Nano Energy 114 (2023) 108644

Contents lists available at ScienceDirect

Nano Energy
journal homepage: www.elsevier.com/locate/nanoen

Self-powered bionic antenna based on triboelectric nanogenerator for

micro-robotic tactile sensing
Dekuan Zhu a, Jiangfeng Lu a, Mingjie Zheng a, Dongkai Wang a, Jianyu Wang a, Yixin Liu a,
Xiaohao Wang a, Min Zhang a, *
a
Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen 518055, China

A R T I C L E I N F O A B S T R A C T

Keywords: Antennae are the main sensory organs of insects which allow them to gather information from environment
Antenna through direct contact and active sensing. In this study, a self-powered bionic antenna (SBA) was developed to
Triboelectric nanogenerators emulate the structure and function of insect antenna for robotic tactile sensing. The sensing component of SBA
Self-powered sensors
based on the triboelectric nanogenerator (TENG) technology enables the conversion of mechanical stimuli into
Porous structure
Micro-robot
voltage signal without requiring external power source, and is capable of identifying different contact materials.
Robotic navigation To imitate the active sensing performance of insect antenna, a two-stage actuator was designed and integrated
into the SBA, enabling active sensing in both the horizontal and vertical planes. The ultra-sensitive, lightweight
and drivable design of SBA renders it highly compatible with micro soft robots featured with low actuation force
and compact size, facilitating both passive and active real-time sensing and making it a promising solution for
future robotic tactile systems.

1. Introduction remain unsatisfactory, as the sensing section is relatively heavy and

cumbersome to install. Additionally, current hair-like sensors are typi­
Organisms in nature have piqued significant research interest due to cally fixed on the robots’ bodies, and the sensing process depends on the
their remarkable abilities and effective solutions for perceiving the overall movement of the whole robot.
environment. Inspired by creatures with diverse sensing mechanisms, In nature, insect antennae resemble mammalian whiskers in
researchers have developed various sensors to equip robots with the appearance, but have a completely different biological structure and
capacity to sense external stimulation and assess motion performance sensing mechanism. The types of antennae vary among insect species,
[1]. In the field of navigation and obstacle avoidance, for instance, with examples including the geniculate antennae of ants and bees,
typical bio-inspired sensors can be divided based on the sensing prin­ filamentous antennae of cockroaches, and rod antennae of beetles[17].
ciple into tactile[2–5], visual[6–8], and ultrasonic sensors[9,10]. Structurally, antennae are all composed of the scape, pedicel, and fla­
Although other types of sensors have advantages in detection range and gellum[18], with various and numerous receptors distributed on the
efficiency, the tactile sensor is the only one that directly reflects collision flagellum, enabling insects to sense stimuli such as mechanical, odor,
and contact force between the robots and environment. Currently, re­ temperature, and humidity[19], and the generated signals are trans­
searchers have drawn inspiration from animals’ hairs and whiskers to mitted to the insect’s brain through the antennal nerve. The scape and
design and fabricate hair-like tactile sensors that allow robots to avoid pedicel parts of insect antenna are driven by associated muscle fibers
nearby obstacles through bump-based or touch-based motion, particu­ and can be controlled at more than two degrees of freedom (DOF), such
larly in dark conditions[11–16]. Similar to the function of mammalian as horizontal sweep and vertical swing, providing them with the ca­
hair and whiskers, the hair-like fiber of these sensors usually has fixed or pacity to actively sense the environment with the receptor directly
very limited ranges of motion, and the contact force is amplified by a contact the external objects[20]. Although a few researches have suc­
leverage structure and transmitted to the receptors around the root of cessfully applied the sensing mechanism of antennae on machines[21,
the fiber. When equipped on micro-scale robots with light weight and 22], these sensors do not meet the specific requirements for micro-scale
low driven force, reliability and adaptability of such hair-like sensors robots in volume and weight.

* Corresponding author.
E-mail address: zhang.min@sz.tsinghua.edu.cn (M. Zhang).

https://doi.org/10.1016/j.nanoen.2023.108644
Received 7 April 2023; Received in revised form 23 May 2023; Accepted 24 June 2023
Available online 26 June 2023
2211-2855/© 2023 Elsevier Ltd. All rights reserved.
D. Zhu et al. Nano Energy 114 (2023) 108644

As a novel method to create sensors with properties of small, light, that utilizes the TENG technology to generate signals by directly con­
battery-free and less wired, self-powered sensing mechanism in robotics tacting external objects and sensitive to different contact materials. The
have drawn wide concern. For the past few years, the triboelectric nerve section is a metal wire that connects the sensing section and the
nanogenerator (TENG) has emerged as a cutting-edge technology and muscle fibers physically and electrically. The muscle fibers are
provides a promising solution for the development of flexible and composed of a two-stage actuator that mimics the contraction and
stretchable self-powered sensors[23–25]. TENG is a technology based on relaxation of insect muscle fibers, controlling the motion performance of
the coupled effect of electrification and electrostatic induction between the whole antenna. The tactile sensing system based on the bionic an­
two different materials, where electrons transfer from one material to tenna was equipped on an insect-scale piezoelectric robot to assist it in
another, resulting in a voltage difference between the two materials. performing tasks in various scenarios, such as obstacle avoidance, in­
Since the introduction of TENG in 2012[26], research into various struction acquisition and environmental recognition. This work dem­
self-powered sensors has flourished, including pressure sensors[27,28], onstrates the light-weight, drivability, and versatility of the bionic
vibration sensors[29,30], acceleration sensors[31,32], and acoustic antenna and reveals the potential of TENG-based sensors in meeting the
sensors[33]. challenges of micro-robots operating in complex environments.
Drawing inspiration from the function and structure of insect
antennae, this paper presents design and fabrication of a self-powered 2. Results and discussion
bionic antenna (SBA) with the features of light, compact and drivable,
for obstacle avoidance of soft micro-robots. The bionic antenna com­ 2.1. Mechanism of self-powered bionic antenna
prises three main components: a sensing section, a nerve section, and
muscle fibers. The sensing section is made of a porous conductive sponge The structure and function of a typical insect antenna, cockroach for

Fig. 1. (a) Anatomy schematic of cockroach antenna system and the process of information transfer from environment to the brain. (b) Optical image of the SBA
equipped micro-robot, which is placed on a leaf. The enlargements are SEM of ACES sensor and 3D model of the two-stage actuator. (c) Schematic diagram of the
working mechanism of the ACES sensor on the bionic antenna. (d) Schematic illustration of the application of SBA in the robotic active sensing.

2
D. Zhu et al. Nano Energy 114 (2023) 108644

example, composed of scape, pedicel, and flagellum are shown in Fig. 1 as Fig. 1(b). As above, the SBA consists of three sections: the silver
(a). The flagellum houses a large number of diverse receptors, including nanowire (AgNW) coated Ecoflex sponge (ACES), a metal conductive
mechanoreceptors, chemoreceptors, contact chemoreceptors, thermo­ wire, and a two-stage actuator. The ACES functions as the sensory
receptors, and CO2 receptors[34]. These receptors respond to relevant component of the bionic antenna, mimicking the function of insect an­
stimuli by generating signals that are transmitted to the insect’s brain tenna receptors. At the macro level, ACES is porous silicone elastomer
through the antennae nerve. Furthermore, the presence of muscle fibers with low density; microscopically, the silicone skeleton is covered by
in the scape section provides antennae with multiple DOF, which en­ AgNWs which serve as both triboelectric material and conductive
hances the range of movement and facilitates active sensing to the network on the silicone surface. The electron microscopic image (SEM)
environment. of the ACES can be seen in Fig. 1(b). According to the working mecha­
Inspired by the mechanism of sensing, locomotion, and signal nism of TENG, the ACES can be classified as a single-electrode TENG and
transmission of insect antenna, in research we have designed and is capable of signal transmission owing to the AgNWs conductive
assembled SBA on the soft piezoelectric micro-robot driven by mecha­ network.
nisms of our previous work[35,36], endowing the robot with both active Similar to how insect antennae use mechanoreceptors and contact
and passive sensing ability. An optical image of the prototype is shown chemoreceptors to gather information about collision and pheromones

Fig. 2. (a) Schematic of the fabrication process of the ACES sample. (b) Optical image of the 10 mm × 10 mm × 10 mm ACES sample. (c) SEM of the ACES sample.
(d) Electrical conductivity test results (the embedded images illustrate the compressing test process). (e) Schematic diagram of ACES working at external TENG and
an enlargement diagram of a pore structure. (f) Schematic diagram of ACES working at internal TENG.

3
D. Zhu et al. Nano Energy 114 (2023) 108644

in the environment, the SBA can obtain information about contact force surface.
and material properties due to the TENG mechanism between the ACES When the external pressure is strong enough (> 100 Pa), the pores of
and external objects (as depicted in Fig. 1(c)). When the ACES comes the ACES can be compressed. To illustrate the process, a single pore of
into contact with an unknown object in the environment, equal amounts the ACES is represented in Fig. 2(f), which is simplified as two opposing
of opposite charges are generated on the surfaces of the object and the silicone plates with AgNWs distributed on the surface. In certain areas,
ACES. Upon separation, the AgNWs network within the ACES conducts the AgNWs are exposed, while in others, the silicone structure is
the charges on the sponge surface to the ground terminal and generates exposed. Therefore, the aligned AgNWs and silicone frame form a pair of
electrical signals, and the magnitude of the signals will reflect material triboelectric materials. In a state without any external force, there is no
properties of external objects and the mechanical stimulation. The metal transfer of charge between AgNWs and silicone due to insufficient
wire carries electrical signals from the ACES to the microprocessor, contact. When external pressure is applied on the ACES, the AgNWs are
similar to the way that insect nerves carry signals to the brain. forced to tightly contact with the silicon frame, leading to contact
The design of the two-stage actuator, as depicted in Fig. 1(b), was electrification on the surface of the pore structure. As triboelectric
inspired by the movement patterns of the cockroach antenna, with charge density of silicone polymer is more negative than metal silver,
simplifying the axis and redundant DOFs (shown in Fig. S1). The two- electrons tend to transfer from AgNWs to silicone frame surface. When
stage actuator mimics muscles in the antenna of insect, endowing SBA the pressure is released and the compressive strain on the sample is
with the ability of locomotion in horizontal and vertical planes. The first reduced, the elastic pores tend to return to their original shape. The
stage of the SBA actuator controls horizontal sweep, which endows the separation of the opposite polarities results in changes in electrical po­
robot with obstacle avoidance and search capabilities, ensuring valid tential on the AgNWs. As the pores deform again, the electrons are
movement. The second stage controls vertical swing, enabling the robot repelled back to the ground, resulting in an instantaneous current and
to detect the ground and edges, thereby ensuring of movement safety voltage change in the opposite direction. The internal TENG open-circuit
(Fig. 1(d)). With the ability of the ACES to recognize different materials, voltage (VOC(In)) can be calculated by the following equation[38]:
micro-robots equipped with the SBA will receive distinct signals upon
σIn x(t)
contact with different materials. This feature can be utilized to provide VOC(In) = # (2)
ε0
"material instructions" to the robots by setting certain materials on
specific areas. where x(t) is the time-dependence distance between the opposite AgNW
and silicone structure. σ In is the triboelectric charge density between
2.2. Preparation and working mechanism of AgNW-coated Ecoflex AgNW and silicone skeleton surface.
sponge
2.3. Sensing properties of ACES
Based on TENG technology, ACES is a porous conductive elastomer
serving as the sensing component in the SBA. The preparation process of To characterize the sensing properties of the ACES, the samples with
ACES is depicted in Fig. 2(a). Sugar particles serve as sacrificial material dimension of 10 mm × 10 mm × 8 mm were prepared. Figs. 3(a)-3(c)
to fabricate porous Ecoflex sponge and AgNWs in ethanol are used for illustrate the external TENG properties and Figs. 3(d)-3(h) describe the
coating. The size and shape of ACES can be adjusted by modifying the properties about internal TENG. The testing process of external TENG is
size and shape of the 3D-printed mold. Figs. 2(b) and 2(c) show the illustrated by the photographs embedded in Fig. 3(b). In the testing
optical and SEM images of an ACES sample with a dimension of process, the ACES sample was set on testing platform and the bottom of
10 mm × 10 mm × 10 mm, respectively. From the enlargement in Fig. 2 ACES sample was connected with a conductive copper foil tape, which
(c), AgNWs are fixed on the silicone framework and form a conductive was grounded through the electrometer. Connected to the dynamom­
network, working as both triboelectric layer and conductive path. Fig. 2 eter, a flat testing plate with Polytetrafluoroethylene (PTFE) attached
(d) shows the compressibility and conductivity of a 10 mm × 10 mm was above the ACES sample and centrally aligned. In the initial state, the
× 8 mm ACES sample, which could be compressed over 80% while still plate contacted ACES surface without contact force. Then the plate was
maintaining stability. In the compressing process, the resistance moved upward, rendering a set separate displacement D between the
changed from 10.3 Ω to 6.1 Ω. The ACES is also qualified to stretch, plate and ACES surface and then back to the start contact position,
bend and twist, which is described as Fig. S2. Conductive property in completing a contact-separate cycling. In this process, the signal of VOC
tensile is further described by the experiments in Fig. S3, which illus­
(Ex) was recorded. Fig. 3(a) shows the VOC(Ex) signals under 0.5 Hz
trates the favorable electrical conductivity and stability. contact-separate cycling process with different D. The variation ten­
Depended on the degree of ACES deformation, the single-electrode dency of peak-to-peak open-circuit voltage of external TENG (VPP(Ex)) is
TENG working mechanism of ACES can be further categorized into shown in Fig. 3(b). The parameter D changed from 0.1 mm to 6 mm
external TENG and internal TENG. As depicted in Fig. 2(e), when resulting VPP(Ex) progressively increasing. As Eq. (1) shows, VOC(Ex) is
external pressure is insufficient to induce deformation of ACES (＜ also determined by σ Ex, which is depended on the external material.
100 Pa), the output signal is determined by the contact event between Consequently, in Fig. 3(c), 8 materials were utilized to explore the in­
the surface of the external object and the surface of ACES. When these fluence to VPP(Ex) in the same testing process of Fig. 3(b) (with D = 6 mm
two surfaces come into contact, an equal number of opposite charges are and 0.5 Hz cycling frequency). Among the materials, PTFE has the
generated on both surfaces. Upon separation, a portion of the charges on largest output as 1.75 V in the polymer materials. Metal materials, such
the ACES surface will flow to the ground due to a reduction in the as copper and aluminum, generated the opposite output, due to their
attraction of charges. Upon re-approaching the object, electrons will be positive charge density compared to ACES. The distinction of output
repelled to the ground, resulting in reverse current and altering the signals in values and symbols makes it possible for the ACES sensor to
potential of ACES. The external TENG open-circuit voltage (VOC(Ex)) can distinguish materials.
be calculated by the following equation[37]: The internal TENG testing process of the ACES is similar to the
σ Ex d(t) external TENG, but with a different moving direction of the test plate.
VOC(Ex) = # (1) The optical images embedded in Fig. 3(e) illustrate the testing process.
ε0
Initially, the test plate made contact with the surface of the ACES sam­
where d(t) is the time-dependence distance between the ACES surface ple, and then it moved downward to compress the ACES until the set
and the external object surface. ε0 is the permittivity of free space and contact force F was reached. Afterward, the test plate was returned to
σ Ex is triboelectric charge density between external material and ACES initial position, releasing the ACES and completing a compress-release

4
D. Zhu et al. Nano Energy 114 (2023) 108644

Fig. 3. (a) External TENG open-circuit voltage (VOC(Ex)) with varying separate distance D. (b) Peak-to-peak open-circuit voltage in external TENG (VPP(Ex)) with
varying D. The optical images embedded are the external TENG testing process. (c) VPP(Ex) of different contact materials. (d) Internal TENG open-circuit voltage (VOC
(In)) of ACES with different applied force (F). (e) Peak-to-peak open-circuit voltage in internal TENG (VPP(In)) with varying F. The optical images embedded are the
Internal TENG testing process. (f) Response and recovery time of the ACES sample. (g) VPP(In) with the compress frequency varying. (h) Stability and durability
performance of ACES sample.

cycle. Because σIn in Eq. (2) is depended on the ACES itself, the value of As the frequency continues to increase, the voltage output rises slightly
VOC(In) is influenced by x(t), which is further determined by the applied and then decreases, indicating the optimal working frequency of ACES is
force. The output signals resulting from F ranging from 0.01 N to 10 N at between 0.1 and 20 Hz The stability of the ACES sample is demonstrated
0.5 Hz loading frequency are depicted in Fig. 3(d). The relationship in Fig. 3(h), where a cycling loading force of 10 N was imposed on the
between F and the peak-to-peak open-circuit voltage of internal TENG ACES for a period of time at 25 ℃ temperature and 40% relative hu­
(VPP(In)), is explored in Fig. 3(e). When F is less than 0.8 N, the sensitivity midity. The results indicate that the ACES can maintain stable output
of the fit line is 1.84 V⋅N− 1, and when the force is larger than 0.8 N, the after 2000 cycles, proving its stability as a mechanical stimuli sensor.
sensitivity is 0.188 V⋅N− 1. It’s worth noting that as compressed, measure ACES samples with different porosity and conductivity were prepared
error will be introduced due to external TENG output of the increased and TENG output performances were tested (Fig. S6-S9). The mass ratio
contact area. The correction method was further explored and supplied of sugar particle: elastomer was used to control the porosity and con­
in the Fig. S4-S5. The response and recovery time of the ACES sample in centration of AgNWs was used to control conductivity of samples. The
the compress-release cycle is shown in Fig. 3(f), where the response time results indicate that higher porosity facilitates the internal TENG output
is 0.13 s and the recovery time is 0.22 s. The frequency response of the but will reduce the external TENG output. Higher conductivity will
ACES under constant external force is depicted in Fig. 3(g). It can be promote both external and internal TENG output.
observed that as the frequency increases from 0 to 0.1 Hz, the voltage
output increases, then remains stable between the range of 0.1–20 Hz.

5
D. Zhu et al. Nano Energy 114 (2023) 108644

2.4. Motion performance of the two-stage actuator

FS = FP # (3)
To mimic the motion of insect antenna, a two-stage actuator was The contraction and recovery process of the SMA spring was modeled
designed for the structure of SBA, as shown in Fig. 4(a). The first stage of utilizing the Lagoudas FEM simulation[39]. Combining the simulation
the actuator (actuator 1) controls the horizontal sweep of SBA, allowing results of the PET film and the SMA spring, the effect of temperature (T)
it detect objects above the ground. The second stage (actuator 2) con­ on the actuator unit deflection angle α was investigated. The working
trols the vertical swing movement, enabling SBA to distinguish different temperature in the FEM simulation was set from room temperature
terrains and ensure driving safety. In both stages of the actuator, shape (25 ℃) to the maximum temperature (140 ℃), taking into consideration
memory alloy (SMA) springs function as muscles, providing the power to of the maximum heat resistance temperature of PET film. The simulation
drive the contraction of the actuator, while the PET film serves as the involved a heating and cooling cycle, where T was increased from 25 ℃
skeleton and provides the recovery power for the SMA spring. The two to 140 ℃, and then decreased back to 25 ℃. The relationship between α
stages of the actuator can be simplified as an actuator unit, which is and T in the working process with varying L (from 2 mm to 10 mm) and
characterized by its length (L), width (W), and thickness (t). L represents constant W (3 mm) is depicted in Fig. S14(a). The results indicate that
the length of the unit, which is equal to the length of the SMA spring and for L greater than 6 mm, the maximum deflection angle (αm) reaches
PET film in the segment. W and t represent the width and thickness of the approximately 180◦ . For L less than or equal to 4 mm, the flexural ri­
PET film in the segment. Information about the fabrication process of gidity of the PET film is greater, and the Fs generated by the SMA spring
SBA is shown in Fig. S10 and the detail of mounting SMA spring on the is insufficient to cause continuous deformation of the PET film, resulting
PET frame is shown in Fig. S11. in a very limited deflection angle at distal end. It is worth noting that the
To control the motion of the SBA, it is necessary to determine the L also affects the initial offset angle (α0), with longer units possessing a
motion mode of the actuator unit. The common working mechanism of larger α0. This is due to the initial stress between the SMA spring and
the actuator unit is depicted as Fig. S12. Both the PET film and the SMA PET film, which is detailed in Fig. S15.
spring were separately simulated using the finite element method (FEM) The parameter L was respectively set for actuator 1 and actuator 2 as
in COMSOL (Figs. 4(b) and 4(c)) and the related parameters is shown in 8 mm and 6 mm with consideration of sweep region, initial deflection
Tab. S1-S3. The simulation model of the PET film is illustrated in Fig. 4 angle and unit size. To optimize the parameter W of the actuator unit,
(b), which also shows the stress distribution in sectional view. The strain the plot of α-T is shown in Fig. S14(b) with varying W (2–4 mm) and
force (Fs) generated by the contraction of SMA spring bends the PET constant L (6 mm). The results indicate that the distal end achieves
film, generating a deflection angle (α) at the distal end. The relationship larger αm when the width parameter is set to 2 or 3 mm. Considering the
between Fs and the length of the contracted SMA spring (L’) is shown in need for compact structure on robots, a width of 2 mm was set as the
Fig. S13, where the initial length L is varied from 2 mm to 10 mm while unit width.
W and t are set constant with 3 mm and 100 µm. As L increases, the PET According to the optimized parameters, the two-stage actuator was
film becomes relatively easily deformed by the SMA spring contraction. fabricated and tested. Fig. 4(d) shows the optical images of the two-stage
In Fig. 4(c), the bending of the PET film generates a constraining force actuator motion process with current input of 1 s Fig. 4(e) shows the
(Fp) that restricts the contraction of the SMA spring. As Fs and Fp can be relationship of α and current input in room temperature.
characterized as a pair of opposing forces, the following relationship
exists:

Fig. 4. (a) The structure of the self-powered bionic antenna and the actuator unit. (b) FEM simulation model of the PET film of the actuator unit. (c) FEM simulation
model of the SMA spring. (d) Optical images of the two-stage actuator working at different current. (e) The relationship between deflection angle α and the current
input at 1 s power-on time.

6
D. Zhu et al. Nano Energy 114 (2023) 108644

2.5. Application in robotics for environment recognition sensing system performed stable under 3.5 m/s wind speed as Fig. S18
illustrates. And material identification test was conducted based on the
Two SBAs were mounted on a miniature piezoelectric soft robot with sensing system (Fig. S19).
electrostatic steer footpads to endow robots with the ability to gather Different scenarios were established to showcase the movement and
information about surroundings through both passive and active touch, sensing capabilities of the SBA. The workspace was a 30 cm × 30 cm
similar to the way cockroach antennae function. The lightweight design square with building blocks. Fig. 5(b) demonstrates the passive sensing
of SBA adds to its appeal for integration with miniature robotic systems. process of the SBA. In the experiment, a blue building blocks wall was
As Fig. S16 shows, the SBA weighs only 0.07 g, making it ideal for constructed as an obstacle for the robot and the SBA was kept stationary
attachment to the micro-robot with a weight of 0.38 g. on the robot. As the robot operated far away from the wall, there was no
The driving mechanism of the micro-robot is shown in Fig. S17 and change in voltage signal output from the ACES sensor and the robot keep
described in our previous work[35,36]. To control the movement of the moving forward. When the ACES on the distal end made contact with the
robot in response to signals generated by ACES, a control system was block wall, a rise in the voltage signal was generated, indicating that the
developed, as shown in Fig. 5(a). The system is composed of a robot had come into contact with an obstacle. After the ACES separated
voltage-controlled current source (VCCS), voltage signal generator and from the wall, the voltage signal dropped back to its original level,
amplifier, SBA, the robot body, and a computer control system. The indicating that the robot had departed from the obstacle. Fig. 5(c) dis­
TENG signal generated by ACES is collected by an electrometer and plays the signal output during the passive sensing process. Supporting
transmitted to the computer, which then controls the power and direc­ information, Movie S1 provides a visual demonstration of the passive
tion of the robot’s movement. In the active sensing mode of SBA, the sensing process.
computer controls switching of the SMA spring to regulate heating time Supplementary material related to this article can be found online at
and current, rendering specific actuators to respond as desired. The doi:10.1016/j.nanoen.2023.108644.

Fig. 5. (a) Schematic of the control system and electrical connection of the SBA and the micro-robot. (b) Photographs of the passive sensing mode of the SBA in
robotic obstacle avoidance. (c) The open-circuit voltage (VOC) signal in (b).

7
D. Zhu et al. Nano Energy 114 (2023) 108644

In addition to passive sensing, active sensing is a more crucial ability veer instead of moving straight. The SBA also swung at larger ampli­
of insect antenna. During active sensing, insects not only sense me­ tudes to detect steps and slopes when the first scan was no signal
chanical stimuli from active contact events, but also make attempt to returned. Since both step and slope terrains fall into the same category of
collect material information about environment, allowing them to make terrain that involves a vertical change in height, but are still safe and
informed decisions about surroundings and navigate accordingly. navigable for the micro-robot. They returned same signal when SBA
Inspired by the behavior of insects, a demonstration of active sensing swung at the large deflection angle. The output signals are shown in
and information collection was designed as Fig. 6(a) shows. In the Fig. 7(b), with every detection the SBA swinging twice. Video of the
experiment, a complex maze was constructed, featuring one entrance vertical active sensing is presented in Supporting information, Movie S3.
and four exits. The robot had to use horizontal touch to sense the walls Supplementary material related to this article can be found online at
and avoid collisions, and find its way to the target exit 3. At each fork in doi:10.1016/j.nanoen.2023.108644.
the maze, different materials were placed on the walls, including PTFE,
Aluminum, and ABS (Acrylonitrile Butadiene Styrene, building block 3. Conclusion
itself). When SBA contacted different materials, the output signal
changed variously due to the TENG sensing mechanism of ACES. Thus, Inspired by the structure and function of the insect antenna, a novel
the robot could receive "material instruction" from the wall, similar to self-powered sensor SBA is proposed and investigated to endow robots
insects collecting chemical information from environment. The material with tactility, which exhibits merits of integration, light-weight, low-
instruction could be used to pre-set the robot’s movement path by cost, drivable and stability.
linking signals to locomotion. At every detect region, the SBA horizon­ The SBA consists of three components: the ACES, a conductive wire
tally sweep twice to ensure the validation of signal. When the voltage and a two-stage actuator. The ACES, which imitates the function of
signal changed by more than + 100 mV, it was regarded as contacting mechanoreceptors and contact chemoreceptors on the insect antenna, is
PTFE and the robot was programmed to turn left. In the range of voltage based on the TENG technology with AgNWs conductive network,
changes from 0 to + 100 mV, the material would be regarded as ABS, achieving self-powered and charge transmission. The ACES can sense
and the robot moved straight ahead. When the voltage changed nega­ both the contact force and material, providing robots with mechanical
tively, the robot was programmed to turn right and the material was stimulation perception and material identification abilities. The two-
regarded as aluminum. Fig. 6(b) displays the voltage signals output with stage actuator is composed of a SMA spring and a PET film skeleton,
different materials. Supporting information, Movie S2 demonstrates the controlling the horizontal and vertical motion of the SBA. The optimized
active sensing process. parameter of the two-stage actuator unit was carried out by the FEM
Supplementary material related to this article can be found online at stimulation, and the influence of heating time and current were explored
doi:10.1016/j.nanoen.2023.108644. by the experiment.
Vertical height variation is prevalent in the natural environment and To validate the sensing reliability and stability of the SBA, a piezo­
poses a challenge to insects’ movement as some terrains exist falling electric micro-robot was equipped with the bionic antenna and tested in
danger. To ensure safe movement, insects utilize their antennae to detect various tasks. The results demonstrate that the SBA endows robots with
vertical height variations and avoid falling danger. This ability was the ability to perform both passive and active sensing, including hori­
replicated in the SBA as depicted in Fig. 7(a). Multiple terrains con­ zontal sweeping and vertical swinging, allowing robots to avoid obsta­
sisting of flat ground, cliff, step, and slope were constructed to test the cles, receive instructions, and avoid falling hazards.
vertical detection ability of the SBA. The SBA was employed to differ­ Future work on the SBA will involve improving its design, control
entiate safe ground and dangerous edges by vertical swing with a fixed system, and integration with other micro-sensors, such as temperature,
amplitude. If the voltage signal changed, it indicated that the front was gas detection, and magnetic field sensors. The use of a pair of SBAs can
flat ground and was safe to pass, but if no signal variation appeared also lead to vector detection, which can be applied in various fields
during swinging, the front may not be the ground and the robot had to including navigation, object tracking, environmental monitoring, and

Fig. 6. (a) Active sensing mode of the SBA at horizontal plane in robotic obstacle avoidance. The micro-robot had to move from entrance to the exit 3, by getting
material instruction from the pre-set materials on the wall. The material was linked to the specific locomotion: PTFE represented turning left, ABS represented go
straight and Aluminum represented turn right. (b) The VOC signals when SBA touched different material (at every corner SBA touches the material twice).

8
D. Zhu et al. Nano Energy 114 (2023) 108644

Fig. 7. (a) Active sensing mode in vertical plane of the SBA. The micro-robot had to utilize SBA to distinguish different terrain, in order to avoid falling (edge) and
find possible paths forward (ground, step and slope). At last, the robot completed safe height reduction by "L" movement. (b) The VOC signals when SBA touched at
different terrain (every detection SBA swing twice).

hazard detection. the preset shape with mounting hole and fold line by paper shears. The
PET film was folded into the shape of SBA frame and mounted with SMA
4. Methods springs as fig. S11 shows, forming the two-stage actuator. The ACES
sample was stamped by a punch with 3 mm × 3 mm dimension, pro­
4.1. Preparation of ACES ducing the ACES sensor. A metal conductive wire inserted into the ACES
sensor at one end, and was adhered by aluminum tape on the PET frame
The prepolymer of Ecoflex 00–30 (Smooth-On, Inc.) was mixed by support platform at another end, completing the fabrication of SBA.
1 A:1B. Sugar particles (200 ~ 400 µm diameter) and prepolymer of
Ecoflex 00–30 were mixed with a mass ratio of 10:3. The mixture was 4.3. Fabrication of the micro-robot
stirred for 15 min and then poured into a 3D-printed mold to cure for
12 h. After curing, the solid mixture was demolded and cleaned with A 28-μm-thick PVDF film (PolyK Technologies, LLC), tailored to
40 ℃ deionized water in an ultrasonic cleaner for 6 h to dissolve the 26 × 22 mm, was deposited with 6-μm-thick silver (Ag) film as elec­
sugar particles. The resulting porous Ecoflex sponge was dried in a trodes on both sides by slot die coating method. One side of the PVDF
convection oven at 80 ℃ for 2 h to remove trapped water in the pores, then was adhered to a PI tape (120-μm-thick for polyimide and 25-μm-
and then soaked into AgNWs in ethanol (CST-NW-S40, COLDSTONES thick for the adhesive) to form PVDF/PI unimorph. Another side of the
TECH, Inc.) for 2 min. After soaking, the Ecoflex sponge was dried in the PVDF was adhered the front and rear legs of robot, which was fabricated
convection oven at 80 ℃ for 15 min, vaporizing ethanol and leaving the by the same PI film. The electrostatic steer pad consisted of a 25-μm-
AgNWs fixed on the silicone frame, completing the preparation of ACES. thick PET frame and a 5-μm-thick PI film with Cr/Al (10 nm/100 nm)
The size and shape of ACES can be adjusted by modifying the size and deposited as the electrode, which can be found in our previous work
shape of the 3D-printed mold. [34]. Eventually, all components were carefully folded and attached.

4.2. Fabrication of SBA 4.4. Characterization

The fabrication of SBA is shown in fig. S10. The PET film was cut into The output voltage of TENG was recorded by an electrometer

9
D. Zhu et al. Nano Energy 114 (2023) 108644

(Keithley, 6514). The displacement was generated by the linear dimensional architected robotic metamaterials, Science 376 (2022) 1287, https://
doi.org/10.1126/science.abn0090.
displacement table (TSA50-C, Zolix). The contact force was recorded by
[10] D. Kang, Y.-J. Cha, Autonomous UAVs for structural health monitoring using deep
the dynamometer (RZ-1, AIKON). A digital multimeter (34470 A, KEY­ learning and an ultrasonic beacon system with geo-tagging, Comput. -Aided Civ.
SIGHT) recorded the resistance variation of the ACES samples when Infrastruct. Eng. 33 (2018) 885–902, https://doi.org/10.1111/mice.12375.
compressed or stretched. [11] P. Xu, X. Wang, S. Wang, T. Chen, J. Liu, J. Zheng, W. Li, M. Xu, J. Tao, G. Xie,
A triboelectric-based artificial whisker for reactive obstacle avoidance and local
To drive the micro-robot, signal generator (DG1062, RIGOL) and mapping, Research 2021 (2021) 1–10, https://doi.org/10.34133/2021/9864967.
amplifier (ATA-2082, Agitek) were utilized. To control the movement of [12] J. An, P. Chen, Z. Wang, A. Berbille, H. Pang, Y. Jiang, T. Jiang, Z.L. Wang,
the SMA spring on the SBA, the voltage controlled current source (KW- Biomimetic hairy whiskers for robotic skin tactility, Adv. Mater. 33 (2021),
e2101891, https://doi.org/10.1002/adma.202101891.
VCCS1000, KangWei) was utilized. [13] J.Z. Gul, K.Y. Su, K.H. Choi, Fully 3D printed multi-material soft bio-inspired
whisker sensor for underwater-induced vortex detection, Soft Robot 5 (2018)
122–132, https://doi.org/10.1089/soro.2016.0069.
CRediT authorship contribution statement
[14] W. Deer, P.E.I. Pounds, Lightweight whiskers for contact, pre-contact, and fluid
velocity sensing, IEEE Robot. Autom. Lett. 4 (2019) 1978–1984, https://doi.org/
Dekuan Zhu: Conceptualization, Methodology, Formal analysis, 10.1109/lra.2019.2899215.
[15] K. Takei, Z. Yu, M. Zheng, H. Ota, T. Takahashi, A. Javey, Highly sensitive
Software, Writing Draft, Project administration. Jiangfeng Lu: Data
electronic whiskers based on patterned carbon nanotube and silver nanoparticle
curation, Software, Writing draft. Mingjie Zheng: Conceptualization, composite films, Proc. Natl. Proc. Natl. Acad. Sci. U.S.A 111 (2014) 1703–1707,
Investigation. Dongkai Wang: Methodology, Resources. Jianyu Wang: https://doi.org/10.1073/pnas.1317920111.
Conceptualization. Yixin Liu: Methodology. Xiaohao Wang: Supervi­ [16] Z.Q. Yu, Y. Guo, J.J. Su, Q. Huang, T. Fukuda, C.Y. Cao, Q. Shi, Bioinspired,
multifunctional, active whisker sensors for tactile sensing of mobile robots, IEEE
sion. Min Zhang: Supervision. Robot. Autom. Lett. 7 (2022) 9565–9572, https://doi.org/10.1109/
lra.2022.3191172.
[17] A. Krishnan, S.P. Sane, Antennal mechanosensors and their evolutionary
Declaration of Competing Interest antecedents, Adv. Insect Phys. 49 (2015) 59–99, https://doi.org/10.1016/bs.
aiip.2015.06.003.
[18] T.J. Prescott, E. Ahissar, E. Izhikevich, Scholarpedia of touch, Atlantis Press
The authors declare that they have no known competing financial (2016).
interests or personal relationships that could have appeared to influence [19] C. Polidori, J.L. Nieves-Aldrey, Diverse filters to sense: great variability of antennal
morphology and sensillar equipment in gall-wasps (Hymenoptera: Cynipidae),
the work reported in this paper. PLoS One 9 (2014), e101843, https://doi.org/10.1371/journal.pone.0101843.
[20] E.M. Staudacher, M. Gebhardt, V. Dürr, Antennal movements and
Data Availability mechanoreception: neurobiology of active tactile sensors, Adv. Insect Phys. 32
(2005) 49–205, https://doi.org/10.1016/S0065-2806(05)32002-9.
[21] H.J. Lee, R. Joyce, G. Kim, W. Kim, S. Han, J. Lee, Antenna-shaped biomimetic
Data will be made available on request. tactile sensor using elastomers and conductive liquid, ACS Appl. Electron. Mater. 4
(2022) 4863–4872, https://doi.org/10.1021/acsaelm.2c00801.
[22] L. Patane, S. Hellbach, A.F. Krause, P. Arena, V. Durr, An insect-inspired bionic
Acknowledgment sensor for tactile localization and material classification with state-dependent
modulation, Front. Neurorobot. 6 (2012), https://doi.org/10.3389/
This work is financially supported by the Shenzhen Fundamental fnbot.2012.00008.
[23] Z.L. Wang, Triboelectric nanogenerators as new energy technology for self-
Research Funds (No. JCYJ20220530143011026). powered systems and as active mechanical and chemical sensors, ACS Nano 7
(2013) 9533–9557, https://doi.org/10.1021/nn404614z.
[24] S. Liu, Y. Li, W. Guo, X. Huang, L. Xu, Y.-C. Lai, C. Zhang, H. Wu, Triboelectric
Appendix A. Supporting information
nanogenerators enabled sensing and actuation for robotics, Nano Energy 65
(2019), 104005, https://doi.org/10.1016/j.nanoen.2019.104005.
Supplementary data associated with this article can be found in the [25] G.Z. Yang, J. Bellingham, P.E. Dupont, P. Fischer, L. Floridi, R. Full, N. Jacobstein,
online version at doi:10.1016/j.nanoen.2023.108644. V. Kumar, M. McNutt, R. Merrifield, B.J. Nelson, B. Scassellati, M. Taddeo,
R. Taylor, M. Veloso, Z.L. Wang, R. Wood, The grand challenges of science robotics,
Sci. Robot 3 (2018), https://doi.org/10.1126/scirobotics.aar7650.
References [26] F.-R. Fan, Z.-Q. Tian, Z. Lin Wang, Flexible triboelectric generator, Nano Energy 1
(2012) 328–334, https://doi.org/10.1016/j.nanoen.2012.01.004.
[27] P.-K. Yang, Z.-H. Lin, K.C. Pradel, L. Lin, X. Li, X. Wen, J.-H. He, Z.L. Wang, Paper-
[1] H. Wang, M. Totaro, L. Beccai, Toward perceptive soft robots: progress and
based origami triboelectric nanogenerators and self-powered pressure sensors, ACS
challenges, Adv. Sci. (Weinh.) 5 (2018) 1800541, https://doi.org/10.1002/
Nano 9 (2015) 901–907, https://doi.org/10.1021/nn506631t.
advs.201800541.
[28] X.X. Zhu, X.S. Meng, S.Y. Kuang, X.D. Wang, C.F. Pan, G. Zhu, Z.L. Wang,
[2] B. Shih, D. Shah, J. Li, T.G. Thuruthel, Y.-L. Park, F. Iida, Z. Bao, R. Kramer-
Triboelectrification-enabled touch sensing for self-powered position mapping and
Bottiglio, M.T. Tolley, Electronic skins and machine learning for intelligent soft
dynamic tracking by a flexible and area-scalable sensor array, Nano Energy 41
robots, Sci. Robot 5 (2020) eaaz9239, https://doi.org/10.1126/scirobotics.
(2017) 387–393, https://doi.org/10.1016/j.nanoen.2017.09.025.
aaz9239.
[29] H. Zhao, M. Shu, Z. Ai, Z. Lou, K.W. Sou, C. Lu, Y. Jin, Z. Wang, J. Wang, C. Wu,
[3] G. Yao, L. Xu, X. Cheng, Y. Li, X. Huang, W. Guo, S. Liu, Z.L. Wang, H. Wu,
Y. Cao, X. Xu, W. Ding, A highly sensitive triboelectric vibration sensor for
Bioinspired triboelectric nanogenerators as self-powered electronic skin for robotic
machinery condition monitoring, Adv. Energy Mater. 12 (2022) 2201132, https://
tactile sensing, Adv. Funct. Mater. 30 (2020), https://doi.org/10.1002/
doi.org/10.1002/aenm.202201132.
adfm.201907312.
[30] S. Li, D. Liu, Z. Zhao, L. Zhou, X. Yin, X. Li, Y. Gao, C. Zhang, Q. Zhang, J. Wang,
[4] G. Gu, N. Zhang, H. Xu, S. Lin, Y. Yu, G. Chai, L. Ge, H. Yang, Q. Shao, X. Sheng,
A fully self-powered vibration monitoring system driven by dual-mode triboelectric
X. Zhu, X. Zhao, A soft neuroprosthetic hand providing simultaneous myoelectric
nanogenerators, ACS Nano 14 (2020) 2475–2482, https://doi.org/10.1021/
control and tactile feedback, Nat. Biomed. Eng. (2021), https://doi.org/10.1038/
acsnano.9b10142.
s41551-021-00767-0.
[31] Y. Chen, Y.-C. Wang, Y. Zhang, H. Zou, Z. Lin, G. Zhang, C. Zou, Z.L. Wang, Elastic-
[5] S. Chun, J.-S. Kim, Y. Yoo, Y. Choi, S.J. Jung, D. Jang, G. Lee, K.-I. Song, K.S. Nam,
beam triboelectric nanogenerator for high-performance multifunctional
I. Youn, D. Son, C. Pang, Y. Jeong, H. Jung, Y.-J. Kim, B.-D. Choi, J. Kim, S.-P. Kim,
applications: sensitive scale, acceleration/force/vibration sensor, and intelligent
W. Park, S. Park, An artificial neural tactile sensing system, Nat. Electron. 4 (2021)
keyboard, Adv. Energy Mater. 8 (2018) 1802159, https://doi.org/10.1002/
429, https://doi.org/10.1038/s41928-021-00585-x.
aenm.201802159.
[6] S. Lowry, N. Sunderhauf, P. Newman, J.J. Leonard, D. Cox, P. Corke, M.J. Milford,
[32] C. Wu, X. Wang, L. Lin, H. Guo, Z.L. Wang, Paper-based triboelectric
Visual place recognition: a survey, IEEE Trans. Robot. 32 (2016) 1–19, https://doi.
nanogenerators made of stretchable interlocking kirigami patterns, ACS Nano 10
org/10.1109/tro.2015.2496823.
(2016) 4652–4659, https://doi.org/10.1021/acsnano.6b00949.
[7] O.M. Omisore, S. Han, J. Xiong, H. Li, Z. Li, L. Wang, A review on flexible robotic
[33] H. Guo, X. Pu, J. Chen, Y. Meng, M.H. Yeh, G. Liu, Q. Tang, B. Chen, D. Liu, S. Qi,
systems for minimally invasive surgery, IEEE Trans. Syst. Man Cybern. 52 (2022)
C. Wu, C. Hu, J. Wang, Z.L. Wang, A highly sensitive, self-powered triboelectric
631–644, https://doi.org/10.1109/tsmc.2020.3026174.
auditory sensor for social robotics and hearing AIDS, Sci. Robot. 3 (2018), https://
[8] N. Van Nhan, R. Jenssen, D. Roverso, Automatic autonomous vision-based power
doi.org/10.1126/SCIROBOTICS.AAT2516.
line inspection: a review of current status and the potential role of deep learning,
[34] H. Nishino, M. Nishikawa, F. Yokohari, M. Mizunami, Dual, multilayered
Int. J. Electr. Power Energy Syst. 99 (2018) 107–120, https://doi.org/10.1016/j.
somatosensory maps formed by antennal tactile and contact chemosensory
ijepes.2017.12.016.
afferents in an insect brain, J. Comp. Neurol. 493 (2005) 291–308, https://doi.org/
[9] H. Cui, D. Yao, R. Hensleigh, H. Lu, A. Calderon, Z. Xu, S. Davaria, Z. Wang,
10.1002/cne.20757.
P. Mercier, P. Tarazaga, X. Zheng, Design and printing of proprioceptive three-

10
D. Zhu et al. Nano Energy 114 (2023) 108644

[35] Y.C. Wu, J.K. Yim, J.M. Liang, Z.C. Shao, M.J. Qi, J.W. Zhong, Z.H. Luo, X.J. Yan, Jianyu Wang is now studying for a master’s degree in
M. Zhang, X.H. Wang, R.S. Fearing, R.J. Full, L.W. Lin, Insect-scale fast moving and Shenzhen International Graduate School, Tsinghua University.
ultrarobust soft robot, Sci. Robot. 4 (2019), https://doi.org/10.1126/scirobotics. He received a bachelor’s degree from Nanjing University of
aax1594. Science and Technology in 2020. His research interests are
[36] J.M. Liang, Y.C.A. Wu, J.K. Yim, H.M. Chen, Z.C. Miao, H.X. Liu, Y. Liu, Y.X. Liu, D. flexible electrode, including multifunctional hydrogel and
K. Wang, W.Y. Qiu, Z.C. Shao, M. Zhang, X.H. Wang, J.W. Zhong, L.W. Lin, physiological electrical signal analysis system.
Electrostatic footpads enable agile insect-scale soft robots with trajectory control,
Sci. Robot. 6 (2021), https://doi.org/10.1126/scirobotics.abe7906.
[37] S. Niu, Z.L. Wang, Theoretical systems of triboelectric nanogenerators, Nano
Energy 14 (2015) 161–192, https://doi.org/10.1016/j.nanoen.2014.11.034.
[38] J. Park, I. Kim, J. Yun, D. Kim, Liquid-metal embedded sponge-typed triboelectric
nanogenerator for omnidirectionally detectable self-powered motion sensor, Nano
Energy 89 (2021), 106442, https://doi.org/10.1016/j.nanoen.2021.106442.
[39] D.C. Lagoudas, Shape Memory Alloys: Modeling and Engineering Applications,
Springer, 2008.
Yixin Liu received his B.E. degree in mechatronic engineering
from Beijing Institute of Technology, China in 2016, and the M.
Dekuan Zhu is now studying for a master’s degree in Shenzhen E. degree in instrument engineering from Tsinghua University,
International Graduate School, Tsinghua university. He China in 2019. He is currently pursuing a Ph.D. degree in data
received a bachelor’s degree from Northwest A&F University in science and information technology in Shenzhen International
2020. His research interest are soft robotics and self-powered Graduate School, Tsinghua-Berkeley Shenzhen Institute (TBSI),
sensors. Tsinghua University. His research interests include wearable
devices, flexible electronics and paper-based electronics.

Xiaohao Wang received the B.E. and Ph.D. from Department

of Precision Instruments and Mechanics, Tsinghua University,
Jiangfeng Lu is currently enrolled in the Shenzhen Interna­ China in 1994 and 1999. He served as a Lecturer in Tsinghua
tional Graduate School at the Tsinghua-Berkeley Shenzhen University from 1998 to 2001 and an Associate Professor from
Institute (TBSI), Tsinghua University, where he is pursuing a 2001 to 2010. He currently serves as a Professor in Tsinghua
master’s degree. In 2021, he graduated with a bachelor’s de­ University from 2010. He is the associate dean of Institute of
gree from Xiamen University. He is interested in soft robotics Ocean Engineering, Tsinghua University and Shenzhen Inter­
for research. national Graduate School.

Min Zhang received the B.E. and M.E. degrees in mechanical

engineering from Jiangsu University and North China Electric
Mingjie Zheng is now studying for a master’s degree in Power University, China in 1998 and 2002, respectively. He
Shenzhen International Graduate School, Tsinghua university. received the Ph.D. degree in mechanical engineering from
He received a bachelor’s degree from Hunan University in Dalian University of Technology, China in 2006. He then
2020. His research interest are soft materials and devices, worked as a postdoctoral researcher in Tsinghua University,
including bioinspired adhesives and robotics. China and The University of Minnesota, USA from 2006 to
2011. In 2012, he joined the Faculty at Tsinghua University,
Shenzhen, China, where he is currently an Associate Professor
with the Graduate School at Shenzhen. He is directing the
Microsensors and Actuators Lab.

Dongkai Wang is now studying as a PhD candidate in Sensors

and Microsystems Lab, Tsinghua-Berkeley Shenzhen Institute
(TBSI), Tsinghua University. He received his bachelor degree
from Harbin Institute of Technology, China, and his master
degree from Johns Hopkins University, USA. His research in­
terest lies in flexible electronics, 3D printing and soft robots.

11