Hailong HU , Fan ZHANG

a School of Aeronautics and Astronautics, Central South University, Changsha 410083, China

b Research Center in Intelligent Thermal Structures for Aerospace, Central South University, Changsha 410083, China

c School of Minerals Processing and Bioengineering, Central South University, Changsha 410083, China

KEYWORDS Analytical modelling;Electronic devices;Interface;Nanocomposites;Self-powered sensors

Abstract Smart sensors are becoming one of the necessities for connecting and detecting surrounding stimuli with tremendous convenience, especially when exploiting a single powerful sensor with multifunctionality.To successfully accomplish the design of a self-powered sensor,serving power is becoming a critical issue because of its continuously consumed energy required by electronics. A variety of nanogenerators aiming for the rational design of self-powered system are reviewed and compared, followed by their recent advances with polymer nanocomposites for self-powered sensors. More importantly, the proposed conceptual design of a self-powered unit/device with triboelectric nanogenerator has been emphasized to eventually realize the practical activities towards multiple detections and human–machine interaction. Finally, challenges and new prospects of rational design of self-powered polymer composite sensors in achieving human–machine interaction/interface are discussed.

1. Introduction

With the advancement of smart cities and Internet of Things(IoT),numerous sensors,as the signal receiving base units,will be installed on various kind of locations,and most of them are installed in inaccessible areas, such as oil and gas pipelines,long-distance transmission lines, long-distance optical cables,oceans, forests, etc. Moreover, the embedding applications of the sensors include structural health monitoring in huge buildings,bridges,substructures of the high-speed rail,and tunnels,etc.In those energy-conversion efficiency applications,sensors are subjected to constant and self-maintaining work for an indefinitely long period of time. However, powering such a large number of sensors is becoming a critical issue that needs to be tackled.One of the alternative methods is to use batteries to serve as a power supplier. However, low capacity is an obstructed and intrinsic defect that requires the battery to face inevitable maintenance and recharge consequences. As a result,using a battery as a source of power will not be suitable in terms of fulfilling the requirements of smart city and IoT.Therefore, the power supply is regarded as a major problem in sensor technology.

To resolve the aforementioned problem (power supply) in sensor technology, the idea of a self-powered sensor and its system was proposed.1Basically, a self-powered sensor is a sensor that automatically gives out an electric signal when mechanically activated without an external power source supply.2Generally, most of the sensors used nowadays are passive, which do not provide any signal with the absence of power supply. However, in terms of the self-powered sensor,the operation power source as provided is self-generated. Triboelectric Nanogenerators (TENGs) and Piezoelectric Nanogenerators (PENGs) exhibit excellent ability to effectively convert various forms of mechanical energy to electric energy,including vehicle movements,3human body motion,4acoustic wave,5etc. In comparison with TENGs, PENGs have the advantages of high sensitivity, lower dimension, durability,and compact structure,hence attracting considerable attention in energy harvesters, self-powered systems and sensor networks.6However, with the development of flexible electronics in implantable biomedical devices,wearable devices,and smart textiles, the frequently used piezoelectric ceramics (such as ZnSnO3,BaTiO3(BTO),PbZrxTi1–xO3)and piezoelectric inorganic semiconductors cannot meet the requirement of ample flexibility due to their intrinsic frangibility.7

Due to the advantages of easy processing,excellent flexibility, and lightweight, piezoelectric polymers exhibited promising potential in wearable and flexible applications.8,9Polyvinylidene Fluoride (PVDF) and related copolymersbased PENGs have been extensively studied to improve piezoelectric output and energy conversion efficiency for practical applications.10–12However, the result of polymer-based PENGs is limited due to the intrinsically relatively low polarization of polymer materials and inferior piezoelectricity to piezoelectric ceramics, or piezoelectric semiconductors, which restricts their applications.13,14To address this drawback,polymer nanocomposites have been configurated by integrating the merits of both polymers and ceramics so as to obtain an optimal result in various applications.15Moreover,nanocomposites comprising flexible polymers and inorganic piezoelectric nanoparticles have been developed for enhancing the piezoelectric output performance of polymeric PENGs,taking advantages of high flexibility of the latter and high piezo-response of the former.6,16,17Among the inorganic piezoelectric nanofillers, BaTiO3nanoparticles have been utilized to enhance the piezoelectric property of polymers due to their advantages of low cost,lead-free and high piezoelectric coefficient.18–20In addition, flexible strain sensors in polymer composites are of special interest because they can undertake large deformation and show high sensitivity instead of metals or semiconductors when being tackled in applications such as personal hearth or sports performance monitoring, soft robotics21–26or human–machine interface.27–31To resolve the increasingly encountered complex environmental detections, integrated sensors or sensor arrays are generally employed with multiple electrodes and interconnects to fulfill the external stimulus detection.32–38However,this will inevitably increase fabrication cost and make the structures pretty complicated. Designing a single sensor with multi-modality is one of the promising solutions that can separately detect temperature, strain, humidity, and other stimuli.31,39More specifically,to realize practical applications via a single sensor,the interactive interfaces shall be integrated with sensor to achieve an interactive system for intelligent/smart sensing,precise digital control, and advanced manufacturing in future.40–43The thermoelectric generator,44–49piezoelectric generator50,51or triboelectric nanogenerator52–58can be used as a desired self-powering unit to power the sensor.

As the wide prevalence of Internet of things,59,60Artificial Intelligence61(AI) and some other intelligent advances such as smart multimodal sensors, robotics and virtual reality,have enormously promoted the progress of effective communication between human and machine interaction/interfacing.62–64

In this work, we present a review of the recent advances of self-powered sensors in polymer composites which prosper the future human–machine interaction. The following aspects will be covered. Firstly, rational design of a self-powered system based on different nanogenerators are thoroughly investigated.The advanced multifunctional strain sensors in polymer composites with superior performance are discussed.Subsequently,to achieve multiple modalities with a single sensor, recent advances of self-powered sensors in polymer composites are reviewed.Moreover,this review places emphasis on discussing the proposed schemes of conceptual design of a self-powered unit/device, where thermoelectric generator, piezoelectric generator or even triboelectric nanogenerator (Fig.1)is proposed with the integration of optical wireless communications and advances of IoT to eventually achieve the practical activities of multiple detections and human–machine interaction.Finally, challenges and new application opportunities of selfpowered polymer composite sensors will be discussed towards achieving human–machine interface.

2. Rational design of a self-powered system

For a self-powered system,the sustainable power supply serves as an inevitably essential part for sensing, signal processing and transmission. To design a self-powered sensor, it is necessary to design a self-power unit to provide the whole energy.The current lithium-ion batteries seem to be a good alternative to provide a steady current to power small-connected devices.However, it can only last for limited few days, which substantially constrain the sustainable development of long-term use of smart devices. Thermoelectric nanogenerator, or triboelectric nanogenerator, or piezoelectric nanogenerator or some other electricity generator such as droplet electricity generator,seems to be promising candidate in serving as the power unit of an integrated system for electronics application. Thermoelectric generator is based on the Seebeck effect to convert heat into electricity by using the high-performance thermoelectric materials. Piezoelectric generator harvests energy through the mechanical vibrations. The mechanism of triboelectric nanogenerator is attributed to electrostatic induction and contact electrification.Based on these varied self-powered sensors fabricated by polymer composites, conceptual configuration of the integrated self-powered sensor system is proposed through the integration of sensor generator, optical wireless communications, and the advances of IoT, which eventually promotes the activities of multiple detections and human machine interaction.

2.1. Nanogenerators based self-powered system

2.1.1. Thermoelectric nanogenerator

Thermoelectric nanogenerator acts as a power generation part.The self-powered sensor is designed based on the Seebeck effect of thermoelectric materials, whose generated voltage can be harvested through the temperature gradient.65The temperature difference can be obtained from any living body or surrounding environment. Thermoelectric materials can be prepared as thin films to make them more suitable for integrating into flexible devices.For the thermoelectric nanogenerator system, holes in p type w and electrons in n-type aim to form the voltage difference and carry the electrical current. As heat starts to flow from hot side to code side through the media of thermoelectrical materials, free charges in semiconductor will also move and convert thermal energy into electrical energy.By using this kind of thermoelectric materials as power generation unit, Kuchle and Love49shared a self-powered sensor technology capable of completing some dynamic and static measurements.

Fig. 2 shows the reported varied designs of thermoelectric nanogenerator to power electronics, where flexible and wearable thermoelectric nanogenerators are made from Bi2Te3in Fig. 2(a).66In addition, photo-thermoelectric nanogenerators comprised of MoS2/Polyurethane (PU) composites as photothermal layer are used for infrared light-harvesting in Fig.2(b).67Fig.2(c)45clearly shows the illustrated thermoelectric effect in stretchable graphene-polymer nanocomposites for self-powered strain sensor system,where temperature distribution along the thermoelectric film is recorded by Infrared (IR)camera (Fig. 2(d)45). To fulfill the practical electronic application with self-powered devices, integrated sensing system with

flexible organic thermoelectric generator as a sensing element has been developed (Fig. 2(e)33), showing its promising as an alternative flexible power sources and offering a new strategy for designing self-powered electronics by the integrated devices. Furthermore, organic electronics are generally featured by their distinct merits of excellent flexibility, low cost,and large-scale production via chemical-processing approaches,which enables organic electronics to be integrated with smart systems for the purpose of complicated tasks.

2.1.2. Triboelectric nanogenerator

The alternative TENG or PENG seems to be a more desired solution.68,69In recent years, triboelectric nanogenerator has been efficiently developed based on contact electrification or electrostatic induction,showing the advantages of high voltage output, lightweight, low cost and tuneable structural design,etc.70Moreover, as the triboelectric nanogenerator can generate electricity from almost all the types of motions, such as touching, vibrating, rotating, etc, it can serve as self-powered sensor for detecting a wide range of motions.71Fig. 3 shows the triboelectric nanogenerator acting as power source, where multiple designs are presented,including single electrode-based sliding triboelectric nanogenerator of Polytetrafluoroethene(PTFE) (Fig. 3(a)72), triboelectric nanogenerator buoy ball exploiting for the application of water wave energy farm(Fig. 3(b)55), paper-based triboelectric nanogenerator for acoustic energy harvesting and sound recording (Fig. 3(c)73),PTFE membrane vibration and electricity generated by the sound wave (Fig. 3(d)73), triboelectric nanogenerator enabling gas detection under the external stimulus surroundings (Fig.3(e)74). Therefore, recent progress of TENG has proven its feasibility to harvest mechanical energy ranging from mechanical motion,vibration,to sound wave,etc.,which can be used as a promising component for self-powered devices and sensor systems.75

Fig. 2 Thermoelectric nanogenerator.

2.1.3. Piezoelectric nanogenerator

Piezoelectric nanogenerator works as an alternative power source.14Our recent experimental results show that the synthesized Carbon Nanofiber (CNF)/BTO/Polydimethylsiloxane(PDMS) polymer nanocomposites nanogenerator can achieve the improved energy output for flexible dielectric electronic devices.76To evaluate the harvested electrical energy in the polymer nanocomposites, the polymer nanocomposites by uniformly blending CNF, BTO, and PDMS are further made into electronic devices, and their piezoelectric voltage output performance is measured as a function of frequency under various CNF contents of 0, 0.5wt% and 1.0wt%, respectively(Fig.4).It is noted that the added BTO contributes to the output voltage owing to its inside piezoelectric active sites.Remarkably, when the added CNF content is 0.5wt%, the maximum output voltage can reach 0.25 V at 20 Hz in Fig. 4(a), which is higher than that of BTO/PDMS polymer nanocomposites without adding CNF in Fig. 4(c). However,when more CNF is added to the nanocomposites, such as 1.0wt% CNF, the degraded output voltage performance is observed, indicating only 0.02 V at its maximum level at 20 Hz in Fig. 4(b). Therefore, the significantly high output voltage is demonstrated in 0.5wt%CNF/BTO/PDMS polymer nanocomposites in Fig.4(d).This can be explained by the critical percolation theory, as lower filler content below the vicinity of percolation contributed to the formed thousands of mini capacitors in the polymer nanocomposites, giving rise to the enhanced dielectric permittivity and thus the output voltage.This kind of experimental design is just taken as an example for illustrating the piezoelectric nanogenerator made by polymer composites, despite their remaining improved energy output.

Fig. 3 Triboelectric nanogenerator acting as power source.

2.2. Droplet-based electricity generator

Fig. 4 Piezoelectric output voltage for CNF/BTO/PDMS, BTO/PDMS polymer nanocomposites and output voltage comparison at 20 Hz.

Droplet-based electricity generator offers another feasible means. Interestingly, the recent advance of droplet-based electricity generator has attracted enormous attention owing to its highly generated power voltage up to 140 V for only one drop of water, which can also be used to light up 100 LED bulbs.77To develop this kind of droplet-based electricity generator, a Field-Effect Transistor (FET) like structure has been designed to achieve instantaneous power density and great energy-conversion efficiency (Fig. 577), showing the increased value up to thousand times when comparing to conventional design without field-effect transistor structure. The principle of this droplet-based electricity generator can be illustrated as follows: triple layers (Al electrode,PTFE film and Indium Tin Oxide (ITO) electrode), which corresponds to the drain, gate, and source of a FET, respectively. When one droplet falls onto the surface of PTFE and begins to spread out and contacts with Al electrode, the whole designed device will be triggered and form a closed current circuit, then charges will be transferred between ITO and Al electrodes, eventually outputting power. As the water droplet keeps falling onto the PTFE surface,tremendously high surface energy density will be accumulated and form a high voltage. Fig. 4 shows the piezoelectric output voltage for different nanocomposites at 20 Hz, where schematic of the device (Fig. 5(a)77) is displayed and four parallel of droplet-based electricity generator devices are prepared on a glass substrate (Fig. 5(b)77). To demonstrate power efficiency of charge accumulated on substrate by droplets (Fig. 5(c)77), one droplet is capable of powering hundreds of commercial LED lights (Fig. 5(d)77). Voltage and current comparison (Fig. 5(e)77and (f)77) between Dropletbased Electricity Generator (DEG) and control devices is also performed to reveal the high energy output by DEG.Control devices are designed based on a lower frequency with less droplets although they share the same loaded charge on the surface of substrate as DEG. This research outcome has a profound meaning on harvesting water energy from new alternatives, such as raindrops instead of conventional oil and nuclear energy,78which will eventually offer a great solution to the upcoming limited renewable energy in the world. On the other hand, this kind of device design can be implemented on a variety of surfaces instead of PTFE, such as the surface of umbrella, ferry, and coastline, etc.

2.3. Piezo-thermoelectric generator

Piezo-thermoelectric generator is proposed to harvest both mechanical and thermal energies.Except for the single thermoelectric generator or triboelectric nanogenerator,alternatively,piezo-thermoelectric generator79has offered a feasible way to simultaneously harvest the waste heat/energy and the mechanical energy (Fig. 6(a)79). Zhou et al.79proposed a piezothermoelectric generator of PVDF based Piezoelectric Generator (PEG) converter (Fig. 6(b)79) and Bi2Te3based flexible thermoelectric generator (Fig. 6(c)79) to use both mechanical and waste heat thermal energy of heating fluids. After configurating the design of series and parallel connections of piezothermoelectric generator, length specific power of 55 μW/cm and 19 μW/cm were respectively achieved under a steadystate flowing fluid. Long-term stability was also performed on this designed piezo-thermoelectric generator,showing good reliability and great potential in the application of robust environments. Therefore, Zhou et al.’s79research work provides a feasible solution to designing a self-powered device by using environmental waste heat and mechanical motions, which paves the way for following design of multiple detections based on a single self-powered sensor.

Fig. 5 Droplet-based Electricity Generator (DEG).77 (Copyright 2020, reproduced with permission from Nature Publishing Group).

Therefore, a variety of generators have been introduced,which include thermoelectric nanogenerator, triboelectric nanogenerator, piezoelectric nanogenerator, the most recently reported droplet based electricity generator, as well as the piezo-thermoelectric generator. In principle, they all show great potential in serving the self-powered based system by offering the continuous power. However, to take the energy density and reliability into consideration, further discussion has to be conducted to identify sensor type in order to achieve the rational design of self-powered sensors with polymer nanocomposites for human–machine interaction.

3. Recent advances in self-powered sensors

3.1. Piezoreisistive strain sensor

Multifunctional strain sensors are designed to realize multidirectional sensing or identify multiple stimuli, while conventional uniaxial strain sensors are generally capable of detecting the motion or strain in one single direction or one stimulus.38,80–84To accomplish the multidirectional sensing,much effort has been devoted to improving the design of conducting networks of strain sensors,such as the structural engineering of sensors in either geometry or shape.85,86However,limited progress has been achieved because of limited sensing range and the instability of networks under large deformations. Generally, multifunctional strain sensors have been categorized into three common types, which are comprising of sensors with self-healing function,87–91integrated sensors with multimodalities,81,83and self-powered sensors.49,55,92–95

A novel type of stretchable piezoresistive strain sensor made of self-healing conductive hydrogel has been introduced,87,90where the conductive hydrogels is ranged from graphene,silver nanowire to carbon nanotube,etc.As a result,fast electrical healing speed and high self-healing efficiency are achieved, showing the values of 3.2 s and 98%, respectively.More interestingly, after the severe deformation, this kind of hydrogel strain sensor can sustain a rather high stretchability up to 1000%, although its gauge factor is only about 1.51.90Human motion detection measurement has been performed on this conductive hydrogel strain sensors, demonstrating a good response and high stability. It is acknowledged that the hydrogen-bonding formed in hydrogel allows the self-healing and reforming ability in strain sensors.For instance,Cai et al.90prepared the cross-linked hydrogel through the reaction between tetrafunctional borate ion and —OH in Polyvinyl Alcohol (PVA) (Fig. 7(a)90). Despite the weak crosslink of hydrogel, it can be dynamically associated and dissociated owing to the proximity of —OH groups and borate ions.Moreover, the adequate mobility in polymer chain and tetrafunctional borate ions supports the hydrogen bond to realize self-healing process without the external stimuli after a certain deformation.

Fig. 6 Piezo-thermoelectric generator realized by waste heat flow, PVDF based Piezoelectric Generator (PEG) converter and Bi2Te3 based flexible thermoelectric generator.79 (Copyright 2020, reproduced with permission from Elsevier).

Except for the hydrogel possessing the self-healing function to strain sensors,80structural engineering is another approach to achieve the self-healing with alterable piezoresistivity,where constructed supramolecular metal–ligand coordination bond and hierarchical conductive network are employed for the structure design in strain sensors.87,88After controlling the morphology of conductive network in elastomer matrix, the tunable piezoresistivity enables a tiny motion monitoring. As shown in Fig. 7(b),87the intrinsic behaviour of metal–ligand coordination bond crosslinks allows the rearrangements of epoxidized nature rubber, which enables the sensors with self-healing efficiency up to 88.6% after deformation or damage occurred in nanostructured sensors integrated with a flexible yarn electrode.87In addition to this, this developed strain sensors can be used to recognize a variety of interactions in human body, such as coughing, deep breathing and pronunciation.

Besides the reported techniques in realizing self-healing strain sensor electronics, techniques of constructing a supramolecular network through hydrogel bonding, metal–ligand interaction, dynamic covalent bonds and ionic interaction/reaction, and incorporating conductive polymers into self-healing materials are becoming other feasible options for developing the self-healing electronics. To improve the shortcomings of soft hydrogel network,a two-step method has been exploited to obtain hydrogel with enhanced mechanical property,where reduced Graphene Oxides(rGO)provides the conductive pathways and a dual-crosslinked hydrogel formed by acrylic acid monomers offers a stable and strong network(Fig. 7(c)88). As a result, the Polyacrylic Acid (PAA)-rGO nanocomposite strain sensors have the advantages of sensing varied human motions, superior self-healing capabilities in both mechanical and electrical damage and the biocompatibility as well. Fig. 7(d)89and (e)89shows the multiple hydrogen bonding of self-healing sensors promoting the human machine interaction, where a facial expression control system is developed to realize the real-time speaking. Based on supramolecular multiple hydrogen bonding elastomer (epoxy natural rubber)in combination with carbon nanotube conductive network,this flexible strain sensor possesses the capability of efficient self-healing with high sensitivity.89

3.2. Integrated sensors with multimodalities

Fig. 7 Piezoresistive strain sensors.

Strain sensors have been commonly used to detect/monitor human motions ranging from simple mechanical stress to complex joints activities; however, only some human motions can be monitored instead of the full representative activities. On the other hand, accurate information of human motions cannot be precisely collected owing to the shortcomings of strains sensors, such as inflexibility and lack of multimodalities.Therefore, the integrated flexible strain sensors are in great need to meet the continuously increasing demands.

Fig. 8 Integrated sensors with multimodalities.

Fig.8(a)30shows a multifunctional flexible health monitoring patch,where reusable sheet,disposable sheet and a variety of sensors are integrated onto a flexible Polyethylene Terephthalate (PET) matrix.30Electrocardiogram (ECG) sensors,temperature sensors and Ultra Violet (UV) sensors are integrated to realize the multifunction in skin electronics,such as monitoring the physical activities of skin temperature, UV light exposure and heart rate,etc. Therefore,this printed multilayered sensor provides a proof-of-concept to design flexible and wearable hearth monitoring electronics with multifunctionalities. Besides the printing technique on designing strain sensor electronics,an integrated textile patch with several components being comprised enables electrophysical monitoring,thermotherapy,activity tracking and even the human–machine interaction.81Electronic textiles are prepared by silver nanowire composites integrated with textiles via a laser patterning process (Fig. 8(b)81), followed by heating process to enhance the bonding between textiles and composites.In addition, electronic textiles can be fabricated via other approaches or methods, such as knitting, weaving, coating, printing and laminating onto fibres or fabrics.96,97However, there remain the issues, such as the limited resolution, low conductivity,instability and complicated fabrication process, which substantially limit the prevalence of integrated sensors with multimodalities.

Despite the current efforts devoted in improving the performance of flexible strain sensors, silk based combo temperature–pressure sensor and layer structured skin electronics are designed to enable the signal monitoring of a number of stimuli, such as temperature, pressure and light, etc (Fig. 8(c)82).The principle of integrated sensors is to well integrate a variety of strain sensors to reach multimodalities,which make complementary progress to a single sensor capable of detecting only one to two stimuli.Moreover,the progress towards improving the performance of integrated sensors lies in perfecting the sensitivity, flexibility, stability and no mutual interference in detecting multiple signals. To further verify the concept and performance of integrated sensors, dual-mode sensors, strain sensor, and supercapacitor (Fig. 8(d)98) are integrated onto a deformable matrix.98The results indicate that this integrated sensor can detect pressure and temperature without any interference. Moreover, supercapacitors enable the dual-mode and strain sensors working wirelessly, showing that it is promising in the application of wireless powered electronics.

3.3. Thermoelectric nanogenerator and piezoelectric sensor

As thermal energy is abundant and universal in surroundings,which can be easily achieved ranging from direct heat source or light source, it is promising to develop thermal energy harvesting technology based on the Thermoelectric Effect(TE)for practical applications. It is reported that a photothermoelectric nanogenerator was fabricated based on the polymer composites of MoS2/PU and TE/Poly(3,4-ethylene dioxythiophene) (PEDOT), showing an electrical output without pronounced spatial temperature gradient.67Moreover, a flexible and wearable thermoelectric nanogenerator realized by composites of Bi2Te3and polyethylene terephthalate film shows a temperature measurement with a resolution of 0.5 K, followed by an output voltage up to 520 mV and a power density of 11.14 mW/cm2.66As a result, all these proposed thermoelectric nanogenerators show great potential into the design of flexible wearable devices as well as low-power wireless sensor networks. However, based on the Seebeck effect of thermoelectric nanogenerator, there are still some breakthroughs waiting to be achieved toward the selfpowered temperature sensor. For instance, the currently harvested output voltage is still quite low with the value up to less than 100 mV at a large temperature difference.

Piezoelectric polymer nanocomposite materials are behaving as one of the most promising components in harvesting energy, sensors and actuators owing to their merit of directly converting mechanical stress to electrical energy or vice versa.99Thus, the mechanical resilience and high piezoelectric coefficient for sensitivity are generally required in piezoelectric polymer nanocomposite materials. Moreover, polymer PVDF ought to meet some specific requirements prior to show strong piezoelectric property,such as large dipole moment in a certain direction, high crystallinity and a perpendicular polar axis to the thickness of a sample.However,when designing a complicated piezoelectric structure, limited mechanical stretchability is hurdling the progress of flexible structural integration toward electronic devices. On the other hand, the underlying mechanism of piezoelectric polymer nanocomposite materials for achieving high piezoelectric performance remains further investigation.

3.4. Current and future challenges in multifunctional strain sensors and their related electronics

Fig. 9100summarizes current and future challenges arised in multifunctional strain sensors and their related electronics.Crucial factor, such as linearity, needs to be considered to achieve a wide range of linear sensitivity without sensing signal drift or delay. To overcome the limited deformability, structural engineering based on a novel geometrical approach or multilayer interlocking geometry has been exploited to attain ultra-high sensitivity without sacrificing the linear response.

Recently,enormous efforts have been devoted to improving the performance and the multifunctional property in strain sensors for the application of wearable electronics and health monitoring. Furthermore, the interactive platform between advanced electronics and human beings is trying to construct,eventually to achieve the progress in next technology advances,such as human–machine interface, Internet of things and artificial intelligence. To successfully perceive the signal and surrounding stimulus, deep learning and machine learning in AI are becoming the crucial components to realize the essential functionality in artificial intelligence, wearable electronics and health monitoring, where the human–machine interaction will play a predominant role in running all these integrated systems.

Other strategies have been endeavored to overcome the limited sensing range and reach high sensitivity with more accuracy in a variety of directions. For instance, using the prestrained materials101or aligned conductive fillers29,102in polymer-based strain sensors differentiates axis angles or directional response. Interestingly, Kim and colleagues have demonstrated a multi-modal strain sensor with unique capability to separate its direction and magnitude of strain,showing a pronounced high selectivity. The successful multidirectional sensing is attributed to the highly-aligned carbon nanofiber working as conductive fillers in elastomer PDMS matrix,which shows the morphology change and resistance variation under directional loadings.39

Despite the achievements made in multifunctional strain sensors, it should be noted that the external energy is adopted to power the multifunctional strain sensors with great structural complexity and high fabrication cost. Alternatively, the interactive system has now been developed,that is to combine the integration of sensors and interactive interfaces.The developed triboelectric nanogenerator has been proposed to function as a signal generator to power the whole system,followed by a rectifying optical communicator and a signal processing unit,which is capable of functioning as a controlled tactile interactive system.40However, these have not met the requirements of multiple detections in a self-powered single sensor.

A green self-powered concept is exploited to convert the external force into a power instead of using batteries as a power supply. The previously reported flexible and selfpowered tactile sensors are based on piezoelectric or triboelectric mechanisms,32,40,103a novel design about the electromagnetic has been introduced to design magnetoelectric sensors.Magnetic powers instead of magnets are uniformly dispersed into polymeric elastomer to prepare magnetoelectric soft composites, which allows for the anisotropic mechanoelectrical conversion and electromagnetic induction.104Several magnetoelectric type soft composites are assembled into electronic devices, whose tactile sensing capacity are utilized and integrated into the design of a smart parking system. This smart parking system is self-powered by the weight of parking car,showing its capability of searching the parking place and recording the parking time through the yielding change of voltage signals by sensors installed at the front and rear ends of a parking car.By altering the several crucial parameters,such as magnetic powder weight ratio, horizontal area of helix, layer number of helixes, as well as the compressing velocity and strain, the output voltage signal can be well controlled,showing its advance in designing a novel type of soft selfpowered tactile sensors and its capability of being integrated into a smart self-powered sensing system. Fig. 10(a)104shows the magnetoelectric soft composites enabling the anisotropic mechanoelectrical conversion and electromagnetic induction,whose high output voltage is also demonstrated in Fig. 10(b).104

Fig. 9 Current and future challenges arised in multifunctional strain sensors and their related electronics.100 (Copyright 2020,reproduced with permission from John Wiley and Sons).

Another type of green self-powered sensors lies in the full self-powered wireless chipless sensors, where no microelectronic components are utilized for the integrated whole system.Thermoelectric,piezoelectric,electrostatic nanogenerator(output energy density ranging from a few μW.cm-2to about 32 mW.cm-2) as energy source are always composed of multiple energy conversion processes, including mechanical stress to electricity, storage and regulation to electronic devices, where energy loss will be induced at each energy conversion stage.To directly use the converted energy in the form of wireless sensing signals,the unprecedented advance for developing the IoT(wireless sensor networks) is put forward. A magnetic resonance based wireless TENG is developed by integrating a capacitive type TENG with an inductor coil, as well as a synchronized microswitch,where pulsed output sinusoidal voltage signal is wirelessly transmitted through this inductor coil and received by another inductor coil at a certain distance.Furthermore,after experimental testing,this self-powered wireless sensing system shows the capability of sensing up to longdistance of 2.00 m away with energy transmission efficiency of about 73%.52Fig.10(c)52shows the schematic diagram of wireless self-powered sensing system; followed by the equivalents circuit of the wireless transmission system in Fig. 10(d).52In addition, the photograph of practical experimental set-up system is shown in Fig. 10(e),52where capacitive type sensor as one of the components is made by the electrospun PVDF nanofiber,as Fig.10(f)52shows.To evaluate the efficiency of sensing distance, the received signal is tested by varying the coil distance ranging from 0.35 m to 2.00 m, showing its efficacy in long-distance sensing. Therefore, apart from the conventional design of self-powered system with the processes of signal rectifying,energy storing and power regulating,transmission chips and other circuit components being integrated, this designed wireless self-powered sensing system can directly provide the power source and sense in remote or even harsh surroundings.Moreover, as the resonant frequency substantially affects the signal output,it can be used as a specific identification in recognizing this simply developed wireless system.

3.5. Self-powered sensing system

Fig. 10 Green self-powered sensing system.

For the conventional nanogenerators mentioned in Sections 3.1–3.4, the principle of triboelectric nanogenerator in realizing energy accumulation is to rely on the coupling effect of contact electrification and electrostatic induction between two solids or between a solid and a liquid.However,the interaction of two pure liquids offers another insight in designing triboelectric nanogenerator. For this novel liquid to liquid triboelectric nanogenerator, a liquid droplet and a liquid membrane are the crucial components. When a falling liquid droplet (40 μL)passes through the suspended pre-charged liquid membrane, a peak power up to 137.4 nW will be generated.105Fig. 11(a)105shows the green concept design of selfpowered triboelectric nanogenerator through the interaction of liquid to liquid,where liquid to liquid TENG collects energy from the rain droplets in an irrigation system. This liquid to liquid TENG is achieved by a liquid droplet passing through either pre-charged fluorinated ethylene propylene film or grounded liquid membrane. In Fig. 11(b),105a diagram illustrates the droplets carrying positive charges because of the friction with air,showing the novel concept of this liquid to liquid TENG. In addition, multiple generations can be achieved under the effect of a charged liquid membrane in Fig.11(c),105where positive and negative changes are caused by the droplets passing through a suspended charged liquid membrane(Fig. 11(d)105). The output current (in Fig. 11(e)105) has been measured based on liquid droplets passing through two liquid membranes with the setup of Fig.11(c),105where the generated current by one droplet is also revealed. Current and voltage generated by water droplets passing through the varied sites of polarized liquid membrane are also compared, as shown in Fig.11(f)105and(g).105As water droplets can be either positively or negatively charged by passing through liquid membrane, the surface charges of water droplets will be detected by a second membrane.As a result,this liquid to liquid TENG can be used to harvest mechanical energy, whose targets are ranging from raindrops, microfluids to even tiny particles.Eventually,based on the green design of nanogenerators,continuous energy generation can be realized through the interface of liquid to liquid. In addition, as the negligible friction force has been taken to accomplish the running of liquid to liquid nanogenerator,less energy loss will be achieved compared with that of conventional TENG.

Fig. 11 Green self-powered triboelectric nanogenerator through interaction of liquid to liquid.105 (Copyright 2019, reproduced with permission from Nature Publishing Group).

To strengthen the concept of green design in nanogenerator, a renewable, sustainable, and biodegradable material,instead of the severe environmental pollution issues induced by the non-degradable synthetic polymers used in conventional nanogenerators, is expected to realize the self-powered device or integrated system.29,60,73,82,100,106Fig. 12(a)60shows the self-powered sensor employed in athletic big data analytics,where the processing procedures,including chemical treatment and hot pressing are demonstrated in preparing flexible wood TENG. To evaluate the performance of this flexible wood TENG,open circuit and short circuit of wood TENG are conducted in Fig. 12(b),60followed by the comparison made between natural wood and treated wood TENG. In addition,wood TENG is employed in designing both a self-powered point distribution statistical system and a self-powered edge ball judgement system (Fig. 12(c)60and (d)60). It is reported that a novel biodegradable material made of natural wood has been exploited to prepare a high-performance TENG through an effective processing approach.Because of this simple and two-step chemical processing way, natural wood achieved with desired mechanical and triboelectric properties enables the development of a sustainable and eco-friendly selfpowered system,where falling point distribution statistical analysis and edge ball judgement can be conducted.Apart from this green design, a configurated single electrode self-powered sensor is prepared by electrospun PVDF PENG,which is capable of monitoring mechanical stimulus/stress. Moreover, after the electrode is damaged,this sensor can still work normally,which is unlikely to happen in a two-electrode sensor.50

Nowadays, in terms of the automatic driving and face recognition, the instant visual image sensing technique is becoming the crucial component of smart system. However,the key question lies in the rapid conversion of optical image to electric signal.To mimic the artificial intelligence neural network structure in human being brain, it is quite promising to enhance the contrast of image, followed by the reduced noise signal during image capture.

Fig. 12 Self-powered sensor in athletic big data analytics.60 (Copyright 2019, reproduced with permission from Nature Publishing Group).

WSe2,as a kind of two-dimensional material,has been used in image recognition with super efficiency.107,108The results show that image sensing can be configurated into an artificial intelligence network, where sensing and image processing will not experience any delay. Moreover, with the photosensitive semiconductor WSe2,a two-dimensional photoelectricity array can be reconstructed,showing the strong light-substance interaction and distinct optical-electrical properties.108Owing to this novel design, supervision and non-supervision study can be monitored with this array device. Furthermore, this sensor can encrypt and classify the images which are projected onto the chips from the optical way. Fig. 13(a)109shows the computing achieved through a vision sensor to realize intelligent and efficient processing, where signals of conventional AI vision sensor experience signal collecting,converting,followed by amplifying and then acting as inputs to an external Artificial Neural Network (ANN). Moreover, interconnected sensors on a chip possess the functions of collecting signals and working as an ANN to recognize features (Fig. 13(b)109).Ultrafast machine vision with 2D material neural network enables image sensing,64as Fig. 13(c)64shows. On the other hand, nanoscale connections of a memristive neural network are designed for brain-like circuits,110as Fig. 13(d)110shows.

Fig. 13 Computing through a vision sensor to realize intelligent and efficient processing.

After reviewing recent advances of self-powered sensors, it is anticipated that one of the most appropriate self-powered sensors will be determined to complete the design of selfpowered system. As much discussion was performed on the following aspects, including piezoresistive strain sensors,integrated sensors with multimodalities, thermoelectric nanogenerator and piezoelectric sensor, current and future challenges in multifunctional strain sensors and their related electronics, as well as the review of self-powered sensing system,we are expecting to understand much more about the differences of their component, performance and the underneath mechanism.

4. Comparison of sensors for self-powered system

To fully illustrate the self-sustainable sensing system,a variety of sensors aiming to contribute the self-generated power have been compared and analyzed, including the aspects of component design, processing technique/style, performance, and potentials sides to be improved for future applications(Table 1). Piezoelectric sensor is getting more and more attention in terms of the research falling on self-powered electronic devices. Strain sensor and multimodal sensor have been playing a predominant role in shaping the daily life. Moreover,after applying the mechanical stimulus, voltage will be generated without the supply of external power source.Thermoelectric sensor is developed based on the Seebeck effect to detect the temperature variation, while triboelectric sensor shows the unparalleled advantages in acting as the self-powered sensors based on its triboelectrification effect/electrostatic induction. Furthermore, thermoelectric nanogenerator can serve as the self-generating power source for conventional strain sensor to improve its sensing performance and stability. Therefore,triboelectric sensor shows great potential in serving as the self-powered component/unit for human–machine interaction.

5. Self-powered sensors prosper human machine interaction

With the popularity of Internet of things and the explosion of interconnectivity,59sensing multiple stimuli is becoming available in a single designed sensor. Fig. 14,111describes the concept of an integrated self-powered sensor system, where TENG sensor and Optical Wireless Communications (OWC)are playing the individual role of event trigger, power source,transmitter and receiver.52When mechanical stimuli are applied on TENG sensor, a high voltage generates and power the LED array lights. Then both OWC transmitter and receiver function normally under each specific condition.

Piezo-thermoelectric generator offers another feasible option by using the surrounding environmental waste heat and the mechanical motions such as touch, contact and vibration, etc. To obtain an improved voltage output, the enlarged temperature difference and high performance (i.e., high piezoelectric constant or high thermoelectric figure of merits) thermoelectric materials can be used to design a piezothermoelectric generator. This design can also be optimized in conjunction with a developed mechanical-thermal-electric coupling model,79where full parameter energy efficiency will be considered to realize a high energy output.

As the integrated self-powered system makes a sensor capable of multiple detections,112–115it is anticipated that the selfpowered electronics will increase more possibilities of breakthroughs for the human–machine interface and provide more solutions to the upcoming energy shortage and environmental polluted issues.

According to Maxwell’s equation, the principle of TENG has been utterly illustrated (Fig. 15(a)63), where current induced by the surface polarization generates the output signal in TENG and serves the electromagnetic wave to comprise the whole displacement current.

To further improve the performance of devices, the device optimization and its stability, durability should be improved to reach the eventually high sensitivity and good stability. To avoid the complex design, device miniaturization and integration is becoming an inevitable trend to achieve the rapid developing of electronics. With the recent advances of IoT, AI and robotics, the design of human–machine interaction system is expected to possess the merits of multimodalities and perfect interaction.

Textile-based electronics (Fig. 15(b)97and (c)97) with high performance and environmental stability capable of being used in harsh surroundings is in a great demand.97A wearable textile based electronic has been developed to achieve real-time human motion monitoring, human–machine interaction and robot learning in harsh surroundings, demonstrating its great potential in sustainable textile-based wearing sensing system.Moreover, machine-knitted washable sensor array textile electronics are striving to concurrently monitor the arterial pulse waves and respiratory signals,which provides a very promising means to achieve the quantitative analysis of chronic diseases.96

Table 1 Comprehensive comparison of various types of sensors.

Fig. 14 Conceptual configuration of integrated self-powered sensor system.111

6. Challenges and conclusions

Thermoelectric generator is based on the Seebeck effect to convert heat into electricity by using the high-performance thermoelectric materials. Piezoelectric generator harvests energy through the mechanical vibrations. The mechanism of triboelectric nanogenerator is attributed to electrostatic induction and contact electrification. However, thermoelectric conversion efficiency is substantially determined by the figure of merit ZT(ZT=S2σT/κ),where crucial parameters including electrical conductivity σ,working temperature Τ,thermal conductivity κ and Seeback coefficient S are all correlated and affect the eventual efficiency.44Therefore, strategies including nanostructure engineering, band structure engineering and nanomagnetic composition optimizing, etc, are exploited to improve the thermoelectric performance, where both high figure of merit value and wide temperature range are anticipated to be harvested for practical applications. In addition, selfpowered sensors with piezoelectric generator show the advantages of self-powering and thinness, while their disadvantages comprising of complicated fabrication process, low output response and low sensitivity have hindered their wide application in electronic devices.For self-powered sensors with triboelectric nanogenerator, small air gaps generally induce low sensitivity and improved sensitivity are required which can be proceeded by enlarging the air gaps in the designed structure of triboelectric nanogenerator. A battery-free shortrange wireless sensor network has been developed by using direct TENG sensory transmission, where wireless 2D/3D control is achieved by varying the connection of textile TENG itself or adjusting the external capacitor.116

Instead of the conventional single measurement provided by a single unit of sensor, multimodal sensors or multifunctional sensor capable of multiple detections with the assistance of recent rapid advances of robotic technology and Internet of things have unprecedently promoted the development of integrated self-powered system for perfecting the human–machine interaction.For instance,to achieve the early diagnosis of cardiovascular disease frequently occurred in aged people115,117and to accumulate the medical information from selfpowered sensors,118self-powered sensor is a sensor that automatically gives out an electric signal when mechanically activated without an external power source. In addition, the selfpowered sensor is flexible electronics that offers a real-time,prompt and comfortable way to monitor disease conditions instead of conventional approaches (such as magnetic resonance imaging, photoplethysmography, etc.). TENG is proposed to be one of the promising alternatives, which behaves as a crucial component in serving self-powered devices owing to its simple configurated structure,high output power density and low operating frequency (in Fig. 16).

On the other hand, to comprehensively reach the unprecedent conversion from environmental energy to electricity,researchers are spending lifetime endeavour on the scientific exploration,where crucial mechanisms of photovoltaic conversion and electromechanical transduction are proposed to solve the energy conversion issue based on semiconducting junctions and piezoelectric insulators. Recently, halide perovskites were reported to show a photoflexoelectric effect, enabling the simultaneous electromechanical and photovoltaic transduction to harvest energy from a variety of energy inputs.119Halide perovskites, known as hybrid organic–inorganic perovskites,provide the alternative as an energy supplier from the external light or vibration stimuli. Furthermore, self-powered interface by integrating both triboelectric and photovoltaics energy harvesting is endeavoring to realise Internet of things home as well as access control applications.69As a result, current existing electronics with sensing modalities are commonly operated by using energy/power supplied from batteries or near-field communication.53,120However, for the next advanced generation of electronic devices with Internet of things, it is anticipated to be operated wirelessly and self-powered.116For instance,the perspiration-powered electronics by accumulating bioenergy source from human sweat provides a feasible route to realize multiplexed and wireless sensing for human–machine interfaces.121Moreover, the largely ignored micro-energy source generated by water temperature variation, sunlight coming through the house, heat from environment, or even pressure induced by a slight physical movement or vibration,can be exploited to power the sensors.Recently,a type of flexible ionic thermoelectric material demonstrated a giant positive thermopower of 17 mV/K, which was attributed to the synergistic effect of thermodiffusion effect and redox couple for thermogalvanic effect.122After being configurated into wearable devices with 25 unipolar elements (each with the size of 5.0 mm × 5.0 mm × 1.8 mm), the accumulated voltage up to 2.2 V can be generated with a peak power of 5 μW by using the body heat. Therefore, this kind of ionic thermoelectric materials shows great potentail in the conversion of heat to electric energy by using ions as energy carriers.

Fig. 15 Self-powered sensors enable human–machine interaction and high-performance sensors towards human–machine interaction and robot learning.

Fig. 16 TENG enables future human machine interaction.

Subsequently,to achieve the human–machine interface with self-powered sensors,these crucial aspects cannot be neglected:(A)Lower the fabrication cost and make the structures simple,such as designing a single sensor with multi-modality instead of integrating a number of sensors with varied functionality; (B)Search for a promising candidate in serving as the power unit in an integrated system for electronics application, such as the potential alternatives of thermoelectric nanogenerator, triboelectric nanogenerator, and piezoelectric nanogenerator; (C)Make full use of the largely ignored micro-energy induced by human movement, body heat or sunlight shining as powers source.Therefore,to sum it up,we are still facing some certain challenges of achieving the comprehensive self-powered sensors with polymer nanocomposites for human–machine interaction,where aspects ranging from physical design principles,selection of materials,nanocomposite processing,sensor structure design to specific potential applications are covered(Fig.17).

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

This study was supported by the Start-Up Funds for Outstanding Talents in Central South University, China (Nos.202045007 and 202044017)and the Open Sharing Fund for the Large-scale Instruments and Equipments of Central South University, China.