Piezoresistive design for electronic skin: from fundamental to emerging applications

There is growing recognition that the developments in piezoresistive devices from personal healthcare to artificial intelligence, will emerge as de novo translational success in electronic skin. Here, we review the updates with regard to piezoresistive sensors including basic fundamentals, design and fabrication, and device performance. We also discuss the prosperous advances in piezoresistive sensor application, which offer perspectives for future electronic skin.


Introduction
Skin is the largest organ that not only protects our bodies from the environment, but also as a system senses and converts external stimuli into physiological signals 1 . The concept of electronic skin (e-skin) that represents a device to mimic natural skin, is developing explosively 2−4 and in order to be user friendly, it aims to rebuild the skin's sensing response to the vital characteristic, e.g. strain, pressure, temperature, moisture, and body fluid composition, and offer enough flexible, durable and selfhealing properties 5−9 . Historically, in 1985 General Electric developed the first skin-like electronic device to offer the sensing capability for robotic arms ( Fig. 1(a)) 10 . This progress initiates the appearance of e-skin in public vision and causes great research enthusiasm with significant developments, e.g. capabilities of shear force measurement 11 , spatial tactile mapping 12 , connection between e-skin and live neuron 13 , bioinspired and large-area man-ufacturing 14−16 , and self-healing property 17 . Meanwhile, the achievement of conductive composite makes e-skin possible to integrate flexibility with high sensitivity. For example, Lou et al. 18 designed a composite for e-skin, by incorporation of reduced graphene oxide (rGO) in a flexible polymer matrix. Park et al. 19 reported a ferroelectric e-skin with a complex three-dimensional (3D) microstructure, which could detect and amplify signals from temperature, pressure, and vibration change simultaneously. Latest developments on the sensitive response to sound, magnetic field, and ultraviolet light are expected to make e-skin be functional beyond natural skin 20,21 . With real-time monitoring, timely feedback, wireless sensing, and ease to carry, there is a strong expectation that e-skin will display more advanced and convenient features than existing health-monitoring devices. Future e-skin shall be able to sense external stimuli and in turn crosstalk with the human neural network 10,22 .
The different receptors of natural skin confer rich sensations 23 . A tactile sensor that mimics response to skin touch demonstrates exciting potential, for the e-skin application in wearable health-monitoring, advanced robotics, and artificial intelligence. For example, e-skin based on a tactile sensor could sense complex mechanical parameters, e.g. strength, position, and time sequence of the applied force 24 . Differently, e-skin based on temperature and chemical sensors could respond to thermal variation 25,26 , and biofluid 5 , respectively. Presently, the working principle of a tactile sensor can be capacitive, piezoelectric and piezoresistive, which accounts for the sensor with a micro-level resolution to mechanical force whilst reducing manufacturing cost and complexity 24 . However, a capacitive sensor may associate with vulnerable sensitivity and parasitic capacitance that give the unstable performance of e-skin 27,28 . A piezoelectric sensor can have the high sensitivity for precise dynamic stimulus detection, except static mechanics or the one at a magnitude beyond nanoscale 29 . In contrast, the piezoresistive sensor enables the detection of both static and dynamic stimulus at high sensitivity, fast response, a large sensing range and better low frequency response, and is advanced in simple assembling and low-cost fabrication 30 . Current versatile designs of piezoresistive sensor find extensive research for e-skin development 31 , leading to the emergence of a wide range of application, including muscle contraction measurement 31 , gait monitoring 32 , voice recognition 33 , robotic control 34 and human-machine interaction 35 . On the other hand, existing challenging issues toward transformation remain not solved yet and waiting for a de novo solution 36 . Here, we present a systemic review on the updated status of the piezoresistive sensor ( Fig. 1(b)). We attempt to uncover the fundamentals, structure design, manufacturing, and performance for piezoresistive innovation. We also discuss the healthcare relevance of piezoresistive sensor-based e-skin and propose future perspectives for this field. This review has significant insights into designing advanced piezoresistive sensors for future e-skin.

Mechanisms of piezoresistive materials
By definition, the piezoresistive effect represents the change of electrical resistance under mechanical stress and strain 37 . It can be quantified by Gauge Factor (GF), which is equal to a ratio of the relative resistance change to applied mechanical deformation: , where R, ρ and γ are the material's resistance, resistivity and Poisson's ratio, respectively 38 . The GF of most metallic materials is between 1.6 and 2, and under mechanical stimuli, it often varies by geometric change 39 . Differently, upon receiving mechanical stimuli, semiconductors can change their density and mobility of charge carriers, leading to a variation to the material's GF 40 .
Piezoresistive effects from metals and conductive inorganic non-metallic materials

Metallic materials
The electrical resistance of a metal depends on its dimension and resistivity. When subjected to mechanical stimuli, the metal follows Poisson's ratio to adjust its dimensions in length and cross-sectional area whilst giving less change to its resistivity ( Fig. 2(a)). As such, the variation in metallic resistance is a function of geometric change, which forms the fundamentals of the piezoresistive effect 41 . For example, a metallic bent strip at a thickness of few microns has application in different strain gauges 42 .
Such a traditional piezoresistive sensor is advanced in low-cost manufacture and high sensitivity, but remains with challenging issues when integrated into a flexible eskin and applied for the detection in different directions 43 . These limitations related to metallic material could be resolved by manufacturing advances. For instance, wafer-scale microfabrication can allow for embedding metal gauge on polymer chip, thereby enabling in-situ stain monitoring 44 and pressure sensing 45 . Further extension of piezoresistive fundamental by integration between metal and polymer, makes the wider and more convenient application of metal for the flexible device. For example, nanometal as the conductive filler can be dopped into a flexible polymer matrix and generates a piezoresistive effect by establishing percolating networks (Refer to 'Nonconductive polymers'). Different from the direct blending, Ferreira et al. 38 reported deposition of the nanostructured zigzag-like ZnO/Ag thin film on polyethylene terephthalate substrate and assembled them into a flexible and stretchable piezoresistive strain sensor. Alternatively, Peng et al. 46 developed a flexible foam for piezoresistive sensor construction, by a liquid metal coating during polymerisation, and obtained a super electrical conductivity, high elasticity, and phenomenal mechanical stability ( Fig. 3(a)). composite films, which showed a percolation threshold at 28.03 wt% of modified-graphite and a piezoresistive sensitivity of 0.274 kPa -1 at 30 wt% of modified-graphite under uniaxial compression (Fig. 3(c)) 51 . This piezoresistance is believed due to the effect of coupling mechanical deformation and electrical property of a percolating carbon network established within the polymeric matrix 51 . Taken carbon nanotube for example, there exist two types of resistances that account for the strainsensing mechanism, including an intrinsic resistance of carbon nanotube and an intertube resistance. The latter occurs when the gap between adjacent carbon nanotubes is small enough to induce a conductive path at the external electronic field both due to a "tunneling" as well as a "hoping " mechanism. In addition to the direct blending, carbon materials could also be fabricated into a piezoresistive sponge, which is capable of overcoming the problems related to conductive nanomaterial coating, e.g. fragile, easy to peel off, and material aggregation 52 .

Piezoresistive effects from semiconductors
Semiconductive materials such as silicon, polysilicon, and silicon carbide, have inherent piezoresistive property, and due to less correlation to geometric change, their piezoresistivity is distinct from metallic materials. The existing theories are based on the modifications to the energy band structure (not only the band gap but also the band warping and splitting). In this paradigm, the strain in a crystalline solid can change the lattice constants and reduce crystal symmetry, leading to the significant shifts in energy band edges 53 . This bandgap shift and the change in electrons mass under strain lead to the change in conductivity and carrier mobility. In another word, the effect of applied force is to change the number and mobility of charge carriers within a material, thus causing large changes in resistivity ( Fig. 2(b)) 40,42,47 . This results in the piezoresistive semiconductor having very high sensitivity and better low frequency response than its piezoelectric effect. Taken silicon for example, its res- opposite trends in resistance change and different direction-dependent magnitudes. For the maximum sensitivity, n-type silicon piezoresistor must be aligned along [100] direction, and p-type piezoresistor along [110] direction 54 . N-type silicon shows a negative GF of about -135 and the p-type silicon a positive GF up to 175 47 ; (2) Doping level -In the highly-doped silicon piezoresistor, the percentage of carriers moving toward the lower energy levels is smaller than that at a low-impurity concentration. This leads to the low -doped piezoresistor having a high GF 47 ; (3) Crystallographic direction -Most theoretical models have the crystal orientation dependence of band structure, electron energy and the effective masses of carriers. In n-type silicon, the average mobility of carriers increases in the direction of tension (longitudinal effect) and declines transverse to that direction (transverse effect) 41 ; (4) Material's elasticity -The semiconductors such as silicon, germanium and gallium arsenide have typically elastic body. The elasticity influences what degrees of the material's deformation to given stress and in turn determines the number and mobility changes of charge carriers; (5) Scale of materials -In semiconductors, the quantum confinement effects exist when the device is sufficiently thin and have important implications for that material's energy band structure. Piezoresistance coefficient can be increased with the decrease in the scale of material. For example, the carbon nanotube can have a GF of ~2900, and the silicon nanowire has a piezoresistance coefficient reaching up to -3550×10 -11 /Pa (Bulk silicon, -94×10 -11 /Pa) 53 . The optimisation of these factors accounts for high sensitivity and repeatable output for the semiconductor-based piezoresistive sensors 54 . Based on polysilicon, Wee et al. 55 reported a piezoresistive micro-cantilever beam that could detect the labeled disease marker. Similarly, a piezoresistive strain sensor integrated with the siliconbased nanowire was demonstrated for accurate nerveprobe operation 56 . Piezoresistive performance of the silicon-based semiconductor is temperature-dependent, and becomes unstable above 150 °C and deteriorated above 500 °C, which limits the application in the high-temperature harsh environment 57 . This is due to the intrinsic electron and hole mobilities of silicon that decrease with increased temperature, leading to raised resistance additionally 58 . Won et al 59 designed a 3D tactile sensor to enable simultaneous recording of pressure, shear force and bending strain along with temperature. Both temperat-ure and the indued thermal expansion were thought to cause the change of resistance. As such, a careful control on the testing condition, addition of a temperature sensor and/or the use of Wheatstone bridge become necessary to offer extra temperature compensation, with the increases in complicity and processing cost 60 . In contrast, silicon carbide shows a high-temperature piezoresistive effect, due to its wide bandgap, high chemical inertness, and radiation resistance 30,61 . Based on the laserinduced silicon carbide, Nguyen et al 62 developed a piezoresistive pressure sensor, with high sensitivity in a wide range of working temperature (198-473 K). Silicon carbide incorporated into the structure of a porous sponge could enable a piezoresistive sensor for monitoring damage in the electromechanical microdevice at high temperatures 63 . However, the development of semiconductive piezoresistivity for e-skin remains challenging because of the high material rigidity and difficult-processing property. By transfer of silicon nanomembrane to a plastic substrate, Park et al. 64 proposed a solution for the assembly of a flexible piezoresistive sensor with excellent sensitivity (Fig. 3(d)). Differently, Gao et al. 65 applied a laser to write silicon carbide on elastomer substrate directly, thereby offering a more efficient method for the future development of the flexible piezoresistive sensor from a semiconductor.

Piezoresistive effects from polymers
The polymeric material is investigated for piezoresistive application, in order to overcome the undeniable shortcomings related to metal and semiconductor, e.g. rigidity, small strain range, complex and high-cost manufacture 66,67 . The polymer has applicable prospects due to the advantages of a simple readout circuit, small crosstalk, and facile manufacture, and it can be conductive naturally or nonconductive incorporated with conductive composition 68,69 . For the composite, resistance raises from the changes in the conductive path, filler density, and filler contact area upon mechanical deformation 70 .

Conductive polymers
Conductive polymers have piezoresistive properties, and their piezoresistance is controlled by geometric change similar to that of metals 72 . Take poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS) for example, the resistance increases in-phase with mechanical strain, probably due to the long separation between its conductive domains 73 . In contrast, the rearrangement and combination of conductive domains could decrease its resistance out-of-phase with the applied strain. Due to these reasons, piezoresistive sensors based on conductive polymers can have effective sensing performance in a small range of mechanical deformation. As such, conductive polymer is often integrated with other materials, in order to obtain the idea piezoresistivity. For instance, to compensate the Ag nanowire-related permeation network Fan et al. 74 embedded PEDOT:PSS in polydimethylsiloxane (PDMS), and obtained a sandwich-structured piezoresistive sensor which had a reliable response, in both a cyclic stretching-releasing test and a continuous loading test (Fig. 3(e)). However, towards the wide applications in piezoresistive e-skin, conductive polymers still need to overcome challenging issues related to their long-term stability against high temperature, moisture, chemical reagent, and other harsh environments 75 .

Nonconductive polymers
Most polymers are insulating in nature and can be engineered to conductive composite by incorporation with conductive materials, e.g. nanoscale metal, carbon, and conductive polymer. Piezoresistive effect of a conductive composite originates from the resistance change upon receiving external force, because of the mechanical deformation in a polymeric matrix, e.g. compress, stretch, and twisting, and the induced variation in filler distribution, e.g. the distance, contact, and conductive path among fillers 76−78 . Specifically, a percolation theory explains the conductivity of polymer composite: when the concentration of filler portion is below a percolation threshold, the composite with separated conductive fillers remains insulating, and it becomes conductive with a rapid decline in resistance when filler concentration is above the percolation threshold to build conductive path (Fig. 2(c)) 47 . In this theory, the first conductive path is built just beyond the percolation threshold, by the formation of filler contacts and thus a percolating network 48 . The overall conductivity of a composite depends on various parameters, including the size, shape, distance, and distribution of conductive fillers 29 , while the total resistance is related to the resistance of each conductive filler and inter-filler contact resistance 79 .
Presently, conductive composites based on the percolation theory are widely applied for piezoresistive sensor development. The polymer composite by blending with conductive filler can meet the requirements of a piezoresistive e-skin sensor, e.g. flexibility and stretchab-ility. In this paradigm, Wang et al. 80 reported a transparent, large-scale, and highly sensitive e-skin based on polymeric composite, which was comprised of Ecoflex and magnetic rGO@nickel nanowire ( Fig. 3(f)). However, it should be noted that the highly conductive and flexible (or elastic) properties of the composite are often difficult to achieve simultaneously. This often endows a piezoresistive sensor from polymer composite with a high sensitivity but a narrow strain range 81−83 . The dispersion and uniformity of conductive fillers in the polymeric matrix may also influence the formation of the percolation path, and both properties could be improved by the effective modification on the filler surface 84 .
Polymer-based conductive composite can also be obtained by the coating and printing of conductive material on the microstructure surface of a nonconductive polymer directly, providing a high potential for piezoresistive application ( Fig. 2(d)) (Refer to Section Piezoresistive effects from metals and conductive inorganic nonmetallic materials).

Piezoresistive structure designs and fabrication
Piezoresistive materials that possess inherent permeability and tunneling effect can respond directly to mechanical load; however, they alone are found with limitation to offer adequate sensitivity, a large detection range, rapid response, or hysteresis performance. This triggers de novo structure design and fabrication, with increasing interest to improve the sensing performance of a piezoresistive sensor. In the following sections, significant progress on these aspects are discussed, including most used fibre, micro/nano-structure, sponge, and additive manufacturing (AM) architecture.

Surface-structured architecture
Structures that are designed in protrusive, cracking or a combined geometry, can be engineered from nano to micro sizes, and function as a piezoresistive portion for the flexible sensors in e-skin. These structures are capable of upregulating the sensitivity of a piezoresistive sensor greatly.

Protrusive structures
Artificial surface structures, such as dome 85 , pillar 86 , pyramid 87 , sphere 88 , and hierarchical structure 89 , are often employed for designing piezoresistive sensors for improved sensitivity. By principle, the micro/nanostructure that connects to sensor electrodes can enlarge piezoresistive change by geometric variations between the electrodes and poorly contacted active material, even if a small mechanical load was applied 90 . Generally, the different high-precision structures are fabricated by photolithography on flexible substrates, although it does involve complex and high-cost procedures. For example, by photolithography Huang et al. 91 reported an ultrahigh sensitivity of piezoresistive sensor successfully ( Fig.  4(a)). Similarly, Bae et al. 89 developed a piezoresistive sensor, comprising of a hierarchical graphene/PDMS array, and demonstrated desirable linearity and high sensitivity (<13 kPa) to external pressure. The response linearity was due to a design of the hierarchical dome structures, which allowed accordingly for increases in the number and contact area of squeezed protuberances upon receiving larger pressure. In contrast, the dome without additional substructures had no linear response to external pressure change.
On the other hand, lots of research efforts are keen to peruse for reducing manufacture complexity, in order to achieve a high fabrication efficiency, reproductivity, and low-cost procedures to improve the translational potential of piezoresistive sensors. In this paradigm, laser-assisted fabrication may offer a highly translational potential and stable controllability in engineering the structures for a piezoresistive sensor, with the easily modulated sensitivity via a precise control on structure geometry, e.g. ridge and dome at different sizes 92 . For example, the micro-cone structure that could enable great contact-area change even under tiny pressure and deformation was reported on both soft and hard substrates by laser fabrication to achieve high-sensitivity performance 7,93 . Hot embossing is another simple and cost-effective method, capable of developing large-scale surface structures with controlled parameters, e.g. height and diameter of micro-domes 80  CNTs/PDMS composite Park et al. 94 studied the performance of a piezoresistive sensor with different structures, e.g. dome, pyramid and pillar, and reported that the dome structure offered the best force sensitivities for the normal, tensile and bending stresses (Fig. 5). This was attributed to the localised stress concentration that increased the sensor's sensitivity and output range. However, a large pressure could cause the dome structure with a declined sensitivity and a short range of linear response 89,95 . In another study, Jia et al. 96 developed an even simple method by sandpaper to obtain irregular protrusions, which could also have an effective contactarea increase upon external pressure. Compared with regular structure designs, irregular pattern replication from sandpapers could serve as hierarchical microarchitectural and form additional conduction paths, to generate excellent piezoresistance with a broader sensing range and higher sensitivity.
In addition to the numerous designs of artificial structure, it seems a more straightforward way to learn from nature. For example, the animal and plant materials have unique hierarchical structures, and take as a template, the manufacture for engineering piezoresistive architecture will be simplified and can offer tunable elastic modulus, excellent flexibility, and adequate robustness to congruently fit the requirements of e-skin 97 . Inspired by the hierarchical surface of lotus leaf, a multiscale design of piezoresistive sensor based on PDMS obtained a high and stable sensitivity (1.2 kPa -1 ) with a wide range of linearity (0 -25 kPa) ( Fig. 4 98 . In another study, a piezoresistive sensor that replicated the micropapillary of rose petals was to mimic the structures on the human epidermis and dermis and showed an ultra-low detection limit (<2 Pa) to detect multiple life activities ( Fig.  4(c)) 99 . Based on carbon nanotubes and graphene, Jian et al 100 developed a high-performance sensor with a unique piezoresistive design, by a combination between natural leave structure and aligned artificial geometry. Similarly, Jang et al. 101 proposed a new piezoresistive design for pulse detection, of which the active portion was an irregular micropatterned PDMS negatively replicated from shark skin and sealed with PEDOT:PSS films (Fig. 4(c)). This design resulted in a significant improvement in the sensor's sensitivity over regular artificial patterns. Future efforts on bioinspiration maybe toward the overall enhancement of sensor performances, including manufacture, mechanical property, and sensing function for eskin application.

Crack structures
By inspiration from the slit organ in a spider, it is proposed that crack-like structures for the design of piezoresistive sensors, can have unique mechanical compliance due to the effective deformation upon receiving even a small variation in mechanical loading, thereby giving ultrasensitive detection performance. With the absence of applied force, the presence of cracks could reduce the initial current significantly by destruction of conductive paths, and upon loading the cracks tended to coalesce for an enhanced sensing capability 103 . Kang et al. 16 were the pioneers who firstly mimicked a spider's slit organ and fabricated nano-crack junctions for designing piezoresistive sensors. Their findings highlighted the effectiveness of cracks toward an ultrahigh sensitivity (~2000 GF at 2% strain) of the piezoresistive sensor, and opened a window for more investigations on piezoresistive cracks from nano to micro scales. Chen et al. 102 reported a piezoresistive sensor with artificial cracks and suggested a principle for the sensor's high sensitivity due to the cyclic disconnection-reconnection of crack juntions under external strain ( Fig. 4(d)). Meanwhile, in order to facilitate crack fabrication, Yang et al. 104 developed a controlled method for the flexible piezoresistive sensor, by which a pre-stretch of 5% strain on Au coated PDMS resulted in cracks in the conductive coating due to its rigid nature and weak interaction at the interface between metal and polymer. Owing to the cut-through channel cracks, the as-fabricated sensor showed superior GFs at different external strains, e.g. GF=200 for ε<0.5%; GF=1000 for 0.5%<ε<0.7%; and GF>5000 for 0.7%<ε<1%. However, applying a large strain to the sensor was found to cause rapid crack propagation and catastrophic coating fracture, which might limit a piezoresistive sensor based on the crack structure for sensing a large-range strain. In order to overcome these problems, Shi et al. 105 developed a simple and scalable bio-templating method, to introduce microcavities in an Au thin film. This method had an effective control on crack patterns, and enhanced the interfacial adhesion between metal and elastomer, thereby offering a high sensing range up to a tensile strain of ~90% by avoiding crack propagation successfully.

Combined structures
Nowadays, increasing evidence demonstrates that a combination of different structures can endow a piezoresist-ive sensor with better performances than the single microstructure alone. The combination gives not only the sensor multi-sensing functionality, e.g. strain and pressure, but also improvements on the overall sensing performances including sensitivity, linearity, detection range, and response time. For example, Ji et al. 86 developed a piezoresistive sensor based on the hierarchical architecture comprising of mesoscale domes and microscale pillar arrays and showed an improved sensitivity of pressure greatly ( Fig. 6(a)). Learning from the leaves of Epipremnum aureum, Zhao et al. 106 developed a hierarchical architecture that combined macroscale veins and plenty of microscale surface pores ( Fig. 6(b)). Compared to the single-structured piezoresistance, this architecture could enable a higher sensitivity for the piezoresistive sensor, due to resistance change from the addition of pores that upon external loading introduced variations in the contact area and conductive path. Another work that replicated banana leaves explored a different hierarchical architecture, comprising of microaligned mountains with secondary and tertiary ridges, and under pressure could obtain additional microcracks to further improve the sensitivity of the piezoresistive sensor 103 . However, replication from the natural Step I Step II Step III  structure is inconsistent and uncontrollable. Chen et al. 107 reported a hierarchical sensor based on the combination of artificially designed PDMS/Ag microstructures (As sensing layer) and polyimide/Au (PI/Au) rough topography (As interdigital electrode; Fig. 6(c)). Due to more contact touchpoints and significantly declined resistance upon loading, the sensor achieved a higher piezoresistive sensitivity than the devices with either a flat sensing layer or smooth electrode, and thus. Similarly, Sun et al. 108 developed a simple and low-cost method for engineering sensitively flexible piezoresistive sensors, by the employment of a pyramid structure and a double-sided rough structure. Because of limitation in the continuous deformation under a high loading pressure, the piezoresistive sensitivity of a hierarchical sensor was usually lower than the single-structured sensor.
However, at small pressures, it was found that a hierarchical architecture as the sensing portion could give very high sensitivity, and its detecting range at the high sensitivity was much larger than that of the single-structured sensor.

Fibrous architecture
Fibre may be an ideal type of piezoresistive structure due to its unique elasticity for e-skin, with the ultra-low lightweight and high length-to-diameter ratio. The piezoresistive fibre can also conform to the geometric variety of different body parts, e.g. wrist and finger 109 . In the fibre form, a piezoresistive sensor can be easily designed from conductive materials, e.g. polymer, carbon, and carbonaceous composite 110,111 . Different from the rigidity of bulk material, fibrous architecture offers the piezoresistive structure that has not only 3D compliant flexibility and stretchability but also a more compressible deformation than 2D materials along the thickness. A fibrous sensor can have longer durability and higher sensitivity, which is probably attributed to an additional piezoresistivity and the more conductive paths upon receiving stretchable/compressive deformation. As such, it is expected to overcome shortcomings related to current eskin, such as inadequate sensitivity, limited detection stability, less scalability, low mechanical stretchability, poor reproducibility, and wash instability.

Natural origin of fibres
It is easy to fabricate piezoresistive fibres through integration of fibrous insulator with conductive material by immersion 112 , coating 6 , and in-situ growing 113 , where the fibres could be natural and engineering origination. For example, Zhang et al. 114 reported a piezoresistive pressure sensor, through the layer-by-layer stacking of PDMS and anchored cotton fabric. A treatment of carbonisation was used to endow the cotton fabric a high conductivity, whilst maintaining well the fibrous integrity and flexibility. Cao et al. 112 obtained natural fibres by cutting the cotton clothes directly and soaked them in a solution containing rGO and Ag nanowires to obtain conductivity ( Fig. 7(a)). Based on cotton fibres, the piezoresistive sensor upon applying loading stress could have physical contacts between the device upper and lower layers, among which the Ag nanowires as a more efficient and rapid charge channel provided a significant resistance change and high sensitivity for the e-skin. However, from the perspective of long-term use, conductive material falling off from fibres should be preferentially avoided, and the advances in dipping technology would promote low-cost piezoresistive sensors engineering simply and quickly.

Electrospun fibres
Electrospinning is a typical process for fibre manufacturing, during which polymer solution or melt is jet spun under a strong electric field, and can be stretched into different diameters away from the nozzle tip. It can generate desirable fibres for piezoresistive sensors, with advantages of low cost, simple procedures, and large-area fabrication. Bi et al. 66 reported a fibrous piezoresistive sensor from polyimide (PI) fibres that supported the insitu growth of Ag particles post electrospinning. It responded to eternal pressures by changing physical contacts among the Ag-modified PI fibres, and this formed the sensor's piezoresistive foundation. Furthermore, the piezoresistive sensor was able to sense temperature change based on an effect known as electron scattering. By integration with a planar PDMS film, Wang et al. 115 electrospun carbonised silk nanofibres to serve as a sensing portion, and thus avoided the complicated procedures required for common surface micropatterning. By this method, the fibre piezoresistive sensor showed good transparency and promising application for future invisible wearables. Using electrospun fibre as supporting, sensing and packaging layers, Zhao et al. 116 obtained a sandwich-structured piezoresistive sensor, to sensitively detect pressure for human motion and vital sign monitoring. The change in fibrous contact points was proposed as the piezoresistive mechanism for sensing loading and unloading processes. In line with this, fibre separation would give resistance increase; and based on the principle, Wang et al 110 developed a piezoresistive sensor using carbon fibre mats to monitor body bending actions ( Fig. 7(b)). Future research on electrospun fibres as a conductive portion to offer piezoresistance may have breakthroughs in sensing performance if the limitation in fibre-mat thickness by electrospinning could be overcome.
Electrospun fibre can also serve as a sensor dielectric portion and forms piezoresistance by tailoring the electronic resistance from point-to-point, point-to-face and face-to-face contacting changes. Liu et al 117 fabricated a piezoresistive sensor that was composed of a dielectronic layer (Electronic polyvinyl alcohol (PVA) nanowires) sandwiched between ultrathin wrinkled graphene films, obtaining excellent piezoresistive sensitivity. Similarly, using electrospun PVA nanowires as a spacer between wrinkled conductive polypyrrole films, Luo et al 118 assembled a piezoresistive sensor with subtly controlled conductive paths. Beyond a simple sensing mechanism of conductive path change, the electrospun nanofibers as a spacer enabled the multiple changes of effective contact, leading to higher sensitivity and wider response range. Alternatively, electrospun fibre can be explored to couple piezoresistance with other properties, e.g. piezoelectric, triboelectric and thermoresistive effects 119,120 , for integrated multi-functional sensors.

Wet spinning fibres
Wet spinning has advantages of facile, fast, low-cost and upscalable fabrication, and is another method of fibrogenesis that finds wide applications in different piezoresistive sensors. Using multi-walled carbon nanotube as conductive fillers and poly(styrene-butadiene-styrene) as a flexible matrix, Yu et al. 121   wrapped carbon nanocomposite), which could be used directly as a piezoresistive strain sensor (Fig. 7(c)). Furthermore, Tang et al. 123 realised non-monotonic responsibility for integrated strain and pressure sensing, by a coaxial composite fibre from wet spinning. This piezoresistive sensor overcame the shortcoming of typical fibre sensors that sensed only a single force and could be used to detect blending strain when placed on joint position and external pressure when placed on inactive positions. However, it still requires more efforts for wet spinning to advance the piezoresistive manufacture and extended applications for e-skin.

Spongy architecture
The 3D architecture from spongy can be designed for piezoresistive sensors, and compared to solid structures it has been reported with improved flexibility for a largescale sensing application 124 . In principle, the more conductive paths will be formed by the more contact points and areas building temporarily upon the spongy receiving compression; and this leads to material's resistance change and piezoresistance for the detection of transient and static deformations. Piezoresistive spongy sensors can be obtained from conductive materials directly or non-conductive materials in combination with other procedures, e.g. coating, impregnation, blending and carbonisation.

Direct fabrication of conductive sponges
Different materials have been explored for the conductive sponge fabrication, including carbon materials (carbon nanofiber, carbon nanotube, and graphene), conducting polymers, and even metals 52 . For example, Zhong et al. 125 prepared a wrinkled graphene foam by freeze-drying with the zinc-chloride template and assembled a piezoresistive pressure sensor that had effective interfacial contacts for long-cycle stability and outstanding mechanical resilience. By an ice-template freezing method, a different sensor based on hybrid rGO/MXenes aerogels was developed, with a high piezoresistive sensitivity obtained from the synergistic effect between rGO (Large surface) and MXenes (Conductivity; Fig. 8(a)) 68 . However, the low mechanical durability of the sponge might hinder its flexible application in piezoresistive e-skin. To address this problem, Zhai et al 126 combined unidirectional freeze-drying and ultrasonic dispersion to form aligned pores from elastic materials and achieved excellent recoverability and wonderful compressibility for piezoresistive application. Using the directional freese-drying followed by annealing, Zhuo et al. 127 developed a facile fabrication of lamellar aerogel (MXene and cellulose nanocrystals; Fig. 8(b)) that showed extraordinary mechanical performances and ultrahigh linear sensitivity in a broad pressure/strain range. Based on an aerogel comprising carbon nanotubes and cellulose nanofiber layers (Modified with wave-shaped rGO), Peng et al. 128 similarly obtained an ultralight and highly sensitive piezoresistive sensor.

Indirect fabrication of conductive sponges
Generally, a highly conductive material makes sponges have small resistance change and a limited sensitivity, to sense low-magnitude loadings 129 . In order to serve this problem, different post-procedure, e.g. dip coating and sputtering, to a sponge from non-conductive materials, can be a simple, efficient and straightforward solution that receives extensively investigations in recent years, whilst insulated sponges can have commercialised availability and/or be fabricated efficiently by combining a segregated structure and sacrificial method. For example, Ma et al. 130 reported a piezoresistive sensor by dip-coating the commercialised sponge with graphene and carbon nanotubes, where conductive networks were effectively built on the sponge's skeletons (Fig. 8(c)). Alternatively, Pang et al. 131 coated multilayer graphene on Ni foam and after chemical etching of the metallic template, obtained intact graphene porous networks in PDMS substrate that were capable of detecting piezoresistive changes to pressure and strain ( Fig. 8(d)). Using the Nibased sacrificial template followed with mechanical fracture, Zheng et al. 132 developed a more stretchable and flexible piezoresistive sensor by electropolymerization, and successfully introduced micro-cracks in the foam skeleton for multifunctional detections, e.g. pressure and strain. The Ni-based scarification method was developed further by Luo et al. 82 , where graphene grew on PDMS before Ni etching away. Through this development, the foam conductivity and mechanical properties could be regulated by the number of graphene layers and the composition's ratio, respectively. These accounted for improved sensitivity and a linearly sensing region of the piezoresistive sensor. Different from the metal templates, Sengupta et al. 32 replicated an ultralight-weight and highly squeezable foam from sugar cubes and obtained the piezoresistance by infiltrating the foam with conductive multilayered graphene. Similarly, Zhai et al. 133 Zhong F et al. used sugar as the sacrificial template of piezoresistive sensing portion, and enhanced the sensor's detection range by building 3D conductive networks from carbonblack-decorated foam. Translation of piezoresistive sponge sensors, however, is still limited in e-skin, due to the brittle nature of con-ductive materials and a weak adhesion (Easily peeling off) on sponge insulate skeletons 52 . In order to solve this problem, a new dip-coating approach was developed by the alternate deposition of positively charged chitosan and negatively charged MXene on the surface of existing spongy skeletons, where chitosan served as a ''binder" to enhance the interfacial interaction between MXene sheets and spongy skeletons 33 . Another problem that hinders the sponge's translational application, is its inability to sense small mechanical stimuli and thus generate detectable deformation and conductive path change. The introduction of cracks into the sponge skeleton can be a solution, with the possibility of transforming tiny stimuli into electrical signals. For instance, by the generation of pre-external compression cracks, an Ag/cellulose coated sponge could have the capability for simultaneously detecting both large and small strains 76 . Using ion-sputtering Au decoration, Wu et al. 134 demonstrated a channel crack-based gold@polyurethane sponge, which achieved similar detectability to both tiny mechanical variation and large motion (Fig. 8(e)). In another study, using supercritical CO 2 foaming followed by hydrogen bond assembly, plenty of microcracks were prepared on 3D interconnected conductive channels 67 . The as-fabricated piezoresistive sensor showed high GFs (up to 13.27) when less than 5% compression strain was applied, and in the response, the microcracks contributed to the synergistic effect between disconnect-connect transition. In addition to the conductivity issues, sponges from sacrificial templates were also reported with restricted porosity previously, and in order to solve this problem, researchers developed an emulsion method that could generate highly ordered and tunable pores. For example, in a graphene sponge, the porous structure accounted for excellent elasticity and piezoresistive effects (GF: 0.79 -1.46) for potential e-skin application 135 . Furthermore, the poor mechanical durability of insulating sponges might also hinder their real application; and similar to the sponges from conductive materials, this can be solved by directional-freezing to form aligned porous architecture 136 .
AM-assisted piezoresistive architecture AM-assisted design and fabrication can advance the development of piezoresistive sensors, by the rapid prototyping of different material architectures as discussed above, e.g. fibre, protrusion, and porous structure 137 . This method allows for a free and customised design of the piezoresistive architecture and sensor shape for specific e-skin application, even with the 3D geometry required a complex and highly precise fabrication 138 . Presently, AM builds a 3D digital model from scanners or computer-aided design software and decomposes the model into a series of Z-direction discrete layers for lay-er-by-layer printing and 3D construct deposition 139−141 . AM has been used to print piezoresistive sensors with redetermined mechanical performances and functionality, including the structured mold, substrate and sensor body, sensing elements, electrodes and the full sensor. In principle, AM printing of piezoresistive architecture relies on certain inks, e.g. a blending of rigid conductive materials into flexible polymer matrixes 142 . Furthermore, the high viscosity of inks, e.g. milled carbon fiber/silicone rubber ink, might require a system capable of depositing the ink at high pressure and a high speed, e.g. drop-on-demand AM printing ( Fig. 9(a)) 143 . Using a nanocomposite liquid suspension, Kwon et al. 31 AMprinted a multi-axial fibre sensor that showed a directly piezoresistive response to the external compression and tensile forces. In another work, Wang et al. 144 developed a printable ink comprising carbon nanotubes and a polymer matrix, of which the carbon nanotube formed an interconnected conductive network for piezoresistive eskin sensing ( Fig. 9(b)). By a printable ink from the carbon nanotube, Si nanoparticle, and silicone elastomer, Tang et al. 145 reported a piezoresistive AM-printing that could have tunability of micropore and conductive network ( Fig. 9(c)). Assembling the AM-printed sensor on the fingertips of an ultrathin rubber glove, it realised real-time detection of the grabbing force and different objects, to show a high potential capacity of monitoring and feedback control in future robotics. In contrast, other studies chose the more cost-effective carbon black as conductive materials, in order to have potential industrialisation. For example, Wei et al. 146 reported a homogeneous ink comprising carbon black nanoparticles mixed with Ecoflex0030 at certain ratios for AM printing. Similarly, Wang et al. 147 obtained another ink from carbon black nanoparticles, NaCl (As sacrificial template) and thermoplastic urethane elastomer, for the fabrication of a fully AM-printed piezoresistive sensor. Kim et al. 148 assembled carbon composite and a thermoplastic triblockcopolymer elastomer electrostatically and obtained a low-cost composite dough to enable AM-printing of the piezoresistive sensor with excellent structural control, stretchability, and high sensitivity (GF: 72) and linearity (0.94) (Fig. 9(d)).
It is becoming clearer that AM can enable the tunable fabrication of piezoresistive architecture to meet different design requirements, e.g. size, shape, and geometry. Based on a cubic-structured design, Huang et al. 149 AMprinted a piezoresistive fibre sensor by layer stacking compositional ink optimisation, for the facial fabrication of a stretchable 3D piezoresistive sensor. The optimisation of ink compositions was done through multiple independent nozzles, at the control of different parameter combinations. Possibly, more in-depth development may need a guiding standard for AM-assisted piezoresistive sensors. For different application purposes, the optimisation of printing materials and fabrication procedures is necessary. The manufacturing processes of AM-assisted architecture remain unacceptable for the mass production of piezoresistive sensors 151 .

Healthcare applications
A clearer picture is emerging that beyond the strain gauge for structural health monitoring and non-destructive testing, the newly development of piezoresistive sensor obtains extensive applications in the e-skin and healthcare field. In this paradigm, the piezoresistive sensor is used for different purposes, such as health monitoring (heartbeat, pulse, and blood pressure), speech recognition, and artificial robots (motion monitoring and control). Such piezoresistive sensors, together with other detections for chemicals and biological signals, can enable multi-functionality towards the early diagnosis of human health and diseases. The following section will focus mainly on the newly emerging applications of piezoresistive e-skin.

Health monitoring
Signals from respiration, heartbeat, and blood pressure are important physiological indexes of human health. By daily monitoring of the slight variations on these indexes, many diseases can be diagnosed early and treated timely. Meanwhile, macroscopic activities such as daily walking and exercise are also meaningful for healthcare monitoring. For example, gait monitoring in individuals that have sustained pain from Parkinson's disease, stroke, multiple sclerosis, and other neurological conditions, can benefit greatly from daily monitoring on the progression of diseases 32 . Thus, it is becoming accepted that a large detection range of piezoresistive sensors is necessary for e-skin, to monitor physiological indexes in both macroscopic and inferior variations, towards a healthy life, disease prevention, and timely treatment.
Additional characteristics that make piezoresistive e-skin able to have information feedback timelier and more overcome the cumbersome defects of existing medical devices, include real-time monitoring, lightweight, easy to carry, and perfect fitting with human skin. In order to develop a personally wearable health-monitoring system for the prevention of potential injuries and disease-related disorders, Gong et al. 152 reported a graphene/rGO/polystyrene-based piezoresistive sensor, with adjustable strain range and GF that enabled high sensitivity for the precise detections of the body's posterior cervical pulse and different motion modes ( Fig.  10(a)). Guan et al. 153 reported a flexible piezoresistive sensor to detect the physiological signals from the wrist and jugular venous pulse with high sensitivity (1.12 kPa -1 ), low detection limit (20 Pa), and ultrawide range (up to 1.2 MPa), which was attributed to a synergistic effect between the nest-like architecture and carbon black percolation network ( Fig. 10(b)). Ramírez et al. 154 developed a de novo strain sensor using a hybrid architecture from graphene, plasticized PEDOT:PSS, and ultrathin-layered granular palladium, and could detect heartbeat and respiration at the same time, thereby offering the potential to identify a slight difference between the heart rates at breathing and not breathing (Fig.  10(c)). This sensor was for the daily monitoring of obstructive sleep apnea, for the professional medical staff to master the patient's condition and make treatments timely.
Piezoresistive sensors that are designed for the detection of relatively small physiological activities (respiration, heartbeat, and pulse), generally require a very low detection limit and high sensitivity. However, it needs to widen the detection range with proper sensitivity if the sensors aim also to detect large human motions. Tewari et al. 155 developed a MWNT-rGO@PEB piezoresistive strain sensor for the detection of large and small-scale movements of the body. This was achieved by the combination of two different structures, which enabled high sensitivity in a wide detection range and the two linear states at different strain values, e.g. GF 150 for 0 -15% strain, and GF 650 for 28-38% strain. Zheng et al. 132 also reported a piezoresistive sensor with excellent performances of high sensitivity and a wide working strain range, for the large-amplitude activity detection including finger bending and wrist movements (Fig. 10(d)). Differently, the performances were achieved based on microcracks on a spongy structure, where the design of cracks was for detecting small-scale activities, e.g. human pulse, and identifying different mechanical stimuli, e.g. pressure force, shear stress, and torsion force.
Further developments show more interest in the functional integration for health detection, e.g. strain visualisation, alarm and wireless sensing ( Fig. 11(a)). For instance, by combining with haptic displays a piezoresistive sensor could be expected to provide the surgeons with haptic pressure feedback and thus, improve the efficacy of minimal access surgery approaches 156 . In another report, Wang et al. 157 developed a piezoresistive sensor that integrated the mechanochromic photon microtubules from interactively full-colour-changeable electronic fibers (Ecoflex/MWCNTs/Ecoflex). As an indicator, the deformation, e.g. strain from 0 to 200% of the fibres could be recorded accurately by the electrical readout. Simultaneously, the photonic crystals generated the strain feedbacks by instantaneous and reversible colour changes. This integration of multi-functions facilitated a wearable device for real-time monitorisation, by digital signals from the resistance and light amplitude changes, as well as colour changes observed intuitively by human eyes. Piezoresistive sensors have also been investigated for alarm application. Chen et al. 158 developed e-skin from the laser-induced graphene with high sensitivity and integration between detection and alarm function. Design of the piezoresistive effect was for sensing mechanical stimuli, whilst thermoacoustic and excellent acoustic performances were for the detection and alarm function   walking motions, and it showed good histocompatibility in vivo and a high potential to dynamically monitor nerve repair process without scarifying animals as required by traditional biomedical evaluations.

Speech recognition
Presently, speech recognition is used widely in humancomputer interaction, such as the voice-controlled electrical operation and the speech-text conversion in  mobile devices (Fig. 12(a)). Instead of sensing the sound waves directly, most sensors detect the movements of the throat by vocal cord vibration and muscle stretching and contraction. This forms the basis of speech recognition by a highly sensitive piezoresistive sensor pasted at the throat 161 . The opposite application of such voice monitoring by piezoresistive sensors made e-skin promising candidates for potential speech recovery via human−machine interaction 103 . Zhang et al. 162 developed a piezoresistive sensor for voice identification, including different words, vibration, and loudness of the sound (Fig. 12(b)). This was based on the heterogeneously structured MXene for the realtime sensing of low-magnitude vibrations. The force from acoustic vibration was distinct at different volumes, and based on the piezoresistive effect, it gave electrical signal changes to detect sound loudness. Based on the piezoresistive design, Yu et al. 163 AM-printed a thinner voice device (2 mm in thickness) and attached it to the throat for detecting vocalizations such as tester's speak and cough. The speech induced throat vibration and in turn resistance change in the sensor. At different loudness on intonations or coughs, the same word accordingly gave differences in the peak magnitudes of curves whilst retaining the similar waveforms. This suggested that the waveform could be a recognisable characteristic of a specific word, while the wave details as the recognisation of different person's voices, leading to prospects promising for voice recognition in future e-skin.
However, the detection of timbre requires high sensitivity and short response time, and the natural frequency of most materials cannot keep up with the vibration of sound waves. Deng et al. 164 achieved timbre recognition by the detection of sound frequency, via an ultra-sensitive piezoresistive sensor comprising of vertical graphene with a high density of cracks at different directions. Furthermore, by the piezoresistive design, Dinh et al. 165 reported a skin-attachable acoustic sensor to avoid background noise interference (Fig. 12(c)). The superior sensitivity (GF=8699) of the sensor allowed for the direct capture of small acoustic vibrations, leading to the elimination of background noises. In fact, the piezoresistive sensor to directly feel sound waves requires even high sensitivity and a very low detection limit. Based on the combination of microscale patterns and cracks, Nie et al. 103 reported a piezoresistive sensor with high sensitivity, which enabled the detection of speech-induced air burst and thus allowed for noncontact speech recognition (e.g. the pronunciation of different single letters). Gao et al. 166 designed the microchannel-confined MXene for engineering a piezoresistive sensor for microforce detection. The sensor to feel sound waves was by the piezoresistive response to pulse forces generated on the sensor's surface, and thus it was able to distinguish different ringing tones at high accuracy in the source waveform outline. Based on MXene@chitosan@polyurethane composition, Li et al. 33 reported a non-contact strain sensor to sense knocking sound quickly (response time: ~19 ms). When a word of 'carbon' was spoken, the sensor was able to identify the two characteristic peaks corresponding to the roots of 'car' and 'bon', respectively. This indicates the potential of the piezoresistive sensor for recognising word pronunciation. However, in the field of speech recognition, there still lacks a standard on piezoresistive design for eskin. For example, it remains unclear what kinds of signal change corresponding to the specific word. Probably, the human eardrum offers us an opportunity to fully understand how the piezoresistive sensor should be engineered. Future research to covert the signals from key voice characteristics in a standard manner will make piezoresistive e-skin more applicable for speech recognition.
Prosthetic and artificial robots Robotic prosthetics and hands require a tactile ability for comfortable and more desirable operations. Under the control of a motion sensor and electrotactile stimulation, robot hands could be programmed for grasping actions. However, as the substitute of human hands, existing robots were found with issues such as inaccurate capture, the poor control of strength, and a limited diversity of movement 167 . Recently, researchers have developed piezoresistive sensors to help robots obtain capabilities of tactile feeling and static distance detection, to sense different objects, textures, and positions in order to avoid a collision for better safety and compliance. For example, Dong et al. 168 integrated a piezoresistive strain sensor into intelligent gloves and by signal recognition and feedback, realised movement control on a manipulator. In another study, Liu et al. 169 manipulated robot arms playing music by gestures, where successful human-computer interaction was built between the robotic arms and a piezoresistive sensor ( Fig. 13(a)). Xu et al. 83 designed a competitive robot that integrated five independent sensors in the rubber gloves for intelligent human-computer interaction, and the piezoresistive sensors made input characters be highly recognisable. With soft porous and elastic piezoresistive sensors, Sencadas et al. 170 developed a gripper to grasp objects with different shapes and textures. The gripper received signal changes when clamping and releasing objects, and matched closely the sensing characteristics of human fingertips based on the compliant contact detection between fingertips and objects.
Large-area detection of pressure and pressure distribution is another very important robotic capability. The strategies that sense surface pressure by the integration of multi-sensor arrays become challenging, due to the difficulties to apply sensor arrays on a large-scale surface, e.g. plenty of sensor units, big-task signal processing, and lots of connecting wires 166 . Fibres fabricated from conductive poly(vinyl alcohol-co-ethylene) were reported with good piezoresistive properties, and a sensor based on the fibrous mats could sense stress change upon irregular touch, map spatial pressure distribution, and display the contacting object shape (Fig. 13(b)) 171 . Liang et al. 172 integrated the e-skin with a piezoresistive sensor and time-of-flight sensor for nursing robot application. The piezoresistive sensor could help the robot to exam object weight and adjust the arm to avoid improper posture (incorrect position) or mechanical failure (excessive weight) for possible slipping. In a different manner, a piezoresistive pressure sensor was applied to the surgical robots for ensuring surgical safety (Fig. 13(c)) 173 . This sensor made surgeons avoid unnecessary surgical injury by improving the awareness of robot-tissue collisions.
Prosthetic limbs are mostly used to recover the selfcare abilities of disabled patients; however, their longterm adaptation tends to cause high possibilities of harm to the patients 33 . Piezoresistive tactile sensors that are integrated with the prosthetic limbs can increase the ownership of those patients to feel the external objects and obstacles, increase their natural feeling, and thus adapt better to the existence of prosthetic limbs 23 . Specifically, messages from the sensor stimulate the residual sensory pathways and in turn, alleviates the prevalent phantom limb pain. The piezoresistive feedback from a prosthetic limb not only promotes the patient's ownership but also allows for the more natural and facile operation through restoring information about their body position and grip forces. Ferreira et al. 174 fabricated an array of piezoresistive sensor units and connected them to a data acquisition system that had a wireless data transmission capability, to measure the interface pressure and pressure distribution of prosthetic alveolar. With help from the sensors, it became possible to analyse pressure distribution during the socket production process, thereby improving the performances of prosthetic alignment and training with comfort feeling for the patients. Similarly, Kim et al. 175 Fig. 13(d)), and realised the connection to biological efferent nerves and showed the information flow to actuate the tibial extensor muscle in the leg.

Concluding remarks and future prospects
Rapid developments on piezoresistive sensors promote e-skin for prosperous applications, with verified functionality towards high sensitivity, fast response and large detection range. As what we reviewed here, the advances in structural design beyond material composition of a piezoresistive sensor may open a promising window for e-skin, thereby becoming one preferred choice for sensor development. Increasing evidences have suggested that a hierarchical design can give often optimal piezoresistive performance over single structure. However, the fabrication with multiple and complex processes faces obstacles towards a facile integration of piezoresistive structure at different dimensions and materials. Meanwhile, future piezoresistive sensors request flexibly ultrathin and ultralight sensing structure. Luckily, additive manufacturing that gets rapid development in recent years may offer possible solutions for advanced piezoresistive sensors. Next-generation of the piezoresistive sensors should further upgrade without a doubt at the functional performances. Future work along this research direction requires solutions to address inevitable problems including (a) to pursue high sensitivity whilst avoiding interference noise, (b) to pursue long-term performance against any damage and (c) to pursue wireless signal transmission and power supply. Presently, the piezoresistive sensor finds a wide application in various wearable devices. However, its portability is limited greatly due to the leads and external power supply. The recent developments on triboelectric nano generators offer a possible solution 176 , which can allow for a functional integration with the piezoresistive sensor in wearable devices. Meanwhile, the wireless self-power and signal transmission can be another feasible solution 177,178 . This can enable the piezoresistive sensor with portable sensing performance within an available distance. Another possible research direction for future work is bioadaptability which shall enlarge translational feasibility of a piezoresistive sensor. As reviewed in this work, the metals and semiconductors studied extensively for traditional strain sensors require the additional flexible designs for e-skin application. Fundamentals such as percolation theory and tunneling effect can guide the de-velopment of different conductive composites at flexibility. The presence of conductive polymer composites represents one of the great breakthroughs in the development of flexible piezoresistive sensors. Unfortunately, increasing reports deliver concerns on the anaerobic infection and inflammation due to a long-term interferential contact between the polymer and skin. However, challenging issues remain not yet solved completely. In this paradigm, in order to be suitable for the applications in e-skin and healthcare field, future work needs to solve problems including but not limited to (a) the mechanical compliance to that of tissue and organ, (b) the antibacterial and biological safety for long-term contact with human skin and (c) the biodegradability for an implant device to monitor tissue repair progress and self-elimination away after service life. In addition to the functional performance and bioadaptability, an environmental friendship should be considered for piezoresistive sensors. The future work on e-skin calls for the green principles for designing the piezoresistive sensors, which shall consider both fabrication and material selection.
Presently, the applications of e-skin are developing explosively. The attempts based on piezoresistive design, for example, health monitoring, speech recognition, robotic device, human-machine interface, and artificial intelligence, are shaping our society toward an intelligent life. In this field, future interests can include a comprehensive biomimicking of skin functionality beyond just tactile response. This will request a piezoresistive sensor to integrate different modules and capable of detecting various basic parameters such as temperature, humidity, sweat and even metabolism. Of course, difficulties related to the integrated design of complicated circuit and the extreme precision of fabrication should be evaluated prior to the pursuit of excellent multi-responses. As discussed in this review, it is also necessary to have the feedback capability for intelligent application. Definitely, techniques that can promote an integrated connection among piezoresistive sensor, different nerves and smart equipment will advance the development of future e-skin.