the inclination of the robot body to the acting pad. Therefore, a relatively small COG height should be chosen to achieve a large driving torque.

Power and control circuit board design

To drive the piezoelectric robot body, an onboard high-voltage generator is fabricated on an FPC power board with three key components, including the low-pass filter and comparator, high-voltage converter, and output load. Pulse width modulation (PWM) waves with frequencies from 300 to 500 Hz are first generated by the onboard MCU (CC2640R2F RSM; Texas Instrument, Inc.). To avoid the influence of the high-order harmonics of the PWM signal on the high-voltage converter, the input PWM signal is attenuated by an integrator and a first-order low-pass filter with a cutoff frequency of 1.75 kHz. As the frequency required to drive the robot is lower than 1 kHz, most of the PWM signal will pass through the two-stage filter and be sent to the high-voltage converter chip (DRV2700; Texas Instrument, Inc.). The flyback mode of DRV2700 is chosen to realize a high-output voltage of around 500 V_pp. The output voltage $V_{H V}$ is programmed by the external voltage divider network shown in Figure 3a and can be calculated as:

FIG. 3.

Overview schematic and photo images of the circuit. (a) The smartphone sends commands to the control board over RF, the frequency control and switch signals are sent to the dual op-amp, high-voltage booster, and the load. The dual op-amp is used for filtering and comparator and two optocoupler relays are used for output voltage control. (b) Photo image of the control board, including low-power wireless MCU and ceramic patch antenna. (c) Photo image of the power board. (d) The wiring and overall layout of the electrical components on the robot. MCU, microcontroller unit; RF, radio frequency.

V_{H V} = V_{D I V} (1 + \frac{R_{8}}{R_{6} + R_{7}}) - V_{o p} \frac{R_{8}}{R_{6} + R_{7}}

(10)

where R₈ is 2050 KΩ, the sum of R₆ and R₇ is 5.502 kΩ, V_DIV is about 1.35 V, and V_op is typically approximated to 0 V. Therefore, the output voltage of the flyback converter $V_{H V}$ can be calculated as 504 V. It is worth noting that the selection of the transformer has a crucial influence on the performance of DRV2700. As the SW node of DRV2700 can boost up to 100 V, the minimum winding ratio of the transformer should be 1:5. Besides, a small inductance value on the primary side is needed to reduce the loss of the parasitic inductance. Therefore, a transformer (ATB322515-0110; TDK, Inc.) with a primary side inductance value of 7 μH and a winding ratio of 1:10 are selected. Finally, a square wave with 500 V amplitude and adjustable frequency is obtained from the power board. A pulldown network operated by a comparator is used at the output terminal to decrease the discharge time. The square wave signal is then divided into three channels to drive the robot body and control the adsorption of the left and right footpads, respectively. Optocoupler relays (APY216S; South Semiconductor, Inc.) controlled by the MCU are used to adjust the voltage applied on the footpads. The wiring between the power board and the robot is illustrated in Figure 3d.

The control and communication circuit are designed on a separated FPC board from the high-voltage power board (Fig. 3b, c). A 5-μm-thick PI film is attached between the control board and the power board for isolation. Under the premise of meeting low power consumption and compact footprint, the control board mainly needs to realize wireless communication and communicate with the power board through the general-purpose input/output (GPIO) port. As shown in Figure 3b, the control board mainly consists of a low-power MCU CC2640R2F RSM and a 2.4 GHz ceramic patch antenna (AN1608-2440; Rainsun, Inc.) with 0603 footprint. The RF-enabled MCU chip CC2640R2F RSM can communicate with the smartphone through the 2.4 GHz wireless network. In this way, the frequency of the driving voltage and the on/off states of the voltage applied on the footpads can be controlled wirelessly by a smartphone. The driving frequency can be adjusted with a precision of 1 Hz, which is essential for the robot's speed control.

A 40 mAH lithium polymer battery (3.7 V; HuiXinLi, Inc.) is used to power the two circuit boards. The total weight of the untethered soft robot with battery and circuits is 2190 mg, including a 380 mg robot body and a payload of 1810 mg. The detailed weight distribution of the robot is shown in Table 1.

Table 1.

Weight Distribution of the Untethered Robot

Component	Weight (mg)
Robot body	380
Battery	840
Control board	245 (MCU, 170 mg)
Power board	725
Total weight	2190

MCU, microcontroller unit.

Results and Discussion

Effect of actuation signal on the running speed of the untethered robot

The effect of driving frequency on the relative velocity of the robot is shown in Figure 4a. As the driving frequency changes from 200 to 1200 Hz, a peak velocity of 2.5 BL/s (60 mm/s) appears at 410 Hz under a voltage of 500 V_pp. The running speed drops dramatically as the driving frequency has a deviation from the peak frequency. When the driving frequency is close to the resonant frequency of the robot structure, the piezoelectric unimorph generates a maximum displacement and therefore, higher contact force between robot legs and ground. So the robot shows the highest running velocity at this frequency. The driving voltage shows a nearly linear relationship with the velocity. The relative velocity increases rapidly from 0.2 to 2.5 BL/s as the driving voltage increases from 220 to 500 V at the frequency of 410 Hz (Fig. 4b). Figure 4c and Supplementary Video S2 shows the velocity test of the untethered robot running at the highest speed. The robot moves through a 300-mm-long acrylic tube in 5.72 s with an average speed of 2.18 BL/s (52.4 mm/s). The highest speed of 2.5 BL/s (60.2 mm/s) is achieved in the 10–20 cm section.

FIG. 4.

Effects of the (a) driving frequency and (b) driving voltage on the relative linear velocity of the untethered robot. (c) Linear velocity test of the untethered soft robot running at the highest speed.

Effect of structural configuration on the mobility of the robot

The experimental results of the effects of footpad position and COG on the turnning ability of the robot are shown in Figure 5a–c. Relative centripetal acceleration, defined as turning centripetal acceleration per body length of the robot, is used to evaluate the turning agility of the untethered robot.³⁰ The footpad position x, that is, the horizontal distance of footpad to the center of robot body, shows a significant effect on the centripetal acceleration and steering radius (Fig. 5a, b). With the increase of x from 8 to 16 mm, the relative centripetal acceleration of the untethered robot increases with quadratic function approximately, which is in agreement with the simulation results shown in Figure 2. A linear relationship between x and steering radius was observed as the robot rotates around the fixed footpad during turning. As shown in the orange and blue regions in Figure 2a and b, a dead zone was found as x value changes from 8 to 13 mm in which the steering failure rate is too high to be acceptable. This is probably because the centripetal force is not high enough to ensure a smooth turning of the robot when x is small.

FIG. 5.

The relationship of (a) centripetal acceleration and (b) steering radius with the footpad position. (c) The effect of COG height on the centripetal acceleration as the footpad distance is 13 mm. (d) The effect of COG location in y direction on the linear velocity and resonant frequency of the robot. (e) The effect of payload on the relative velocity and resonant frequency of the robot when the COG position in y direction is zero. (f) Performance improvement of robot with arc-shaped legs in comparison with rectangular-legged robot. The payload in (a–d, f) is 1.81 g.

Finally, the x value of 13 mm was chosen to give enough centripetal force while keeping the steering radius low. In this case, the untethered soft robot with a payload (1810 mg) of more than 4.7 times its own body weight can realize a maximum centripetal acceleration of 0.07 BL/s² with a turning radius of 18 mm, corresponding to a turning speed of 18.0°/s.

By adjusting the PI suspension frame, the height of the circuit and battery is changed to investigate the effect of COG height on robot performance (Fig. 5c). With the increase of COG height from 2 to 6 mm (measured from the plane of PVDF layer), the relative centripetal acceleration decreases rapidly from 0.07 to 0.02 BL/s². A larger overturning torque is generated by the inertia force during turning at a larger COG height, which intends to separate the fixed footpad from the ground and therefore, decrease the driving friction force and centripetal acceleration.

The horizontal position of COG is adjusted by changing the position of the battery. As COG moves from the rear leg to the foreleg, the linear velocity of the robot increases slightly from 1.75 to 2.35 BL/s (Fig. 5d). When moving straight, the driving force of the robot mainly comes from the foreleg.⁸ As COG is close to the robot head, the friction force applied from the ground to the foreleg increases due to the increase of weight distribution on the foreleg. Thus, the moving speed is improved. However, shorting risk increases if the battery is mounted near the robot head where high-voltage modules on the circuit are located. Besides, the resonant frequency of the structure increases, and vibration of the battery occurs as the battery is moved to the middle of the frame, which limits the deformation of robot body and the additional load capability of the robot. Therefore, the COG is finally located in the rear region of the robot.

The linear moving speed of the robot decreases as the payload increases due to the attenuation of the vibration amplitude and resonant frequency (Fig. 5e). A tethered robot without carrying payload gives a relative linear velocity of 3.78 BL/s (91 mm/s), which is in agreement with our previous work.³⁰ After adding payloads with COG of zero in the y direction, the running speed decreases gradually from 3.78 to 1.75 BL/s under the payload of the full weight of circuit and battery with a payload to body weight ratio of 4.8. As the ratio of payload to body weight increases to 7.8, the speed drops to 0.85 BL/s. The effect of the arc-shaped leg on the centripetal acceleration is verified by comparing with the rectangular-legged robot. As shown in Figure 5f, the relative centripetal acceleration has been improved more than twice when the arc-shaped leg is adopted.

The turning capability of the untethered soft robot is also demonstrated by turning the robot 360° around a fixing point with a radius of 18 mm and a rotation speed of 18.0°/s and 15.6°/s in counterclockwise and clockwise directions, respectively (Fig. 6a, b and Supplementary Video S1). The speed difference is probably due to the slight structural asymmetry of the robot resulting from the fabrication tolerance. The untethered robot can be controlled by a smartphone wirelessly. An android application is developed on the smartphone to adjust the driving frequency and on/off state of the voltage applied on footpads manually. The communication distance of Bluetooth is about 1.5 m, and the response time of the footpads is <500 ms. The trajectory control of the robot is demonstrated on an S-shaped path with four 90° angles. It takes 43.4 s for the robot to pass through the 580-mm-long route (Fig. 6c). The trajectory of the robot is depicted with red dots and the density of the dots illustrates the speed of the robot as the time interval between adjacent dots is the same.

FIG. 6.

An untethered robot moves counterclockwise (a) and clockwise (b) with the angular velocity of 18°/s and 15.6°/s, respectively. (c) Trajectory control of the robot on an S-shaped path by remote operation of a smartphone.

Robustness and rough surface adaptability

The robustness of the untethered soft robot is verified by adding and removing a 200 g weight (91 times the weight of the untethered robot) on a running robot. As shown in Figure 7 and Supplementary Video S3, the weight generates a large deformation on the foreleg of the robot, while the robot recovers immediately after the weight is removed. Due to the flexibility of the all-polymer body, the mobilities of the robot, including moving and turning capabilities are not affected under large strain.

FIG. 7.

Robustness test of the untethered soft robot under the squeeze of a 200 g weight. The robot recovers rapidly and keeps mobility afterward to show good robustness to large strain. (a–f) The movement of the robot at different time during squeeze and release.

The mobility of the robot on rough surfaces, including unpainted wood, painted wood, linen, and sandpapers in 180 and 2000 grit is also tested. As shown in Supplementary Section 2, Supplementary Figure S2 and Supplementary Table S1, the robot can move forward on unpainted and painted wood surfaces at a velocity of 1.6 and 1.9 BL/s, respectively. When running on sandpapers, the thin robot legs will be stuck in the small abrasive grits and the robot cannot move forward easily. In most cases, the robot will turn around its rear leg on both 180 and 2000 grit sandpaper and the electrostatic footpads are not working due to the low electrostatic force on the rough surface. On a soft and rougher linen surface, the robot cannot move as its legs are trapped in the loose weave. To improve the adaptability of soft robots to complex terrains, another work on robots with curved legs has been conducted and submitted elsewhere.

Power consumption and COT

The continuous operating ability is evaluated by driving the robot running in an acrylic tube with a fully charged battery. The runtime and distance are recorded until the robot stops due to low power as the battery voltage drops from 4.1 to 3.2 V. According to the results of three tests, the average continuous runtime for robots equipped with a 40 mA·h lithium battery is 21.5 min, and the maximum travel distance is around 58 m. Therefore, an average speed of 45 mm/s is achieved.

The power consumption of the robot is calculated from the voltage and current of the battery on a running robot. As shown in Figure 8a, the charge/discharge curve of the battery is first plotted by a battery cycler (S4000; Maccor, Inc.) with a charge current of 40 mA. The discharge current of the battery on a running robot, that is, 65 mA, is measured by a digital multimeter (Keysight 34470A). By recording the battery voltage before and after robot running, the power consumption can be calculated from the integral of voltage and current according to the battery discharge curve. Then the COT of the robot can be calculated as:

FIG. 8.

(a) Charge and discharge current of the battery used in three cycles. (b) COT of terrestrial arthropods (triangle),^31,32 tethered rigid robots (light square),^33–39 untethered rigid robots (dark square),^40–44 tethered soft robots (diamond),⁸ untethered soft robots (circle),^9,25,30 and this work (star) versus body mass. COT, cost of transport.

C O T = \frac{P}{m g v} = \frac{253.3 m W}{2.2 \times 10^{- 3} k g \times 9.8 m ∕ s^{2} \times 0.045 m ∕ s} = 261

(11)

Figure 8b shows the comparison of COT with respect to the body mass for terrestrial arthropods (green),^31,32 tethered rigid robots (solid yellow),^33–39 untethered rigid robots (purple),^40–44 tethered soft robots (brown),⁸ untethered soft robots (blue),^9,25,30 and this work (red star). Thanks to the fast running speed and convenient gait control, some tethered rigid robots have demonstrated comparable COT performance with insects in the literature.^32,36 While the COT of untethered rigid robots increases due to the extra load from driving circuit and power source. COT control for soft robots could be even more challenging because of the low efficiency of soft actuators and difficulty in accurate gait control due to their low structural stiffness.^9,25,30 Specifically, the COT of a reported untethered soft robot with a payload of 780 mg is 1670 with a running speed of 12 mm/s.⁹ This work demonstrates the lowest COT of 261 among untethered soft robots due to the high linear moving speeds, light weight, and low power circuit design.

Conclusions

Agile locomotion and trajectory maneuverability are demonstrated on a 24-mm-long untethered soft robot integrated with power, control, and communication electronics. An arc-shaped leg structure is adopted to self-regulate the friction force of the two electrostatic footpads by inclination-induced redistribution of the large payload from the integrated electronics. The location of COG is optimized using a 4-degree model. The resulting relative running speed of 2.5 BL/s is higher than most untethered insect-scale soft robots reported. The robustness of the untethered soft robot to large strain is demonstrated by compressing with 91 times robot's weight. With a low-power consumption design, an allowable travel distance of 58.6 m is achieved for the robot equipped with a 40 mA·h lithium polymer battery. The resulting COT of 261 is the lowest among all untethered small soft robots. These achievements have advanced the practicability of insect-scale soft robots and proved their huge application potential.

Experimental Section

Fabrication of the untethered soft robot

A 28 μm PVDF film (PolyK Technologies, LLC) with 24 × 22 mm in size is first deposited with 6-μm-thick silver film on both sides by slot die coating method (Supplementary Fig. S3a). The robot body, supporting frames, and arc-shaped legs are fabricated from a 120 μm PI film with a 25 μm sticky silicone layer (Keyun Tape, Inc.) by a vinyl cutter (CAMESO 3; Silhouette, Inc.). The electrostatic footpad is composed of a 25 μm thick PET skeleton and 5-μm-thick PI film with 10 nm Cr/100 nm Al films as the electrode. Both the PET skeleton and the PI/Cr/Al electrode film is patterned by the vinyl cutter. The detailed fabrication method of the footpad is shown in our previous work.³⁰ Finally, the PVDF film, PI frame, legs, and footpads are carefully folded and attached to the PI body by the silicon adhesion layer, as shown in Supplementary Figure S3d. The installed legs give an angle of 70° with the ground. The circuit board and battery are mounted on the PI supporting frame and 50-μm-in-diameter silver wires are used to connect the power board with the robot body and footpads. The overall dimension of the fabricated robot is 24 × 32 × 13 mm.

Velocity test of the untethered robots

Due to the slight asymmetry of the robot body structure caused by the manufacturing tolerance, the robots cannot run straightly for a distance longer than 30 cm, which is the distance used for the linear velocity measurement. The electrostatic footpads are not activated during the straight-line velocity test to avoid the drag force. In this case, an acrylic tube with an inner diameter of 50 mm was used as a guide to constrain the running direction of the robot (Supplementary Fig. S4). The rectangular legs of the robot work well on the curved tube inner surface and the two rectangular corners in contact with the curved surface can generate enough friction force to make the robot move quickly. This is proved by comparing the linear velocity of the robot running in the tube and on the flat surface. On flat surface, the highest speed of the robot calculated from a segment on the robot's path is 2.6 BL/s. While the highest velocity inside the acrylic tube is 2.5 BL/s, very close to that on flat surface. Therefore, we choose acrylic tube as a guide for the measurement of the linear velocity of the robots in all cases.

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 Shenzhen Fundamental Research Funds (JCYJ20180508151910775).

Supplementary Materials

Supplementary Section 1

Supplementary Section 2

References

Rus

, Tolley

. Design, fabrication and control of soft robots. Nature, 2015; 521:467–475.

Arazoe

, Miyajima

, Akaike

, et al. An autonomous actuator driven by fluctuations in ambient humidity. Nat Mater, 2016; 15:1084–1089.

, Lum

, Mastrangeli

, et al. Small-scale soft-bodied robot with multimodal locomotion. Nature, 2018; 554:81–85.

Wehner

, Truby

, Fitzgerald

, et al. An integrated design and fabrication strategy for entirely soft, autonomous robots. Nature, 2016; 536:451–455.

Tolley

, Shepherd

, Mosadegh

, et al. A resilient, untethered soft robot. Soft Robot, 2014; 1:213–223.

Marchese

, Onal

, Rus

. Autonomous soft robotic fish capable of escape maneuvers using fluidic elastomer actuators. Soft Robot, 2014; 1:75–87.

Hines

, Petersen

, Lum

, et al. Soft actuators for small-scale robotics. Adv Mater, 2017; 29:1603483.

, Yim

, Liang

, et al. Insect-scale fast moving and ultrarobust soft robot. Sci Robot, 2019; 4:eaax1594.

, Liu

, Cacucciolo

, et al. An autonomous untethered fast soft robotic insect driven by low-voltage dielectric elastomer actuators. Sci Robot, 2019; 4:eaaz6451.

10.

Kim

, Laschi

, Trimmer

. Soft robotics: A bioinspired evolution in robotics. Trends Biotechnol, 2013; 31:287–294.

11.

Rich

, Wood

, Majidi

. Untethered soft robotics. Nat Electron, 2018; 1:102–112.

12.

Duggan

, Horowitz

, Ulug

, et al. Inchworm-Inspired Locomotion in Untethered Soft Robots. In: 2019 2nd IEEE International Conference on Soft Robotics, Seoul, 2019; pp. 14–18.

13.

Aydin

, Molnar

, Goldman

, et al. Design of a Soft Robophysical Earthworm Model. In: 2018 IEEE International Conference on Soft Robotics, Livorno, 2018; pp. 24–28.

14.

Zhang

, Wang

, et al. Design and modeling of a parallel-pipe-crawling pneumatic soft robot. IEEE Access, 2019; 7:134301–134317.

15.

Lin

, Leisk

, Trimme

. GoQBot: A caterpillar-inspired soft-bodied rolling robot. Bioinspir Biomim, 2011; 6:026007.

16.

Wang

, Lee

, Rodrigue

, et al. Locomotion of inchworm-inspired robot made of smart soft composite (SSC). Bioinspir Biomim, 2014; 9:046006.

17.

Umedachi

, Vikas

, Trimmer

. Softworms: The design and control of non-pneumatic, 3D-printed, deformable robots. Bioinspir Biomim, 2016; 11:025001.

18.

Koh

, Cho

. Omegabot: Biomimetic Inchworm Robot Using SMA Coil Actuator and Smart Composite Microstructures (SCM). In: 2009 IEEE International Conference on Robotics and Biomimetics. Guilin, Guangxi, 2009; pp. 1154–1159.

19.

Huang

, Kumar

, Jawed

, et al. Chasing biomimetic locomotion speeds: Creating untethered soft robots with shape memory alloy actuators. Sci Robot, 2018; 3:eaau7557.

20.

Cao

, Qin

, Liu

, et al. Untethered soft robot capable of stable locomotion using soft electrostatic actuators. Extreme Mech Lett, 2018; 21:9–16.

21.

Jung

, Koo

, Nam

, et al. Artificial annelid robot driven by soft actuators. Bioinspir Biomim, 2007; 2:S42–S49.

22.

, Zou

, Zhao

, et al. Soft wall-climbing robots. Sci Robot, 2018; 3:eaat2874.

23.

Tang

, Li

, Chen

. U-Turning an Agile Robotic Cube by a Soft Dielectric Elastomer Resonator. In: 2018 IEEE International Conference on Soft Robotics, Livorno, 2018; pp. 333–338.

24.

Chen

, Zhao

, Mao

, et al. Controlled flight of a microrobot powered by soft artificial muscles. Nature, 2019; 575:324–329.

25.

Must

, Kaasik

, Põldsalu

, et al. Ionic and capacitive artificial muscle for biomimetic soft robotics. Adv Eng Mater, 2015; 17:84–94.

26.

Lee

, Prosser

, Purohit

, et al. Bioinspired hygromorphic actuator exhibiting controlled locomotion. ACS Macro Lett, 2013; 2:960–965.

27.

Shin

, Ha

, Lee

, et al. Hygrobot: A self-locomotive ratcheted actuator powered by environmental humidity. Sci Robot, 2018; 3:eaar2629.

28.

de Rivaz

, Goldberg

, Doshi

, et al. Inverted and vertical climbing of a quadrupedal microrobot using electroadhesion. Sci Robot, 2018; 3:eaau3038.

29.

Baisch

, Ozcan

, Goldberg

, et al. High speed locomotion for a quadrupedal microrobot. Int J Rob Res, 2014; 33:1063–1082.

30.

Liang

, Wu

, Yim

, et al. Electrostatic footpads enable agile insect-scale soft robots with trajectory control. Sci Robot, 2021; 6:eabe7906.

31.

Jensen

, Holm-Jensen

. Energetic cost of running in workers of three ant species, Formica fusca L., Formica rufa L., and Camponotus herculeanus L. (Hymenoptera, Formicidae). Comp Physiol, 1980; 137:151–156.

32.

Reinhardt

, Blickhan

. Level locomotion in wood ants: Evidence for grounded running. J Exp Biol, 2014; 217:2358–2370.

33.

Erdem

, Chen

, Mohebbi

, et al. Thermally actuated omnidirectional walking microrobot. J Microelectromech Syst, 2010; 19:433–442.

34.

Ebefors

, Mattsson

, Kälvesten

, et al. A Walking Silicon Micro-Robot. In: 1999 Proceedings of the 10th International Conference on Solid-State Sensors and Actuators, Sendai, 1999; pp. 1202–1205.

35.

Greenspun

, Pister

First Leaps of an Electrostatic Inchwormmotor-Driven Jumping Microrobot. In: Hilton Head Solid-State Sensors, Actuators, and Microsystems Workshop, Hilton Head Island, 2018; SC Vol 3.

36.

Driesen

, Rida

, Breguet

, et al. Friction Based Locomotion Module for Mobile MEMS Robots. In: 2007 IEEE/RSJ International Conference on Intelligent Robots and Systems, San Diego, CA, 2007; pp. 3815–3820.

37.

Contreras

, Drew

, Pister

KSJ

. First Steps of a Millimeter-Scale Walking Silicon Robot. In: 2017 IEEE 19th International Conference on Solid-State Sensors, Actuators and Microsystems, Kaohsiung, Taiwan, 2017; pp. 910–913.

38.

Jayaram

, Shum

, Castellanos

, et al. Scaling Down an Insect-Size Microrobot, HAMR-VI into HAMR-Jr. In: 2020 IEEE International Conference on Robotics and Automation (ICRA), Paris, 2020; pp. 10305–10311.

39.

Doshi

, Goldberg

, Sahai

, et al. Model Driven Design for Flexure-Based Microrobots. In: 2015 IEEE/RSJ International Conference on Intelligent Robots and Systems, Hamburg, 2015; pp. 4119–4126.

40.

Birkmeyer

, Peterson

, Fearing

. DASH: A Dynamic 16g Hexapedal Robot. In: 2009 IEEE/RSJ International Conference on Intelligent Robots and Systems, St. Louis, MO, 2009; pp. 2683–2689.

41.

Goldberg

, Zufferey

, Doshi

, et al. Power and control autonomy for high-speed locomotion with an insect-scale legged robot. IEEE Robot Autom Lett, 2018; 3:987–993.

42.

Churaman

, Currano

, Morris

, et al. The first launch of an autonomous thrust-driven microrobot using nanoporous energetic silicon. Microelectromech Syst, 2012; 21:198–205.

43.

, Zhu

, Liu

, et al. A Fast-Moving Electrostatic Crawling Insect. In: 2017 IEEE 30th International Conference on Micro Electro Mechanical Systems, Las Vegas, NV, 2017; pp. 761–764.

44.

Hollar

, Flynn

, Bellew

, et al. Solar Powered 10 mg Silicon Robot. In: The Sixteenth Annual International Conference on Micro Electro Mechanical Systems, Kyoto, 2003; pp. 706–711.

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.

1.21 MB

0.09 MB

16.02 MB

0.78 MB

0.00 MB