Anisotropic Tactile Sensors: Constructive Designs, Challenges, and Emerging Applications

Abstract

Recent advancements in human–machine interaction technologies have driven significant interest in tactile sensors for health monitoring, movement detection, and the progression of intelligent robotics. However, most existing sensors rely on isotropic materials or structures, limiting their ability to detect stimuli from multiple directions simultaneously, which can be efficiently mitigated by incorporating anisotropic architectures. Despite their promising potential, the development of anisotropic tactile sensors remains nascent and necessitates more comprehensive synthesis and generalization of the current state. This review offers a thorough analysis of anisotropic tactile sensors, delving into their sensing mechanisms, performance metrics, materials, and structural designs. It also explores their applications in intelligent systems and critically evaluates the current developmental status and outlines the challenges to be addressed, providing essential insights and innovative solutions to propel advancements in this emerging research area.

1. Introduction

The term “tactile” encapsulates various mechanical stimuli such as contact, sliding, and pressure that are sensed by mechanoreceptors beneath the skin and subsequently relayed to the brain. , Human skin, which houses a diverse array of mechanoreceptors including two slow-adapting (SA-I, SA-II) and two fast-adapting (FA-I, FA-II) types, facilitates the concurrent detection of static pressure and dynamic vibrations across different frequencies (Figure ). − For example, Merkel discs (SA-I) play a key role in perceiving mechanical pressure and low-frequency vibrations arising from alterations in shape and texture. Skin’s tactile perception enables individuals to sense and react to environmental stimuli present in their environment, aiding in the recognition of pressure, temperature, motion, and texture for prompt object identification. − Electronic skin (e-skin) mirrors the functionalities of human skin and empowers robots to categorize objects and respond to external stimuli, , playing a crucial role in diverse applications spanning health monitoring, − motion tracking, − and human–machine interaction. − Advances in artificial skin and wearable technologies have catalyzed the development of tactile sensors, crucial components of electronic skin systems. While these sensors effectively monitor surface characteristics and detect object movement amplitudes, capturing detailed shape information and movement directions, especially in complex three-dimensional motions, remains a formidable challenge.

1.

Distribution of mechanoreceptors in biological skin and the functions that the skin can perform. The Meissner corpuscle (FA-I) perceives light pressure stimuli on the skin and discriminate the distance between two contact points. The Pacinian corpuscle (FA-II) responds to small pressure, vibration, and tension. Merkel discs (SA-I) detect light tactile stimuli, while Ruffini corpuscles (SA-II) are particularly sensitive to pulling stimuli and also detect flexion and extension of joints. , Additionally, SA-I mechanoreceptors measure the softness of objects during pinching. When the hand slides over an object’s surface, both FA-I and FA-II differentiate its roughness and texture. Moreover, SA-II receptors can detect object stickiness during adhesive contact.

To realize multidimensional signal detection, a prevalent approach in research involves the deployment of sensor arrays. These arrays produce unique signals in response to diverse stimuli, facilitating real-time sensing and differentiation of the intensity, location, and pattern of various external stimuli like soft sliding, touching, and bending. While multisensor systems have demonstrated effectiveness, the increasing complexity of target systems necessitates scaling up the number and size of tactile sensors, leading to escalated costs and heightened system intricacy. , Among these, most tactile sensors have an isotropic conductive network, where the distribution of conductive elements is random. These isotropic sensors are typically limited to detecting stimuli in a single direction, resulting in a unidirectional response capability. Additionally, due to their mechanical isotropy, when a tactile sensor is simultaneously stretched or compressed, significant deformations occur in other directions, making it difficult to distinguish between the direction of strain and the applied force. In contrast, anisotropic tactile sensors possess the capability to capture three-dimensional information, enabling the detection of pressure in multiple dimensions and the direction of object motion simultaneously. − The use of a single anisotropic sensor not only enables efficient discrimination of external stimuli but also determines their direction, thereby addressing issues related to the superposition effects encountered in multiple sensor setups. , In anisotropic tactile sensors, the normal force aids in the perception of contact and object shape, while the tangential force provides information about dynamic interactions, including surface texture recognition, object slip detection, and tool manipulation. −

The advantage of anisotropic tactile sensors over isotropic sensors lies in the enhanced electrical and mechanical performance in specific directions. , By designing anisotropic signal output capability, such as by orienting the arrangement of conductive fillers or adding conductive pathways in particular directions, distinct conductivity gradients can be established. When external stimuli cause the sensor to deform, the electrical signals in different directions will exhibit significant differences, allowing for the discrimination of both the magnitude and direction of the applied force. Additionally, anisotropic structural design, such as prestretching or directional freezing, can create a directional internal structure that enhances the sensor’s mechanical performance in specific directions. , When external forces deform the sensor, strain in one direction will not cause significant deformation in another, enabling determination of both force magnitude and direction. In most cases, both approaches enhance mechanical properties and directional signal variations. Over the past decade, anisotropic tactile sensors have been reported, including unidirectional pressure/strain-sensitive anisotropic tactile sensors and multidirectional-sensitive anisotropic tactile sensors. − These sensors show significant differences in sensitivity in different axes or planes. When unidirectionally stretched or compressed, these sensors exhibit minimal deformation in orthogonal directions, effectively reducing or eliminating signal crosstalk in sensor arrays. Furthermore, anisotropic tactile sensors incorporating surface microstructures or oriented internal conductive fillers have been designed to detect object weight, sliding direction, and texture. ,− Additionally, these sensors optimize specific physical attributes such as sensitivity, linearity, and detection range, particularly in specific orientations. − These enhancements arise from the presence of numerous oriented internal pores, which offer additional conductive pathways. While prior reviews have focused on categorizing sensor types or discussing isotropic sensors for unidirectional detection, comprehensive summaries of anisotropic sensor structures remain scarce. , Given advances in sensor materials, it is imperative to review operational principles, materials, developmental status, and future outlooks pertaining to multidimensional signal detection.

This review provides an overview of the current development status of anisotropic tactile sensors, covering their operational principles, performance assessment methods, materials, structural designs, and applications (Figure ). Herein, it first explains the principles of anisotropic tactile sensors with different sensing mechanisms, followed by an exploration of performance characterization methods. Subsequent sections analyze the types of substrates and conductive materials employed, elaborating on design approaches for these sensors, including integrated arrays, oriented structures, and microstructured sensors. Furthermore, the paper highlights innovative applications of anisotropic tactile sensors. Due to limited reporting on anisotropic structures in tactile sensors, it also introduces wearable smart sensors with anisotropic properties. Finally, the review summarizes future trajectories and challenges for anisotropic tactile sensors, aiming to inform and propel advancements in the field.

2.

Structure of this review. (i) Classification of the tactile sensors: piezoresistive, piezoelectric, capacitive, triboelectric. (ii) Component materials of tactile sensors: polymers, metals, and inorganic nonmetallic materials. (iii) Six key performances of flexible sensors: tensile stress–strain curves, sensitivity, linearity, drift, response/recovery time, and stability. (iv) Structure of anisotropic tactile sensors: multiple sensors. Reproduced with permission from ref . Copyright 2020, Royal Society of Chemistry. Oriented structure and microstructure. (v) Application of anisotropic tactile sensors: athletic training, reproduced with permission from ref . Copyright 2024, Wiley-VCH. Health surveillance. Reproduced with permission from ref . Copyright 2020, Wiley-VCH. Object identification. Reproduced with permission from ref . Copyright 2023, Wiley-VCH. Human–machine interactions. Reproduced with permission from ref . Copyright 2020, Wiley-VCH.

2. Classification of Tactile Sensors

Tactile sensors determine the magnitude and direction of applied stimuli by converting stimuli (pressure, tension, etc.) into various types of signals. The operational principles of anisotropic and isotropic tactile sensors are fundamentally similar, as both convert stimuli into signals through resistive, capacitive, piezoelectric, triboelectric, or similar mechanisms. The primary difference lies in the capability of anisotropic sensors to simultaneously track forces and directional movements of these forces across multiple dimensions. This section offers a succinct summary of the mechanisms and structural configurations of various anisotropic tactile sensors.

2.1. Piezoresistive Sensors

Piezoresistive sensors function by detecting changes in resistance in response to external pressure or tension. Typically, these sensors consist of a sensitive material layer sandwiched between two electrodes or positioned on a finger electrode, as illustrated in Figure a. The piezoresistive effect is quantified by the following equation: ,

$$\frac{\mathrm{\Delta }R}{R_0}=\frac{R-R_0}{R_0}=\frac{I-I_0}{I_0}$$ 1

where R is the real-time resistance, ΔR is the change in resistance after applying pressure, R ₀ is the initial resistance, I is the real-time current, and I ₀ is the initial current. The operation of piezoresistive tactile sensors is governed by two primary mechanisms. The first mechanism involves modifications in the semiconductor material’s energy band structure upon external deformation, leading to a shift in resistivity. For example, in P-type semiconductor silicon, external force alters the energy band structure, reducing conductivity. The second mechanism relies on changes in interfacial contact resistance. Pressure on the sensor causes the conductive materials in the porous matrix to make contact, increasing conductive pathways and decreasing resistance. Conversely, stretching the conductive material leads to microcrack formation, disrupting the pathways and increasing resistance. Anisotropic piezoresistive sensors typically employ two approaches: the directional alignment of internal conductive substances (such as fibers and nanosheets) or the design of anisotropic structures. , The oriented arrangement of these conductive components enhances electrical conductivity along the parallel direction, resulting in a significant resistance change during deformation compared to the perpendicular direction. By engineering the substrate material with an anisotropic structure, significant differences in resistance change rates across different directions can be achieved under identical deformation. For instance, Ye et al. introduced an anisotropic aerogel tactile sensor with exceptional compressive strength, strain, shape recovery, and fatigue resistance. Illustrated in Figure a, under compression, internal conductive materials form additional pathways, reducing resistance. Although most developed tactile sensors are piezoresistive owing to high sensitivity, structural simplicity, and rapid fabrication, their relatively high energy consumption limits broader usage.

3.

Schematic diagrams of major sensors based on different sensing principles: (a) piezoresistive, (b) piezoelectric, (c) capacitive, and (d) triboelectric. (e) Comparison of the advantages and disadvantages of four types of sensors.

4.

Different types of tactile sensors: (a) piezoresistive sensor with a directional pore structure that exhibits rapid resilience during compression and release. Reproduced with permission from ref . Copyright 2024, Elsevier. (b) Piezoelectric sensors combining two piezoelectric effects can exhibit anisotropic signal output capabilities when compressed. Reproduced with permission from ref . Copyright 2019, American Chemical Society. (c) Schematic diagram of an anisotropic capacitive sensor for detecting the direction of tangential force. Reproduced with permission from ref . Copyright 2021, American Chemical Society. (d) Schematic diagram of the contact–separation working principle of a triboelectric sensor with a fingerprint microstructure. Reproduced with permission from ref . Copyright 2022, Elsevier. (e) Schematic diagram of a magnetic sensor with multidirectional force sensing capabilities. The top layer is the contact layer, the middle layer is the partition, and the bottom layer is the 3D Hall sensor. Reproduced with permission from ref . Copyright 2024, Wiley-VCH. (f) Schematic diagram of color changes in structurally colored cellulose liquid crystal hydrogels during compression. Reproduced with permission from ref . Copyright 2020, National Academy of Sciences. (g) Structural diagram of VBTS developed based on hydrogel. The sensor carries a camera module, light source module, and contact module. Reproduced with permission from ref . Copyright 2024, Wiley-VCH. (h) Kind of mechanical deformations employed during device testing as OFET-based mechanical sensors. Reproduced with permission from ref . Copyright 2024, Wiley-VCH.

2.2. Piezoelectric Sensors

Piezoelectric flexible tactile sensors identify pressure in a specific direction using the piezoelectric effect. When external force is applied to a piezoelectric material in a specific direction, its internal crystal structure deforms noncentrosymmetrically, resulting in polarization and generation of positive and negative charges, as demonstrated in Figure b. These minor charges are captured by an external circuit and converted into a current or voltage signal after amplification. When pressure is removed, the piezoelectric material’s internal potential returns to its equilibrium. These sensors typically consist of two parallel electrodes with an enclosed piezoelectric layer, which are categorized as organic or inorganic materials. A common organic material is PVDF, while typical inorganic choices include ZnO and BaTiO₃. The two main design methods for anisotropic piezoelectric tactile sensors are piezoelectric material orientation design and microstructure design. , Directional alignment of piezoelectric materials (e.g., polyvinylidene fluoride and barium titanate) to enhance crystallinity and polarization uniformity along the target axis enhances piezoelectric output performance in that direction. Additionally, constructing microstructures on the surfaces of piezoelectric layers and electrodes can selectively amplify strain sensitivity in a specific direction by enabling preferential stress transmission, thereby creating directional disparities in mechanical-to-electrical conversion responses. For example, Milani et al. introduced a BaTiO₃-based tactile sensor (Figure b), demonstrating anisotropic behavior under low-frequency (0.1–1 Hz) compressive stress (5 kPa). The sensor determines the sample’s orientation concerning the stress direction through analysis of the output signal’s phase and amplitude modulation. Piezoelectric tactile sensors provide self-powered functionality, are more energy-efficient than piezoresistive sensors, and excel at vibration detection due to their capability to instantly produce high potentials. However, they have limitations in detecting static deformations due to their high internal resistance.

2.3. Capacitive Sensors

A capacitive sensor comprises two parallel electrodes separated by a dielectric material. As illustrated in Figure c, capacitive sensors function by detecting changes in the capacitor’s structure when a voltage is applied. This adjustment induces opposite charges on the electrodes, resulting in the determination of capacitance. The capacitive effect is described by the formula for capacitance:

$$C=\frac{{\epsilon }_0{\epsilon }_{\mathrm{r}}A}{d}$$ 2

where ε₀ is the vacuum permittivity, ε_r is the relative permittivity, d is the distance between the parallel plates (the thickness of the dielectric material), and A is the effective area of electrode overlap. The formula underscores the significant impact of both A and d on capacitance. Specifically, A is valuable for detecting shear forces and strain, while d is instrumental in detecting normal forces and strain. These distinct sensitivities to varied parameters enable capacitive sensors to operate as anisotropic sensors. In capacitive tactile sensor design, graphene , and carbon nanotubes , are commonly employed as conductive materials. For anisotropic design, the focus is on structural orientation. By designing oriented structures, it is possible to facilitate deformation of the sensor in a specific direction, thus obtaining a more pronounced change in the capacitive signal. For instance, Chai et al. introduced a shear-sensitive sensor sensitive to tangential forces, utilizing carbon nanotubes (CNTs) as conductive fillers. As shown in Figure c, the capacitive sensor has a different structure at the top and bottom, so that the sensor signals at the top and bottom are significantly different after being subjected to shear forces in different directions. Benefiting from this difference, the sensor can recognize the direction of the shear force. Capacitive sensors exhibit exceptional performance in detecting multiaxis deformations with high sensitivity, cyclic stability, and robustness. They also offer the advantage of being energy-efficient. However, the measurement accuracy may be compromised by the viscoelastic nature of the elastomeric substrate. Moreover, while isotropic capacitive sensors are prevalent, anisotropic tactile sensors based on capacitive sensing are a limited development.

2.4. Triboelectric Sensors

The inaugural flexible sensor based on triboelectric principles was introduced in 2012. Owing to their self-powered operation and longevity, triboelectric nanogenerators (TENGs) are extensively employed in wearable sensor design. − As illustrated in Figure d, the operational concept of TENGs involves generating equal and opposite charges on two different materials upon contact, which is influenced by their surface characteristics. Through electrostatic induction, the accumulated opposite charges on the rear electrode produce a potential difference during cyclic motion, inducing electron flow between the electrodes and generating a current output. TENGs operate through four principal modes: vertical contact–separation (C–S), lateral sliding (LS), single-electrode (SE), and freestanding triboelectric layer (FT). − In the domain of anisotropic tactile sensor development, the contact–separation (C–S) mode, depicted in Figure d, prevails. This mode operates on the principle that external force induces relative movement between the textured substrate and the triboelectric layer. Electrostatic induction causes charge accumulation near the contact area, establishing an electric field with one material surface positively charged and the other negatively charged. Continuous external force application results in repetitive contact and separation between the substrate and the triboelectric layer, generating a sustained electrical signal. For triboelectric sensors, the anisotropic structure is usually designed to improve the electrical signal output capability in a specific direction. However, triboelectric sensors with anisotropic structures struggle to distinguish forces in different directions due to the contact–separation mechanism. To distinguish between normal and tangential forces, a microstructure design on the sensor surface is usually required. For example, Qu et al. utilized eutectic gallium–indium (EGaIn) liquid metal and silicon to craft a flexible, fingerprint-like triboelectric tactile sensor (FTTS). As demonstrated in Figure d, this sensor operates in both contact–separation and stretching modes, serving diverse applications. With a microstructural design and triboelectric principle, the FTTS boasts a rapid 1.01 ms response time and a low detection threshold (detecting forces as light as 16.88 mg). It offers 225% stretchability, facilitating pressure testing and material identification. While triboelectric tactile sensors present high sensitivity, broad detection range, and self-powering advantage, they may exhibit compromised stability and durability under prolonged repetitive friction.

2.5. Other Sensors

2.5.1. Magnetic Tactile Sensor

The magnetic tactile sensor fundamentally operates by employing a blend of magnetic materials and flexible substrate as the magnetic field source. Upon applying an external force to the sensor, the mixture experiences deformation or displacement. The change in the magnetic field is detected by adjacent magnetoceptive sensors, determining the magnitude and direction of the applied force. By structuring the magnetic sensor with a biomimetic design, it can acquire the ability to detect forces in multiple directions. For example, Zhao et al. drawing inspiration from human skin and fish lateral lines, created a detachable magnetic soft tactile sensor enabling wireless three-dimensional force sensing through a centripetal magnetization arrangement and a theoretical decoupling model (Figure e). This sensor facilitates precise robotic manipulation and flow-based navigation, with an accuracy deviation of under 1.03%. Magnetic tactile sensors are recognized for their high resolution, mechanical durability, straightforward fabrication, and simple structure. Nonetheless, they remain susceptible to noise interference.

2.5.2. Optoelectronic Tactile Sensor

Another type of sensor is the anisotropic optoelectronic tactile sensor, which relies on the structural light variations in oriented materials when the materials are stretched to measure tensile or normal forces. These sensors can also be made conductive by incorporating electrically conductive materials, rendering them electrically conductive, thereby enabling force magnitude detection via electrical signals. For example, Zhao et al. created a color-changing optoelectronic tactile sensor. By incorporating hydroxypropyl cellulose (HPC) in the hydrogel to form cholesteric liquid crystal photonic structures, the hydrogel exhibits vibrant structural colors responsive to external stimuli such as temperature, pressure, and tension changes, which are represented as optical signals (Figure f). Although these sensors provide the most intuitive signal output, they are difficult to integrate and to ensure uniformity of color change during stretching.

2.5.3. Vision-Based Tactile Sensor

The vision-based tactile sensor (VBTS) can sensitively sense normal and tangential forces, which is very close to the tactile information dimension of the human hand. The VBTS interprets tactile information through visual perception mechanisms (e.g., optical imaging, image analysis), and its unique perception principles (e.g., light reflection, deformation texture capture) give it a natural advantage in distinguishing between tangential (shear) and normal (pressure) forces. VBTS typically consists of an elastomer (e.g., silicone, hydrogel) and a vision acquisition system (camera, light source). When an external force is applied, the elastomer undergoes deformation and its surface texture, transmittance, or color changes are captured by optical imaging and interpreted as a tactile signal by image analysis. In detail, the main structure of the VBTS consists of a contact module, an illumination module, a camera module, and an information processing module. The contact module contains a contact body, a reflective layer, a protective layer, and a marking layer. When the reflective layer is deformed at the contact surface, the deformed area causes an irregular offset of the reflection path. Normal forces (pressure) mainly cause vertical deformation of the elastomer (e.g., thickness compression, surface depression), resulting in localized changes in light intensity or texture contrast. Tangential forces (shear) mainly cause lateral slip or shear deformation of the elastomer, resulting in directional distortion and/or displacement of the surface texture. When light changes caused by this irregular offset are collected by the camera module, the tactile type and size are obtained through information processing. For example, Liu et al. developed a vision-based tactile sensor based on a mechanically responsive color-changing hydrogel (Figure g). When the hydrogel touches a material surface, the material’s surface causes the hydrogel structure to change, producing color variations. These color changes are captured by the camera module to sense the hardness, shape, and spatial position of the material.

2.5.4. Organic Field Effect Transistor Tactile Sensors

As a three-terminal electronic device distinct from two-terminal counterparts, organic field effect transistors (OFETs) are constructed using organic semiconductor materials. The three terminalssource, gate, and drainform the core electrode architecture, complemented by a substrate and an organic semiconductor layer in their typical structural configuration. , Operationally, applying a gate bias induces carrier accumulation at the interface between the insulating layer and organic semiconductor, forming a conductive channel between source and drain. Under a source–drain voltage, carriers are injected from the source into the semiconductor layer and transported through this channel to the drain. Mechanically, external pressure perturbs intermolecular spacing and molecular alignment within the semiconductor layer, modulating the degree of π–π conjugation and thereby influencing carrier transport characteristics. The implementation strategy for achieving anisotropic sensing in flexible devices using OFETs primarily hinges on modulating the directional transport properties of carriers within the organic semiconductor layer. This can be realized through material orientation design and device structure optimization. Organic semiconductors, such as poly-3-hexylthiophene (P3HT) and small-molecule pentacene, inherently exhibit anisotropic carrier mobility when their molecules are axially aligned. Alignment methods for organic semiconductors include stretch-induced orientation, template-guided orientation, and shear-induced orientation. These techniques effectively manipulate the molecular arrangement, leading to the desired anisotropic transport characteristics. Directional carrier transport can be achieved by designing interdigitated electrodes along specific axes and incorporating oriented nanofibers into the insulating layer. For instance, Cosseddu et al. employed a meniscus-guided printing method to induce the oriented alignment of pentacene (Figure h). They fabricated different strain sensors by designing anisotropic parallel patterns with a single orientation in the crystal direction and isotropic patterns radiating outward from the crystal center, which were then superimposed on interdigitated and helical electrodes, respectively. By comparing the signals from these different sensors, it was found that parallel-aligned crystal domains enable efficient charge-carrier transport along the crystal’s main axis (parallel to the channel length), enhancing the electrical conductivity in this direction. In contrast, the transport efficiency is significantly lower in the perpendicular direction, resulting in anisotropic electrical characteristics. OFETs offer advantages such as high sensitivity and ease of integration. However, challenges remain in ensuring the homogeneity of the piezoelectric film, and these devices are also susceptible to environmental influences.

In summary, different anisotropic sensors and their means of anisotropic realization are presented in this section. These means are roughly categorized into two groups: oriented structure design and surface microstructure design. Among the various types of flexible sensors mentioned above, the most common anisotropic sensors are piezoresistive sensors. This is due to the fact that the directional design of their conductive networks leads to significant anisotropic signal output capability. As a result, increasing reports on piezoresistive sensors have become available. However, for practical sensors (e.g., multimodal sensors that acquire multiple signals from an object), it is often more efficient to use multiple subsensors with different signal output mechanisms within a single sensor. In addition, anisotropic sensors are capable of outputting larger electrical signals in specific directions. Therefore, stacking and combining multiple anisotropic sensors to meet increasingly complex sensing requirements are the focus for future research.

3. Performance

When crafting flexible tactile sensors, it is essential to consider the key performance metrics including the flexibility, sensitivity, linearity, detection range, response and recovery times, drift, stability, resolution, and crosstalk for accurate data acquisition in various applications (Figure ). The primary goal of tactile sensor design is to imitate and even surpass the capabilities of human skin through the use of flexible and highly responsive materials. However, isotropic tactile sensors often face challenges in achieving the multidimensional signal detection capabilities of human skin. For anisotropic tactile sensors, performance parameters are expanded to include the detection of multidirectional forces. These performance benchmarks critically influence the functionality and longevity of anisotropic tactile sensors. In this section, we outline the common performance indicators.

5.

Six key performances of flexible sensors: (a) tensile strength at break, (b) sensitivity, (c) linearity, (d) drift, (e) response/recovery time, and (f) stability.

3.1. Flexibility

The flexibility of flexible tactile sensors primarily originates from the substrate, determined by intrinsic characteristics like Young’s modulus and elongation at break. Young’s modulus reflects the material’s resistance to deformation, where a lower modulus indicates a more flexible substrate. Elongation at break represents the maximum deformation tolerance for a tactile sensor before failure. Illustrated in Figure a, elongation at break serves as a standard gauge for a sensor’s flexibility, defined by the following formula:

$$\epsilon =\left(\frac{L-L_0}{L_0}\right)100\%$$ 3

In this equation, ε represents the strain, L denotes the stretched length, and L ₀ stands for the initial, unstrained length. As illustrated in Figure a, the horizontal axis corresponds to the deformation rate while the vertical axis indicates the applied tensile force. It can be observed that prolonged stretching results in a sharp decrease in applied stress, leading to rupture due to excessive elongation. The fracture strain, defining the strain at the point of breakage, emerges as a pivotal parameter. While the flexibility of flexible tactile sensors is primarily determined by the substrate material, research has indicated the potential for enhancing flexibility in hydrogels through hydrogen bonds with conductive fillers like MXene and graphene. These evenly dispersed conductive fillers act as physical cross-linking nodes via hydrogen bonding, rendering the hydrogels more resilient and significantly improving their deformability.

3.2. Sensitivity

Sensitivity quantifies a sensor’s ability to transduce mechanical stimuli into electrical signals, which is calculated as the ratio of signal change to pressure change. The sensitivity formula is expressed as

$$S=\frac{\mathrm{\Delta }E/E_0}{\mathrm{\Delta }P}$$ 4

In the equation, ΔE denotes the change in the measured electrical signal (such as resistance, current, voltage, or capacitance) from its initial value, ΔP represents the change in external stress, and E ₀ is the initial value of the electrical signal. Typically, in a plotted graph (Figure b), the horizontal axis illustrates sensor deformability due to external force, while the vertical axis shows the rate of change of the electrical signal with respect to its initial value. In Figure b, the sensor’s sensitivity is indicated by the slope of the curve, which is often expressed as the gauge factor (GF). Within various deformation ranges, the sensor exhibits varying sensitivities. For anisotropic tactile sensors, a significant difference in the gauge factor across different directions indicates superior anisotropic performance. By implementing anisotropic structures like microstructure such as microstructures (e.g., pyramids, pillars) and internally oriented architectures (including aligned fibers and oriented pores), sensor sensitivity can be notably enhanced. This sensitivity boost arises because anisotropic sensors generate more conductive pathways or experiencing larger deformations in response to minimal external stimuli compared to isotropic sensors.

3.3. Linearity and Linear Response Range

Linearity quantifies the relationship between an electrical signal and an external stimulus, reflecting the deviation of a tactile sensor’s static curve from a linear approximation. Actual sensor curves frequently deviate from ideal straight lines, necessitating an assessment of sensor performance by comparing deviations across sensors. Linearity is commonly represented as a percentage of the maximum variation in the electrical signal relative to its peak value, calculated using the following equation:

$$\delta =\frac{\mathrm{\Delta }{\mathrm{Y}}_{\max }}{Y}100\%$$ 5

Here, δ represents the percentage linearity value, ΔY _max indicates the maximum deviation between the sensor’s standard curve and the fitted linear regression line, and Y corresponds to the sensor’s maximum scale output. In this formula, the smaller the linearity percentage, the closer the measured linearity value aligns with the actual value. Linearity can also be gauged by the linear fit coefficient, R ² (ranging from 0 to 1); a higher R ² signifies better linearity. As depicted in Figure c, the red dashed line illustrates a strong linear response with a higher R ² value, which aids in signal processing and calibration. Conversely, the blue curve demonstrates a nonlinear response with a lower R ² value, which poses challenges for data analysis during operation. Therefore, when striving for high test accuracy in tactile sensor design, strategies should prioritize enhancing R ² . The linear response range delineates the pressure span within which the transducer generates a consistent signal as the external force varies. Within this range, the transducer converts external pressure into an electrical signal, displaying a linear curve. Sensors may exhibit multiple linear ranges primarily due to an uneven creation rate of conductive paths or short circuits during deformation of the sensitive material. Current research on tactile sensors concentrates on achieving heightened sensitivity across a broad operational span. Tactile sensors with anisotropic structures are widely acknowledged for enhancing sensitivity and linearity through increased conductive pathways during external stimulation, leading to an expanded contact area, thereby offering improved sensitivity and linearity.

3.4. Drift

Drift in flexible sensors refers to the unintended fluctuation in the sensor’s output signal over time under a constant input (Figure d). This phenomenon can be attributed to various factors, including material degradation, environmental influences, and stress relaxation. For example, ionic hydrogel sensors may undergo water loss during operation, which exacerbates signal drift. Viscoelastic materials like polydimethylsiloxane (PDMS) exhibit gradual creep deformation under continuous stress, resulting in changes in the sensor’s electrical properties such as resistance and capacitance. In ionic and electronically conductive hydrogel tactile sensors, ambient humidity variations can alter conductivity, leading to signal drift. To counteract the effects of drift, strategies such as improving interfacial adhesion or implementing robust encapsulation methods for hydrogel materials can be employed.

3.5. Response and Recovery Times

Response time indicates the duration required for a tactile sensor to produce a stable output signal upon pressure application, reflecting the sensor’s speed in responding to stimuli. A shorter response time implies a quicker response (Figure e). Recovery time denotes the duration required for the tactile sensor signal to stabilize after pressure removal, showcasing the sensor’s ability to restore stability after stimulus removal (Figure e). Designing tactile sensors involves strategies such as avoiding highly viscoelastic materials or implementing microstructural designs in such materials to effectively decrease response and recovery times.

3.6. Stability

Stability in tactile sensors denotes their capability to uphold their original performance after extended or repeated usage. Since stability assessments are conducted across a fixed range of forces or deformations, sensor stability can be depicted by the cyclic variations in the electrical signal ratio ΔE/E ₀ over a specified duration and number of pressure or relaxation cycles. The sensor’s performance retention is calculated by determining the signal fluctuation rate at the test conclusion compared to initial values. In the coordinate system illustrated in Figure f, the horizontal axis typically represents cycles or time, while the vertical axis signifies the electrical signal. Throughout the continuous force applications and releases, the output signal exhibits periodic stability. Tactile sensors exhibiting minimal signal degradation after extensive testing. The selection of substrate materials significantly influences the stability of tactile sensors. For instance, sensors with hydrogel substrates may encounter challenges like moisture loss, mechanical property deterioration, and material aging over time, affecting overall performance. , Moreover, stability is influenced by environmental conditions. Hence, for specific operational settings, precise selection of substrate material and structural design is imperative. , For instance, hydrophobic materials are preferred for underwater tactile sensors, while biocompatible or adhesive materials are essential for biomedical applications. −

3.7. Resolution and Crosstalk

In sensor technology, resolution is defined as the capacity to detect the smallest measurable change within a sensor’s effective deformation range. A sensor with a resolution of 0.1 N, for instance, requires a minimal force of 0.1 N to alter its electrical output. Tactile sensors, which aim to replicate the skin’s sensitivity to minor stimuli, typically demand higher resolutions. Often, employing microporous and microstructural designs enhances sensor resolution. Crosstalk, a phenomenon where electrical signals from adjacent sensors interfere, thereby reducing the accuracy of signal measurements, frequently occurs in multisensor configurations. For example, sensors in an array may touch each other or compress adjacent circuitry when deformed. Utilizing arrays of subsensors with anisotropic structures proves to be an effective strategy, as they are sensitive only along a specific axis. Therefore, deformation perpendicular to the sensitivity axis does not significantly affect the deformation in the parallel direction, thus effectively minimizing crosstalk.

3.8. Multidirectional Force Detection

To equip tactile sensors with multidimensional information perception resembling human skin, it is vital not only to detect unidirectional forces but also to differentiate multidirectional forces including normal force, shear force, and torsion. This capability facilitates identifying strain, pressure, stretch, and torsion. , In general, the ability to detect multidirectional forces derives from differences in electrical signals in different directions. For example, strain–stress curves differ significantly between the X-axis and Z-axis directions. Another example is polarity differences in electrical signals when forces in different directions are applied to the sensor. The detection of multidimensional information can be achieved through three strategies: (1) integrating multiple sensors to analyze diverse stimuli and their orientations, ,, (2) designing tactile sensors with unidirectional sensitive structures to enhance sensitivity variances across different orientations, ,,, and (3) engineering sensor microstructures to detect multidirectional forces. , Employing tactile sensors with distinct sensing functions and separately detecting various signals helps prevent signal interference, such as combining strain sensors with pressure sensors for multidimensional perception. Alternatively, implementing sensor arrays of identical tactile sensors can monitor object pressure and motion direction within a plane. For individual sensors, designing mechano-/electro-anisotropic sensors that are sensitive in a specific direction or plane effectively avoids signal crosstalk, such as utilizing stretchable fibers for strain sensitivity without pressure sensitivity, and aerogels for pressure sensitivity without strain sensitivity, enhancing sensor performance. , Incorporating surface or internal microstructures in tactile sensors is also crucial for detecting multiple forces. Surface microstructures like microneedles, hemispheres, cones, and pyramids yield distinct deformations and generate different electrical signals when subjected to normal and shear forces.

4. Constructure Materials of Sensors

Tactile sensors function by employing a sensitive layer situated between two electrodes, which comprises sensitive and substrate materials. Anisotropic tactile sensors primarily employ pressure-sensitive materials capable of generating various electrical signals such as resistance, voltage, and capacitance upon pressure application. Substrate materials typically consist of soft polymers with a degree of flexibility. The performance of sensors is intricately linked to the sensitive materials, which are categorized into metals, polymers, and inorganic nonmetal materials. Furthermore, sensitive materials can range from 0D (zero-dimensional) to 3D (three-dimensional) based on dimensionality. This section details the material composition of anisotropic tactile sensors based on these classifications. The use of conductive polymer fillers in anisotropic sensor fabrication is minimal and thus will not be elaborated on in this context. Polymers can be categorized based on their mechanical properties into rubber, plastics, and fibers, with several polymers commonly used in sensor substrates outlined below.

4.1. Polymeric Materials

Polymeric materials, valued for their lightweight, mechanical flexibility, compatibility, and moldability, are commonly employed as substrates in crafting anisotropic tactile sensors. The flexibility of flexible tactile sensors heavily relies on the flexibility of the substrate material. This flexibility is typically assessed through parameters like tensile strength, Young’s modulus, fracture stress, hardness, and density, each providing insight into the sensor’s deformation capacity to different degrees.

Generally speaking, the substrate materials for anisotropic flexible sensors can be categorized into rubber, polyester, plastic, cellulose, and gel. Typical examples include PDMS, , thermoplastic polyurethane (TPU), polyvinylidene fluoride (PVDF), cellulose, polyacrylamide (PAAm), poly­(acrylic acid) (PAA), and poly­(vinyl alcohol) (PVA). These materials are not inherently anisotropic but are anisotropically designed to exhibit anisotropic electrical signal output capabilities. For example, the conductive filler inside is induced to be oriented using an external field, and the cellulose or hydrogel fibers are designed to be oriented. For different systems, a rational choice of substrate is critical. Wearable sensors requiring skin conformity must prioritize nontoxicity and biocompatibility, for example, PDMS and TPU, which are not light water, nontoxic, and biocompatible. For implantable sensors with stricter requirements, hydrogels are preferable due to their high water content and enhanced conformability.

Due to better biocompatibility and flexibility, some common conductive polymers are also used in the design of flexible sensors. Examples include polypyrrole, polyaniline, and polythiophene. However, these conductive polymers typically form nanoparticles, so incorporating them into anisotropic sensors depends on the structural design of the flexible substrate. For example, the conductive polymers are grown in situ on the surface of cellulose, and the cellulose with the conductive polymers is oriented through anisotropic structural design to give the sensor anisotropic conductivity. Specifically, depositing polypyrrole (PPy) and polyaniline (PANI) on the surface of cellulose and using cellulose’s one-dimensionality to orient the polymers can give the sensor significant anisotropic conductivity (Figure a). , Lin et al. coated PPy on the surface of cellulose nanofibers (CNF) before incorporating them into PVA and then prestretched it to form an anisotropic conductive hydrogel. The hydrogels exhibited significant directional differences in both conductivity and sensitivity.

6.

Tactile sensors with polymers and metals as conductive fillers. (a) Anisotropic hydrogels obtained by the prestretching method to induce orientation of PVA-PPy@CNF hydrogels. Reproduced with permission from ref . Copyright 2025, Wiley-VCH. (b) Oriented arrangement of silver nanowires, and sensor detection of eye movement direction. Reproduced with permission from ref . Copyright 2021, Elsevier. (c) Silver-plated wavy elastomer surface. Reproduced with permission from ref . Copyright 2018, Wiley-VCH. (d) Structure of anisotropic liquid metals senor. (e) Schematic diagram of a light pole bar showing the relationship between microdeformation and resistance. The test results of resistance testing of ACG in both vertical and horizontal directions using a digital multimeter. Reproduced with permission from ref . Copyright 2024, Wiley-VCH.

4.2. Metals

Metallic materials have been a focal point of scientific research for a long time. In the domain of anisotropic flexible tactile sensors, researchers commonly incorporate metals into microstructures or nanostructures to meet the demands of anisotropy. Generally, metallic materials used in anisotropic tactile sensor design are classified as 1D (one-dimensional) nanowires. − Among them, silver nanowires (AgNWs) are widely used in anisotropic flexible tactile sensors due to their high electrical conductivity (6.39 × 10⁷ S/m) and thermal conductivity (429 W/(m·K)). , As metal particles themselves lack anisotropy, they are usually combined with other materials to achieve anisotropic properties. Two common approaches involve designing metals as 1D nanowires and orienting them through specific fabrication processes, or using an external field to orientate the metal nanoparticles. For instance, Ko et al. developed a self-healing triboelectric tactile sensor founded on the anisotropic arrangement of silver nanowires, enhancing directional force sensitivity to enable the detection of low-pressure movements below 100 Pa (Figure b). The sensor’s anisotropic dielectric properties enabled the detection of ocular motion in eight directions. An alternative method involves depositing metal particles on oriented or microstructured polymer materials, achieving anisotropy through the design of the flexible substrate. Kim et al. proposed a high-sensitivity, stretchable strain sensor that integrated AgNWs into elastomer composite films (Figure c). Uniform dispersion of AgNWs on the microstructured surface endowed the sensor with piezoresistive capabilities, facilitating strain detection under bending and pressure. The sensor demonstrates high sensitivity (with a GF of approximately 81 for strains >130%), stretchability (150%), and long-term reliability (10,000 cycles at 150% strain). In addition to traditional metals, the use of liquid metals in tactile sensors has been increasingly reported. Eutectic gallium indium (EGaIn) liquid metal is commonly employed in producing anisotropic soft tactile sensors. While liquids are usually isotropic due to their random flow, liquid metals can create permanent anisotropic structures when combined with stimuli-responsive solids. For instance, Peng et al. mixed gallium with conventional metals like indium and bismuth, utilizing gel encapsulation after isolation with copper oxide to fabricate an anisotropic conductive gel (ACG) displayed in Figure d, providing a method to confer anisotropic conductivity on flexible sensors. The ACG exhibits ultrasensitivity to minor deformations due to its intrinsic anisotropic conductivity, acting as an insulator horizontally and conducting vertically (Figure e). Thanks to its anisotropic structure, the ACG can accurately and simultaneously detect subtle strains and vibrational frequencies with accuracy rates up to 99%.

4.3. Inorganic Nonmetallic Materials

Most inorganic nonmetallic materials used in anisotropic tactile sensors are carbon-based conductive materials, which offer greater environmental friendliness and renewability compared to metal-based conductive materials. Carbon-based materials such as carbon nanotubes, graphene, and MXene exhibit high conductivity; consequently, research centers on these materials.

CNTs are unique one-dimensional materials made of carbon atoms, with radial dimensions within the nanoscale and axial dimensions within the microscale, which exhibit exceptional thermal, electrical, and mechanical properties. Oriented CNT sheets typically exhibit superior electrical conductivity (1.3 × 10⁶–3.2 × 10⁶ S/m) compared to CNT network films. Consequently, CNTs are often structured in oriented forms to transfer their outstanding chemical, mechanical, and electrical characteristics to the macroscopic level, contributing to high sensing performance. In the development of anisotropic tactile sensors, techniques such as magnetic and electric field alignment of CNT orientation are commonly employed. Liu et al. first induced the perpendicular alignment of MWCNTs under a weak magnetic field of 0.6 T to create vertically aligned carbon nanotubes/PDMS (VCPs) with anisotropic properties, and then enhanced its anisotropy by incorporating semicircular microstructures, and named the final sensor m-VCP (Figure a). The combination of these aspects significantly boosts the sensitivity of the flexible sensor in the pressure direction. In the low strain range (0–6%), the sensitivity of m-VCP (GF = 9.208) is 49% higher than that of the unoriented microstructured tactile sensor (m-CP) and 86% higher than that of VCP. A prevalent strategy involves dispersing CNTs in substrate materials and subsequently creating anisotropic tactile sensors using methods such as 3D printing, directional freezing, or surface microstructuring. − For example, Xia et al. designed a tactile sensor based on carbon fiber (CF) and MWCNT-filled PDMS, as depicted in Figure b. The CNTs inside the sensor exhibit a directionally aligned morphology due to shear flow, resulting in different sensitivities of the sensor in different directions.

7.

Tactile sensors with carbon-based materials as conductive fillers. (a) Sensor structure with internally oriented carbon tubes and schematic diagram of pathways in different compression directions. Reproduced with permission from ref . Copyright 2023, American Chemical Society. (b) Schematic diagram of the internal structure of a tactile sensor developed based on CNTs. Due to the action of shear force, a large number of carbon tubes maintain consistent orientation. Reproduced with permission from ref . Copyright 2021, Wiley-VCH. (c) Schematic diagram of a sensor composed of RGO and PDMS developed using directional freezing. The RGO is assembled along the temperature gradient direction, resulting in an anisotropic structure for the sensor. Reproduced with permission from ref . Copyright 2017, Royal Society of Chemistry. (d) Schematic diagram of MXene-based conductive hydrogel with oriented structure. The hydrogel has good stretchability and compressibility. Reproduced with permission from ref . Copyright 2021, Wiley-VCH. (e) Structural diagram of MXene-modified porous wood sponge skeleton and its structure in different directions. Reproduced with permission from ref . Copyright 2023, Elsevier.

Graphene, a carbon allotrope, consists of sp²-hybridized carbon atoms arranged in a single-layer hexagonal lattice structure. It is renowned for its exceptional optical, electrical properties (e.g., conductivity of 10⁸ S/m), and mechanical strength and often serves as a conductive filler in tactile sensors. Graphene is commonly combined with other conductive fillers or nonconductive polymer substrates to enhance network conductivity. , Designing anisotropic structures using graphene as a conductive filler to achieve directional signal differentiation presents significant challenges, stimulating extensive research efforts. − One prevalent method for constructing anisotropic tactile sensors involves embedding two-dimensional graphene into flexible materials and designing macroscopic structures. Lu et al. proposed a honeycomb-like reduced graphene oxide (RGO) foam synthesized through unidirectional freeze-drying and mechanical compressiondepicted in Figure cand encapsulated within PDMS. The PDMS shell and honeycomb structure collectively provide exceptional flexibility, high stretchability, sensing sensitivity, and anisotropic mechanical and sensory performance perpendicular and parallel to the stretching direction.

MXene, a two-dimensional inorganic compound similar to graphene, comprises transition metal carbides, nitrides, or carbonitrides with a thickness of a few atomic layers. , The MXene layers are linked by van der Waals forces, which allow electrons to move freely to create conductive pathways. Additionally, the MXene surface is rich in hydroxyl groups and terminal oxygen atoms, which exhibit electrical conductivity skin to that of transition metal carbides. Consequently, MXene is commonly employed as a conductive filler in sensors. Due to extensive hydrogen bonding on its surface, it is often combined with hydrogels in sensor design to enhance hydrogel stretchability and impart electrical conductivity. Drawing inspiration from the ordered muscle structure, Lin et al. fabricated anisotropic MXene/PVA conductive hydrogels (PMZn-GL) using the directed freezing and immersion in ZnSO₄ solution (Figure d). The anisotropic nature of PMZn-GL hydrogels is evident in their superior mechanical properties and electrical conductivity horizontally versus vertically. Leveraging MXene’s excellent conductivity enables the hydrogel to produce a stable resistance signal at a low voltage of 0.05 V, offering safety and portability. Widely used as a conductive filler, MXene is prominent in aerogel design for tactile sensors. For example, Li et al. transformed wood blocks into compressible wood sponges with low density and highly anisotropic porous structures through thermal treatment and then functionalized them with MXene to create conductive, highly compressible MXene-wood (MW) sponges (Figure e). These sensors feature a broad operational range of 0–25 kPa and a wide linear sensing range of 5–50% and 0–180° and exhibit excellent long-term reliability over 5000 cycles at a 30° bending angle. Suitable for monitoring physiological signals, these sensors can detect pulse beats, finger movements, and wrist flexions.

In summary, the selection of materials for anisotropic sensors should prioritize 1D or 2D (two-dimensional) substances, as these materials inherently exhibit pronounced anisotropy and tend to orient readily within polymer substrates. The development of sensors based on such materials facilitates the rapid fabrication of anisotropic conductivity. For zero-dimensional (0D) materials lacking intrinsic anisotropy, a viable strategy involves incorporating them with 1D nonconductive substances to engineer anisotropic sensor designs. However, in the context of resistive sensors, most existing hydrogel-based anisotropic designs rely on ions to enhance mechanical toughness or regulate morphological structuremodifications that can significantly affect the anisotropic signal response by blurring conductivity differences. Therefore, amplifying the contrast in conductivity signal output under directional stimuli constitutes a critical focus for future research. On another front, anisotropic aerogel sensors demonstrate versatility across piezoresistive, piezoelectric, capacitive, and triboelectric sensing modalities. Yet, their susceptibility to irreversible deformation during sustained compression necessitates encapsulation with resilient elastomers to improve compressive resistance and cyclic stability. In conclusion, addressing these challenges requires future investigations into anisotropic sensors to focus on two key avenues: (1) rational selection of substrate matrices and filler materials to optimize structural anisotropy, and (2) development of innovative material combinations that reconcile conductivity with mechanical durability and signal specificity.

5. Construct Design

There are usually three ways to build anisotropic tactile sensors. The first approach uses multiple subsensors to form a multisensor system. In the pursuit of more precise signals and versatile detection capabilities for tactile sensors, researchers have integrated isotropic with anisotropic sensors, endowing them with multidimensional sensing capabilities. The second approach designs anisotropic tactile sensors with an internally oriented structure, featuring a flexible substrate as the outer layer and sensitive material as the inner layer. This configuration offers numerous advantages by enhancing stability, reliability, and longevity. The outer flexible substrate provides deformability and protection to the internal sensitive material against chemical and physical damage. Recent research has shifted toward microstructural design, aiming to mimic human skin or biological exoskeletons for tactile sensor development. The third method incorporates surface microstructures such as pyramidal, cylindrical, or hemispherical shapes into the flexible substrate and sensitive materials to increase sensitivity while reducing resistance. This section outlines the structural design of anisotropic tactile sensors focusing on three directions: multisensor arrays, internally oriented sensors, and surface-microstructured sensors.

5.1. Multisensor

The combination of multiple subsensors capable of detecting signals of different types or orientations allows a sensor system to achieve multidimensional detection capabilities. Owing to its procedural simplicity, this approach is commonly employed in the design of anisotropic sensors. Based on distinct structural configurations, such systems are classified into the following two categories: integrated sensors − and sensor arrays.

5.1.1. Integrated Sensors

Integrated sensors, composed of multiple subsensors vertically stacked, demonstrate robust multidimensional sensing abilities, capable of detecting diverse signals and measuring both normal and tangential forces. − Two dominant design strategies exist: one stacks subsensors of identical material with varied sensitivities, and the other approach stacks subsensors outputting different signals. For sensors with identical electrical characteristics, insulation between layers with a PDMS flexible film or insulating diaphragm is critical. Sensors intended to detect multidirectional signals must exhibit significant anisotropic deformability or conductivity, enabling discernment between directionally distinct signals under deformation. For instance, Li et al. developed an anisotropic hydrogel with an internal conductive filler of precisely aligned carbon fibers using 3D printing (Figure a). The anisotropy of this sensor is caused by the flow field, where the shear flow of the conductive liquid as it is ejected from the 3D printing nozzle aligns the carbon fibers. By combining a needle-like microstructure of hydrogel with two anisotropic hydrogels into a single unit, the integrated sensor can detect stimuli along the X, Y, and Z axes, exhibiting distinct sensitivities along each axis. Notably, the sensitivity differs significantly between the X and Y axes when stretched in distinct directions, while the Z-axis detection is facilitated by the needle-like microstructures. This design allows the sensors to precisely detect changes in electrical signals in the specific direction of deformation, making them highly suitable for human motion signal detection. Furthermore, leveraging intricate designs that merge natural and synthetic materials, anisotropic integrated sensors can be crafted. Drawing inspiration from the complex structure of articular cartilage tissue, Jiang et al. used topological and zipper-shear chains to combine soft and hard materials, resulting in a wood and hydrogel composite hydrogel. This design strategically integrates soft and hard materials, balancing reinforced composite properties with flexibility and high toughness. In their study, the modulation of the sensor’s electrical conductivity and tensile properties was achieved by merging polyacrylamide (PAAm), a soft material, with wood, a hard material from which lignin and hemilignin had been removed. Controlling the wood’s fiber pore structure in intermediate segments enhanced substantial anisotropy (Figure b). This sensor has a large difference in sensitivity not only at large deformations (0–150%) but also at small bends (0–5%).

8.

Different types of integrated sensors. (a) Schematic of a tactile sensor integrated by the same hydrogel and its anisotropic sensing capability. Reproduced with permission from ref . Copyright 2024, Elsevier. (b) Schematic representation of the tactile sensor integrated by wood hydrogel and its sensitivity in different orientations. Reproduced with permission from ref . Copyright 2024, Wiley-VCH. (c) Schematic of the integrated sensor consisting of piezocapacitive and piezoresistive sensors and their respective sensitivities. Reproduced with permission from ref . Copyright 2023, Springer Nature. (d) Schematic diagram of an integrated sensor consisting of triboelectric and piezocapacitive sensors and its capability of detecting forces in different directions. Reproduced with permission from ref . Copyright 2025, Elsevier.

In comparison to sensors made up of subsensors detecting identical signals, sensors created by incorporating subsensors that perceive diverse signal modes exhibit increased complexity. For instance, the successful transmission of capacitive signals mandates the placement of electrode layers both above and below the sensor. ,, In such configurations, a subsensor designed to emit resistive signals may also serve as an electrode layer. Furthermore, when a friction electric sensor is coupled with either a piezoresistive or a piezocapacitive sensor, the electrode layer within the friction electric sensor is capable of generating either resistive or capacitive signals. , Ma and colleagues successfully developed anisotropic strain and deformation-insensitive pressure sensors using rigid-flexible synergistic stretchable substrates (Figure c). The sensor’s fabrication involved four steps: (1) depositing PVA on electrospun fibers; (2) screen-printing conductive ink and liquid metal on PVA; (3) prestretching to induce microcracks; (4) electrospinning TPU onto PVA for protection. This layered approach not only secures the placement of the PVA but also results in anisotropic electrical conductivity and capacitive signal output capabilities in the sensor. By layering liquid metal sensors between conductive ink sensors, this design enables detection of surface deformations and measurement of vertical pressure. Nonetheless, both resistive and capacitive sensors necessitate an external power source. In light of energy conservation efforts, an increasing number of integrated sensor designs are incorporating frictional electricity and piezoelectric sensors. , Guo et al. have innovated a self-powered, multimodal wireless tactile sensor (TC-MWTS) utilizing triboelectric and capacitive coupling effects (Figure d). This sensor consists of three strata: the triboelectric normal force sensing layer (TNFSL), the medium stretchable Ecoflex layer (MSEL), and the capacitive shear force sensing layer (CSFSL). The unique presence of dual time-frequency domain information in the wireless signals enables this sensor to output normal force and shear force signals as voltage amplitude and eigenfrequency, respectively, without necessitating crosstalk and complex signal decoupling. However, constructing multilayered electronic skin devices introduces challenges in ensuring adhesion between the sensory layers and overcoming mechanical mismatches.

5.1.2. Sensor Arrays

Sensor arrays are created by arranging either identical or varied sensors on a circuit board. ,, Below the board, an extensive network of circuits links the sensors, operating independently without signal interference. Homogeneous sensor arrays adeptly identify item movements and collect data on the sensors’ surfaces. Additionally, these boards can be customized into shapes (e.g., insoles or hand forms) for specific applications. For example, Zhou et al. developed a breathable plantar pressure sensor via electrospun fiber film, offering high sensitivity and a broad detection range. Illustrated in Figure a, an individual’s walking posture can be broken down into eight distinct movements, each associated with a unique contact point between the foot and sensor arrays. This granular differentiation enables the detailed capture and analysis of human locomotion to identify irregular motion patterns. Similarly, Lin et al. described a piezoelectric tactile sensor array with a skin-like multilayered structure and row-column electrodes. This sophisticated array incorporates a signal processor and logic algorithms, adapting to various external stimuli detection methods. Demonstrated in Figure b, the sensor precisely locates stimuli by comparing piezoelectric signals from upper and lower layers, effectively determining the stimulus’s direction based on differential subsensor responses. To imbue sensors with more complex functionalities and enhance the accuracy of function recognition, sensors capable of discerning intricate information may be arrayed. For example, by integrating strain and pressure sensors within an array, Liu et al. fabricated a three-dimensional tactile sensor array that simulates the capabilities of Merkel cells and Ruffini corpuscles in human skin. This array, illustrated in Figure c, was achieved through a meticulously controlled mechanical assembly and a diverse packaging approach for 3D devices. Its distinct design, featuring systematically positioned pressure and strain sensors, facilitates the concurrent detection of both pressures and strain. Furthermore, varying intensities of strain within this 3D sensor enable the differentiation of object softness and the determination of local curvature via distinct distributions of normal and shear forces. Moreover, Cui et al. engineered a bimodal sensor that can discern both softness and force exerted by objects, merging a protrusion designed for softness detection with a pressure sensor, as depicted in Figure d. A compilation of 112 such bimodal sensors was integrated into a hand-shaped sensor array, equipped with an adaptable external control system. This hand-shaped array accurately identified four physical characteristics of a human and four organ modelsaddressing both healthy and pathological stateswithin an abdominal simulator, achieving 98% accuracy. Although sensor arrays present detailed spatial data and reduce signal interference, they confront challenges like complex circuit designs and extensive wiring. In this light, the development of anisotropically structured tactile sensors is deemed highly beneficial.

9.

Different types of sensor arrays. (a) Schematic representation of a plantar tactile sensor array based on a breathable spun fiber membrane composition and the human walking posture it can detect. Reproduced with permission from ref . Copyright 2024, Wiley-VCH. (b) Schematic representation of a human skin-inspired tactile sensors array, and the sensors’ ability to detect pressure and the direction of motion of a small insect. Reproduced with permission from ref . Copyright 2021, Wiley-VCH. (c) Schematic structure of a flexible tactile sensor with three-dimensional force detection capability. Reproduced with permission from ref . Copyright 2024, AAAS. (d) Structure of a bimodal tactile sensor capable of detecting softness and pressure and a palm-like sensor array comprising it. Reproduced with permission from ref . Copyright 2022, Wiley-VCH.

In summary, the integration of multiple sensors and the design of sensor arrays enable discriminate perception of external forces’ direction and magnitude. Sensor integrationby embedding subsensors with multimodal signal outputsmimics the skin’s multimodal sensing ability, allowing differentiation between normal (perpendicular) and tangential (shear) forces through distinct signal responses. Sensor array designs, conversely, facilitate efficient discrimination of surface object motion direction and force magnitude through spatially distributed sensing elements. However, challenges persist; integrated sensors suffer from interlayer mismatch (e.g., mechanical/electrical property discrepancy between layers), while sensor arrays face alignment precision and crosstalk interference issues. These limitations are significantly mitigated by using the anisotropically structured subsensors, whose directionally dependent mechanical/electrical properties reduce postprocessing complexity for force decoupling.

5.2. Internal Orientation

A single tactile sensor requires the design of its internal structure and materials to achieve differentiation between the direction and magnitude of applied forces. Typically, three methodologies achieve this. The first method structures the conductive filler within the flexible sensor into a directional layout. For example, in piezoresistive sensors, conductive material alignment yields higher conductive pathway density along the parallel versus perpendicular axis. This arrangement accentuates the sensor’s sensitivity under compression or stretching along specific orientations, generating pronounced anisotropic responses. The second approach forms the flexible sensor with an inherent anisotropic configuration, diversifying its response under compression and stretching across different axes. This structural anisotropy differentiates mechanical behavior and produces distinct electrical signal variations upon deformation. The third strategy employs self-assembly to design the nonconductive component of the sensor for optoelectronic signal output. This method is predominantly used in optoelectronic tactile sensors by merging resistive sensing with dynamic color changes from photonic crystals induced by stretching. The inherent structural colors of these ordered crystals cause the sensor’s color to transition from red to blue under stretching or compression. Furthermore, infusing conductive materials into elastomers or hydrogels surrounding these crystals bestows electrical conductivity.

5.2.1. Orientation of Internal Conductive Materials

The design of anisotropic sensors that orient the conductive fillers usually relies on an external field. Anisotropic material design using external fields (e.g., magnetic, flow, electric, or force fields) is an effective approach. Magnetic field-assisted alignment has historically dominated anisotropic sensor design. ,, However, out-field preparation of anisotropic materials is widely used in actuators, with fewer applications in sensors. , The orientation of the magnetic field to the filler is highly dependent on the permeability of the filler, the viscosity of the precursor fluid, and the strength of the applied magnetic field. For example, Zhang et al. designed a hydrogel precursor solution with dispersed MXene to orient in parallel orientation by varying the magnetic field parameters and MXene additions, so that the hydrogel sensor was able to produce a doubling of the difference in electrical conductivity between the parallel and perpendicular orientation directions (Figure a). Because MXene surface unsaturated bonds form hydrogen bonds with water, stretchability increases along the alignment direction. The prepared sensors have better tensile properties in the parallel direction than in the perpendicular direction and have fast tensile response time and recovery time, which can be used for anisotropic strain sensors. In the past, anisotropic flexible sensors prepared using liquid metal have poor conductivity in the relaxed state because the elastomer wraps around the liquid metal, causing it to form a small number of conductive pathways. Yun et al. constructed liquid metal anisotropic flexible sensors using magnetic field-assisted orientation of the liquid metal in PDMS, as shown in Figure b. The liquid metal anisotropic flexible sensors were constructed using a magnetic field-assisted orientation of liquid metal in PDMS. Under a magnetic field of 1 T, the carboxy-Fe magnetic particles are aligned. This alignment cuts the liquid metal into small droplets and also connects the liquid metal to form a conductive network. The resistivity of the sensor is 1.34 kΩ·m in the orientation direction and increases to 13.2 kΩ·m in the perpendicular orientation direction. As shown in Figure b, for the same deformation, the sensor has different conductive paths in different directions, which leads to different conductive properties. The flow field-induced filler orientation inside the sensor is also an effective way to achieve anisotropy, which is based on the principle that the conductive filler in the precursor fluid is orientated by extruding the precursor fluid from the screw or capillary tube and letting the velocity gradient be perpendicular to the direction of fluid motion. , Bao’s team used shear flow and nanolimited domain effects to achieve multiscale ordering and alignment of conjugated polymers in stretchable semiconductors. The preparation process is shown in Figure c; shearing of the semiconductor within the solution using patterned microgroove-coated blades resulted in macroscopic alignment of conjugated polymer nanostructures along the direction of charge transport. At the same time, the nanorestricted domain effect aligns the chain conformation, lowering the energy barrier for charge-carrier transport. The semiconductor conductive films are stretched at 100% with a pronounced orientation structure in the restretching direction. Recent studies have demonstrated anisotropic tactile sensor designs via 3D/4D printing, , reports have begun to appear on the design of anisotropic tactile sensors using 3D printing. The principle is to control the fluidity and viscosity of the liquid while using a flow field to orient the internal conductive fibers so that the fibers gain conductivity in a parallel oriented structure. For example, Huang et al. used 3D printing to design a conductive silicone rubber (CSR), in which conductive carbon fibers were added and the flow and viscosity of the CSR were altered with a thixotropic agent. Due to the flow field, CSR fibers are oriented along the printing direction, leading to anisotropic conductivity and mechanical properties. The fibers inside the sensor have a clear morphology oriented along the extrusion direction, as can be seen in Figure d. Moreover, with obvious anisotropic conductivity, the sensor has different resistance change rates in parallel and perpendicular orientations at the same bending angle.

10.

Anisotropic tactile sensors with internal conductive packing orientation. (a) Conducting hydrogel obtained by external magnetic field orientation of MXene, and its anisotropic stretchability and conductivity. Reproduced with permission from ref . Copyright 2023, Elsevier. (b) Anisotropic liquid metal sensor obtained by magnetic field-induced orientation of iron powder and its conductive principle. Reproduced with permission from ref . Copyright 2020, Elsevier. (c) Schematic representation of the oriented structure of the conductive film and the fibers inside it obtained by shear flow. Reproduced with permission from ref . Copyright 2019, Springer Nature. (d) Sensors with internal conductive fiber orientation obtained by 3d printing, and the anisotropy detection capability of the sensors. Reproduced with permission from ref . Copyright 2018, Elsevier.

In conclusion, this section summarizes the design of conductive materials within sensors, oriented via typical external field-induced methods (e.g., magnetic/flow fields), endowing sensors with anisotropic electrical signal output during deformation. Owing to conductive filler surface functional groups and their intrinsic 1D or 2D properties, mechanical properties along the orientation direction are also moderately enhanced. Future research may focus on modifying filler surface groups and via in situ growth of additional materials on conductive fillers to concurrently improve conductivity and mechanical performance.

5.2.2. Orientation of the Sensor Structure

Anisotropic design by structuring the sensor is more workable than designing the conductive material with an external field because it is simpler to manipulate. In the past reports, there are four main means of designing anisotropic sensors with anisotropic structures. The first is the anisotropic design of polymers using directional freezing, , the second is the use of prestretching to orientate long chain polymers, − the third is the modification of the wood structure, − and the fourth is the directional rearrangement of the conductive fibers. ,,

Orientation freezing techniques utilize a bottom-up temperature gradient to align polymers within a precursor liquid. The underlying principle involves placing a special mold containing the precursor liquid into a colder liquid environment. Due to the high thermal conductivity of the metal at the mold’s base, numerous oriented ice crystals form at the bottom, facilitating polymerization or aligning monomers and leading to an internally oriented structure. This technique enables the preparation of anisotropic polymers, which can be transformed into aerogels through freeze-drying. , For instance, Yu and his team developed an electrically conductive hydrogel using this method, featuring a reinforced concrete-type structure (Figure a). The interior conductivity was enhanced by AgNWs, while the anisotropic mechanical properties resulted from the conductive backbone established by the directional freezing. The direction parallel to tensile current is defined as x-axis, and the orthogonal direction is defined as y-axis. The sliding of a glass rod on the hydrogel surface produces compressive and cumulative tensile deformation perpendicular to the sliding direction. When the rod moves along the x-axis, the sum of compressive strain resistance (ΔR _c > 0) and cumulative tensile strain resistance (ΔR _Ty ≈ 0) exceeds 0. When moved along the y-axis, the combined change in compressive (ΔR _c < 0) and tensile (ΔR _Tx > 0) deformation resistance is negative. Thus, the sensor can identify the direction and speed of the rod’s motion.

11.

Sensors with anisotropic structures. (a) Structures of AgNWs/PAAm hydrogels with anisotropic structures obtained by directional freezing and signal changes of small rods moving in different directions on their surfaces. Reproduced with permission from ref . Copyright 2022, Springer Nature. (b) Hydrogels obtained by prestretching plus salting strategy, and their anisotropic conductivity and sensitivity. Reproduced with permission from ref . Copyright 2025, Elsevier. (c) Hydrogel sensors obtained from delignified balsa wood with deformability and current signals in different directions. Reproduced with permission from ref . Copyright 2022, Elsevier. (d) Schematic synthesis of oriented carbon fibers and their sensitivity curves in different directions. Reproduced with permission from ref . Copyright 2022, Royal Society of Chemistry.

By applying an external force to prestretch the polymer, the polymer chains align along the direction of the stretch, thereby enhancing its mechanical properties in that specific orientation. Following this prestretching, immersion in a specific ionic solution alters the polymer’s aggregation state (via the Hofmeister effect), resulting in a modulus-contrast structure. , This method is widely employed in the design of anisotropic ion-conducting hydrogels. For instance, Lin et al. developed robust anisotropic hydrogels composed of chitosan and P­(AM-AA) through a salting-out-assisted stretch remodeling treatment (Figure b). This process involves prestretching the hydrogel and then inducing salting out to lock in the structure, thereby endowing the hydrogel with anisotropic ionic conductivity and mechanical properties. Importantly, the conductivity of the hydrogel in the direction parallel to the stretch is twice that of the perpendicular direction. Similarly, the sensor’s sensitivity nearly doubles in different orientations. Such sensors can be used to gesture recognition and measuring the intensity of facial expressions.

The assembly of cellulose into oriented structures by directional freezing is an important means of constructing anisotropic sensors, but cellulose is often obtained by stripping wood of its components, leading to complexity in sensor preparation via this method. The use of natural balsa wood to build elastomers is considered to be a simple and efficient solution for anisotropic sensor design, as it is easily prepared while retaining its original anisotropic structure. A typical scheme is to cotreat balsa wood with NaOH/NaClO or NaOH/Na₂SO₃ solutions to remove the hemilignin and lignin and other cellular contents, leaving only the cellulose skeleton. , Anisotropic wood sensors can be constructed either by adding other conductive substances to the cellulose skeleton or by introducing polymers for encapsulation. For example, Yan et al. fabricated an all-wood hydrogel using PAM-coated wood (Figure c). Thanks to the anisotropic wood structure and the Fe³⁺ metal–ligand bonds inside the hydrogel, at least 50% compressive strain exhibits significantly divergent compressive strengths along orthogonal axes. When used for throat vibration detection, the signals exhibited significant directional divergence.

Through textile methods such as 3D printing , or electrostatic spinning, − large quantities of conductive fibers can be obtained. After the conductive fibers are arranged in an oriented direction and encapsulated in an elastomeric film (e.g., PDMS), anisotropic flexible sensors can be obtained. The sensors obtained in this way not only have oriented arrangement of conductive fillers but also have an anisotropic structure. At the same time, such sensors exhibit significantly different sensitivities in different directions within the same operating interval. Hu et al. fabricated a carbon nanofiber anisotropic film by employing electrospinning techniques combined with calcination, where fibers, tasked with signal detection in various directions, were isolated by a PDMS film (Figure d). This anisotropic film demonstrates considerably divergent conductivity and sensitivity in the two orientations. Specifically, the sensitivity along the direction parallel to the fibers is recorded at 442 kPa^–1, whereas it drastically drops to 35 kPa^–1 in the perpendicular direction. These conductivity discrepancies directly cause the distinct piezoresistive properties of the fiber sensors discussed above.

However, anisotropic structure design solutions are not limited to the above four common ones. For example, anisotropic microporous sensors have been fabricated by combining melt blending and chemical etching, and sensors with multidimensional force detection capability can be obtained by making fibers into fabrics with knitted structures. − Future research will likely yield additional anisotropic structures as understanding of anisotropic structural mechanics advances.

5.2.3. Oriention of Non-Conducting Material

Both the anisotropic design of the structure of the sensor and the conductive filler are aimed at making the sensor exhibit different electrical signal output capabilities when stimulated in different directions. Researchers are beginning to design sensors without anisotropic conductivity to have anisotropic optical properties. These designs typically rely on the self-assembly of the material to form ordered columns of structures that give them stretching color-changing properties. Self-assembly is the process by which small unitary components spontaneously organize into an ordered arrangement via noncovalent bonding interactions. For instance, during preparation, molecules of cholesteric liquid crystals self-assemble into a helical, layered configuration. This configuration is optically anisotropic, selectively reflecting light. Consequently, polymers configured in this manner display variable layer spacing under tension or compression, leading to a gradual shift in the sensor’s reflected light wavelength from longer to shorter wavelengths. − Based on this phenomenon, Lagerwall et al. developed a stretchable, color-changing fiber, as depicted in Figure a. This fiber can stretch up to a minimum of 200%. Typically, the fibers are initially extruded through a syringe and subsequently left to evaporate. During evaporation, the filament transitions into a ribbon shape, while the encapsulated liquid crystal molecules self-assemble into helices. Consequently, when these fibers are woven into clothing and stretched horizontally, they exhibit a color shift from green to blue. Notably, the light sensor designed in this manner solely modifies the color without integrating an electrical signal. Combining optical and electrical signals, the sensor can achieve a dual-signal output mode. Cellulose nanocrystals (CNCs) self-assemble into cholesteric liquid crystals in liquids slowly, with further reduction in their layer spacing via gradual evaporation. By embedding this self-assembled CNC structure in a polymeric low-eutectic solvent, Li et al. developed a color-changing hydrogel exhibiting ionic conductivity. This approach effectively overcomes the fragility of CNC-based elastomers and grants them significant stretchabilityup to 1163.7%. Color shifts progressively from red to blue as the sensor elongates from 0 to 450%. Owing to its ionic conductivity, the sensor outputs a resistance signal under tensile stress. Furthermore, when affixed to a joint, it can detect bending (Figure b). Additionally, Ma et al. introduced a cholesteric liquid crystal elastomer with ionic conductivity, which merges the elasticity of rubber with the structural color of helical nanostructures, facilitating both color and resistance changes within a 150% deformation (Figure c). The elastomer’s structure remains stable without internal water, making it suitable for human–computer interaction via an interactive system comprising the sensor and a display module. Stretch-induced color variation is not limited to liquid crystal sensors; other sensors made from photonic crystals also achieve stretching discoloration through self-assembly. For instance, Gong et al. developed an optoelectronic skin (o-skin) that integrates electronic (e-skin) and photonic skin (p-skin) properties (Figure d). This was achieved by self-assembling highly ordered laminar structures (dodecylglycerol clathrate, DGI) in a solvent with surfactant assistance. Subsequently, the PDGI/PAAM compound was synthesized by encapsulating this architecture within a polyacrylamide (PAAM) network through in situ polymerization. The layer spacing of PDGI aligns with the visible-light spectrum, allowing the optoelectronic skin to exhibit color shifts under stretching or compression. Unlike cholesteric liquid crystals, this material also demonstrates anisotropic electrical conductivity, attributable to parallel channels within the PDGI layers that facilitate rapid ion migration. Photonic signal detection, unlike electrical signals, offers advantages such as low energy loss and clear visualization. Nonetheless, achieving self-assembly of photonic crystals or liquid crystals with anisotropic conductivity remains challenging.

12.

Anisotropic sensors for stretching color change obtained by self-assembly. (a) Schematic representation of the preparation of a cholesteric liquid crystal fiber with tensile discoloration and its tensile discoloration. Reproduced with permission from ref . Copyright 2022, Springer Nature. (b) Schematic representation of the color change and sensitivity of CNC liquid crystal hydrogel during stretching. Reproduced with permission from ref . Copyright 2022, Wiley-VCH. (c) Color change in stretching, color change in squeezing, and capacitance change curves of cholesteric liquid crystal hydrogels. Reproduced with permission from ref . Copyright 2024, Royal Society of Chemistry. (d) Diagram of the extrusion and tensile color change of PDGI/PAAm and its tensile stress–strain curve in both directions. Reproduced with permission from ref . Copyright 2022, AIP Publishing LLC.

This section outlines three structural design approaches for anisotropic sensors. First, conductive materials within sensors are oriented using external field-induced methods (e.g., magnetic/flow fields), thereby endowing sensors with anisotropic electrical signal outputs during deformation. The surface functional groups of conductive fillers and their inherent 1D/2D characteristics significantly enhance mechanical properties along the orientation direction. Future research may focus on modifying filler surface groups and achieving simultaneous enhancements in conductivity and mechanical properties through in situ growth of additional materials on conductive fillers. Second, the substrate’s internal structure is engineered into anisotropic architectures. Methods such as directional freezing, prestretching, natural wood retreatment, or spinning yield transducers with anisotropic structures that typically exhibit pronounced differences in mechanical properties and directional electrical signal responses during deformation. Third, nonconductive force-induced color-change sensors rely on internal photonic crystal orientation: stretching or compression induces structural color changes due to photonic crystal rearrangement. However, few studies have reported directional conductivity differences in this category. Future work on this sensor type may explore integrating anisotropic conductivity to expand functionality.

5.3. Surface Microstructure

Designing surface microstructures for sensors involves fabricating micron-sized protrusions on the sensor surface, which is much simpler than designing anisotropic conducting networks and anisotropic structures for sensors. These microstructures not only enhance the sensor’s surface area but also improve the contact area under external forces, thereby increasing the sensor’s sensitivity to various stimuli. Additionally, microstructures offer diverse mechanical responses and expand the sensor’s detection capabilities. Methods for microstructure design encompass the template method, 3D printing, , electrospinning, , and magnetic assistance. The template method is widely adopted in flexible sensor design due to its precision and repeatability in incorporating microstructures and nanostructures This process typically entails etching microstructured and nanostructured patterns onto a silicon wafer using techniques such as laser and wet etching. A mixture of flexible materials, frequently PDMS and PU, combined with a sensitive material precursor, is then applied to the etched wafer, cured, and subsequently peeled off to fabricate tactile sensors with intricate microstructures. Microstructured tactile sensors are categorized into two types: regular geometric and bionic structures, based on their microstructural morphology.

5.3.1. Regular Geometry Microstructure Sensor

Since 2010, Bao et al. have published a series of work on microstructured sensors. Three classical surface microstructures, micropyramid, , micropillar, and microhemisphere , structures, have then become common in sensor design. These artificial microstructures of regular geometries have different properties. For example, Jiang et al. prepared highly sensitive capacitive shear force sensors with a surface morphology of micropillar arrays using the template method. As shown in Figure a, the structure of the sensor is a micropillar array sandwiched between two gold electrodes. This array ensures an initial capacitance value. When a shear force is applied to the sensor, it causes the distance (d) and overlap area (A) between the top and bottom electrodes to change. The capacitance changes (ΔC) caused by Δd increase the capacitance value, while the changes (ΔA) caused by ΔA decrease the capacitance value. When a shear force is applied to the sensor at different angles, the changes in Δd and ΔA are also different, resulting in different ΔC. Relying on this ability, the sensor is able to detect the direction and magnitude of the shear force. The sensor thus exhibits anisotropic sensitivity to shear forces from different directions, enabling angle detection over a range of 90°. However, this microcolumn sensor does not have good detection of normal forces despite its anisotropic sensitivity. Microhemispheres are often designed to detect both normal and tangential forces. When a normal force is applied to the surface of the microhemisphere sensor, it tends to compress downward. Wang et al. proposed a three-dimensional flexible force sensor, as shown in Figure b. It consists of an axisymmetric hemispherical protrusion and four quarter-circle electrodes of equal size. When pressure is applied to the surface of the tactile sensor, the contact area gradually increases as the pressure increases, leading to the enhancement of the electrical signal, which can be used for normal force detection. Moreover, when nonperpendicular force or sliding causes the sensor to produce an asymmetric change in contact area, different electrodes will produce different responses, which can be used for tangential force detection. Based on this tangential and normal force detection capability, the sensor-equipped robotic arm can provide a firm grip on the object. When gripping a bottle, the sensor detects the tangential force, then it will start gripping the bottle and detect the output voltage Voc. When Voc representing the normal force is below the threshold, the robot will move further until it is above the threshold, which indicates a correct gripping position. Moreover, when Voc representing the tangential force is found to be greater than the threshold value, it means that the bottle is not sliding and the gripping is successful. Although this design enables normal/tangential force detection, it is almost flattened under high pressure, resulting in a limited detection range and linearity.

13.

Microstructured tactile sensors with regular geometry. (a) Sensing mechanism of the initial, bending, and collapsing microcolumn dielectric layer shear force sensor and the variation of the signal curves of the sensor for different directions and magnitudes of shear force. Reproduced with permission from ref . Copyright 2024, Wiley-VCH. (b) An array of triboelectric sensors consisting of four hemispheres and its mechanism for detecting forces in different directions. Reproduced with permission from ref . Copyright 2024, American Chemical Society. (c) Performance plot of resistive tactile sensors with geometrically interlocked micropyramid, microhemisphere and microcolumn structures against normal force, tangential force, tensile, and bending signals. Reproduced with permission from ref . Copyright 2018, Springer Nature. (d) Tactile sensors with a three-layer micropyramid structure and their ability to recognize different textures. Reproduced with permission from ref . Copyright 2018, Wiley-VCH.

The use of two microhemispherical flexible sensors in a geometrically interlocked structure can effectively improve this phenomenon. In order to investigate the intrinsic relationship between microstructure and force-sensitive performance, Park et al. designed three microstructured carbon nanotube/elastomer composites, namely, cylindrical, dome (hemisphere), and pyramid, and superimposed their microstructured surfaces to form a geometrically interlocked structure. As shown in Figure c, sensors with all three interlocking structures exhibit multidirectional force detection capabilities, capable of detecting normal force, tangential force, tension, and bending. Meanwhile, after comparison, it is found that the microdome sensor has the best normal force tensile and bending detection capability, and the microcylindrical sensor has the best tangential force detection capability. In order to detect small pressure and vibration, the use of micro pyramid structure is a better choice because of its narrow top and wide bottom structure; the stress will be concentrated at the apex of the pyramid, resulting in having large structural deformation and sensitivity. Inspired by the epidermis–dermis and external microstructure of human fingerprints, Cao et al. assembled three layers of micropyramid structures together to form a tactile sensor, as shown in Figure d. The sensor has an ultralow resolution and is capable of detecting typical response curves in the range of 45–550 Pa. Because differences in item surface textures generate distinct signals, the sensor discriminates between items when gently swept over surfaces.

5.3.2. Bionic Microstructure Sensor

Given the extensive research on artificial microstructures with standard geometries, researchers are increasingly turning to turn to biological entities for fabricating biomimetic microstructures. Unlike sensors with microstructures created through inverse molding on silicon wafer surfaces, certain biologically inspired structures can be directly replicated and molded from plant epidermis. , For instance, the common lotus leaf, exhibiting myriad irregularly distributed micron-sized protrusions, not only enhance signal detection but also facilitate anisotropic sensing. Employing this technique, Wang et al. crafted microstructured tactile sensors as depicted in Figure a. These sensors exhibit no coupling effect between resistive and capacitive signals, enabling the capacitive measurement of pressure and the resistive measurement of strain. Owing to the signal amplification provided by the microstructures, the sensor adeptly differentiates among objects of varying materials and weights. The surface of reed leaves, featuring regular microgrooves and sparse micropapillae distribution, exhibits a more complex morphology than that of lotus leaf-inspired sensors. Drawing inspiration from reed surfaces, Wang et al. developed a piezoresistive pressure sensor with both a broad linear pressure range and elevated sensitivity through the employment of four stacked parallel layers mimicking reed-like structural substrates (Figure b). Owing to its distinctive anisotropy, this sensor not only detects normal force but also discerns tangential force direction. When the force direction aligns with the microstructure orientation, the angle is defined as 0°. As this angle increases toward 90°, the rate of resistance change similarly increases.

14.

Tactile sensors with bionic microstructures. (a) Structure and mode of operation of a lotus leaf microstructured tactile sensor with a geometrically interlocked structure. Reproduced with permission from ref . Copyright 2023, Wiley-VCH. (b) Schematic representation of the surface morphology of the reed leaf micropattern sensor with its ability to detect the direction of shear force. Reproduced with permission from ref . Copyright 2022, Springer Singapore. (c) Schematic of a fingerprint-structured tactile sensor and its ability to recognize fabric texture. Reproduced with permission from ref . Copyright 2022, Elsevier. (d) Schematic diagram of a Liesegang flexible hydrogel with periodic hard ridges. Reproduced with permission from ref . Copyright 2023, Wiley-VCH.

The objective of tactile sensors is to mimic the human skin’s ability to perceive objects. Consequently, some researchers have designed fingerprint textures on the surface of tactile sensors to enhance their detection capabilities. For instance, Fu et al. employed a template method to create a bimodal tactile sensor featuring a fingerprint microstructure. As depicted in Figure c, this sensor supports two output modes: capacitive and triboelectric and successfully captures both static and dynamic data. This sensor is capable of conducting sliding tests from any point on the fabric. Given the varied geometric characteristics of fabric surfaces, the vibrations resulting from the sensor–fabric interaction differ; consequently, sliding vertically or horizontally produces distinct voltage outputs, enabling effective identification of fabric type and sliding direction. Furthermore, innovative methods have been developed to construct tactile sensors with similar fingerprint-like microstructures. Wu et al., for example, utilized the classical Liesegang reaction-diffusion phenomenon to produce a highly stretchable artificial ionic skin, incorporating periodic rigid ridges with fingerprint-like patterns into a soft hydrogel (Figure d). These ridges exhibit a piezoresistive effect independent of deformation. By combining hydrogels with an elastomer featuring a fluorinated surface coating, the sensor gains the capability to detect triboelectric signals. This configuration enables operation in both piezoresistive and triboelectric modes, facilitating fine texture recognition. Its efficacy was confirmed through tests on everyday objects, where it demonstrated exceptional spatial resolution and successfully distinguished between various items, including precise identification of eight fruit types.

This subsection concludes that both designing regular geometric microstructures and mimicking plant/animal surface microstructures can enhance sensor sensitivity. Directional forces induce distinct deformations in these microstructures, yielding differential electrical signal variations that enable discrimination of force magnitude and direction. However, most microstructure sensors rely on multisub-sensor integration, which often encounters interlayer mismatch issues (e.g., mechanical/electrical property discrepancies between layers). The microstructure is also susceptible to damage during deformation, reducing service life. Therefore, solving the problems of interlayer mismatch and low service life is the focus of future research.

6. Innovative Applications

Recent years have seen a rapid advancement in wearable technology, resulting in practical application of anisotropic tactile sensors. These sensors offer significant benefits in various domains including sports training, health monitoring, robotics, and human–computer interaction. This section delves into the diverse applications of anisotropic tactile sensors, categorizing them based on their distinct functionalities.

6.1. Athletic Training

During joint motion, between 60 and 70% of the weight-bearing load is transmitted through the medial compartment. Deviations toward either valgus or varus directions influence load distribution significantly. Inappropriate exertion postures can result in increased compartmental loads, potentially heightening the risk of degenerative changes in articular cartilage and other joint structures. Employing tactile sensors at specific joints and force application points, alongside additional signal acquisition devices, enables the assessment of posture and force accuracy. This integration mitigates injury risks and enhances athletic performance. A pragmatic method for analyzing motion behavior involves designing sensor arrays with anisotropic tactile sensors. For example, to observe runners’ gait, Zhai et al. utilized a bilateral convergent directional solidification casting technique to fabricate a negative Poisson’s ratio (NPR) anisotropic porous foam (Figure a). The anisotropic structure ensures minimal lateral deformation; under uniaxial compression, the tactile sensor experiences insignificant lateral change. Therefore, integrating nine sensors into a system ensures that compressing one sensor does not lead to contact among adjacent ones, thereby eliminating electrical crosstalk. A sensor array insole can monitor pressure distribution and gait nuances, distributing force uniformly across each subsensor during a complete foot strike and localizing pressure when landing on toes or heels. Through thoughtful design, the insole attains intricate functionality. For instance, Dai et al. developed a muscle-like anisotropic conductive composite comprising PDMS, carbon black (CB), and TPU foam (PCTF) through directional freezing. When aligned horizontally, the sensor’s pores allow for notable compressibility and resilience, functioning effectively as a piezoresistive sensor. In this orientation, the sensor-equipped insole facilitates gait analysis (Figure b). Conversely, when pores are arranged vertically, the sensor exhibits superior compression resistance. By integrating negatively charged PCTF with positively charged Ecoflex, a triboelectric sensor is formed. This configuration not only provides excellent compressive resistance and swift response times but also enables the sensor to accurately detect supination and pronation during running activities (Figure c). Each subsensor produces consistent voltages during a normal running gait, whereas varied signals are generated during phases of supination and pronation. For more intricate movements, such as those in shooting practice, anisotropic sensors capable of monitoring both slippage and pressure at joints are crucial for evaluating the appropriateness of posture and force. Bao et al. developed a multidimensional sensor inspired by the structure of human muscles, integrating three subsensors. In this system, an MXene-embedded ZnO nanowire array on an isotropic gradient-wrinkled polyurethane film functions as a strain-insensitive pressure subsensor, while two orthogonally stacked aligned segments of polyimide/polyurethane film serve as pressure-insensitive anisotropic strain subsensors. This unique anisotropic structure enables the sensor to identify and quantify forces along the X, Y, and Z axes by analyzing electrical signals from the three subsensors. When applied to shooting motions, differences in subsensor strain correlate with variations in posture, facilitating the assessment and correction of shooting posture based on signal patterns (Figure d). To simultaneously monitor movements and correct postures in badminton players, Liu et al. employed a one-step 3D printing technique to create a layered carbon nanofiber (CNF)/PDMS nanocomposite strain sensor. During fabrication, the shear flow from the printing process aligns the CNFs into a network that forms uniform microcracks under stretching, allowing the sensor to exhibit anisotropic responses to external deformations. Three sensors were placed on the left, center, and right sides of the wrist, with the circumferential and axial directions designated as the X-axes and Y-axes, respectively. These sensors generate distinct signals corresponding to specific movements. For instance, during a proper grip, all sensors deform nearly simultaneously within a 1 to 1.5 s interval, with greater resistance changes along the Y-axis compared to the X-axis. Conversely, incorrect postures produce inconsistent signal timings (Figure e). Furthermore, striking postures generate varied sensor signals, providing additional data for motion analysis (Figure f).

15.

Anisotropic tactile sensors for sports training: (a) Structures and arrays of anisotropic porous foams and arrays comprising them for gait monitoring. Reproduced with permission from ref . Copyright 2023, Elsevier. (b) Principles and implementation of PCTF arrays for gait monitoring. Reproduced with permission from ref . Copyright 2024, Elsevier. (c) Principles and implementation of PCTF arrays for detecting external and internal rotation of the foot. Reproduced with permission from ref . Copyright 2024, Elsevier. (d) Current signal diagram for the detection of shooting postures by a tactile sensor system with three anisotropic subsensors. Reproduced with permission from ref . Copyright 2024, Wiley-VCH. (e) Anisotropic strain sensors are useful for detecting the ability to hold a racket correctly and (f) grip detection for hitting a badminton ball. Reproduced with permission from ref . Copyright 2024, Wiley-VCH.

6.2. Health Surveillance

Anisotropic tactile sensors integrated with smart devices can monitor human vital signs such as pulse and respiration. Their unique structural design enhances tactile sensing capabilities, offering enhanced sensitivity due to their anisotropic nature. , Chen et al. developed a hydrogel with a biomimetic anisotropic structure using polypyrrole-dopamine/poly­(vinyl alcohol) (DA-PPy/PVA) through a directional freeze–thaw strategy. This hydrogel, characterized by excellent electrical conductivity and an oriented structure, monitors electrocardiogram (ECG) and electromyography (EMG) signals. As shown in Figure a, the DA-PPy/PVA hydrogel exhibited higher PQRST waveform strength and a superior signal-to-noise ratio compared to Ag/AgCl electrodes. Additionally, the sensor effectively captured EMG signals. Beyond signal detection, these sensors also play a vital role in in joint health protection, such as monitoring potential joint inflammation. Hong et al. designed a high anisotropic piezoelectric network composite (HAPNC) sensor based on a kirigami structure, which enhances piezoelectric phase connectivity and piezoelectric performance, providing high-dimensional anisotropy. This structural anisotropy enables the sensor to measure planar stress in any direction and determine both its magnitude and direction. The dimensional extension of HAPNC sensors facilitates the collection of joint motion data. A monitoring and alert system integrated with smart devices was developed to track neck and shoulder joint movements and provide alerts for prolonged sedentary behavior (Figure b). Advancements in medical technology emphasize the need for a comprehensive analysis of pulse wave signals, including their periodicity, wave width, heart rate, and waveform characteristics. A 3D dynamic simulation of pulse waves has emerged as a promising solution. By combining tactile sensors, 3D dynamic simulations, and advanced technologies like computer vision, medical robots can be developed for detailed patient assessments. Tian et al. created a flexible, high-resolution tactile sensor array based on a pressure-sensitive tunneling mechanism. The pyramidal microstructure generates additional conductive channels and synergizes with an HfO₂ tunneling layer, achieving high sensitivity, a wide measurement range, and excellent spatial-temporal resolution. Furthermore, the noise suppression properties of the HfO₂ layer provide a high signal-to-noise ratio, enabling the detection of weak signals. As illustrated in Figure c, Tian et al. utilized this sensor to construct a 3D dynamic pulse map, capturing the spatial distribution of pressure. This system successfully mapped pulses across various application scenarios, including different acupoints (CUN, GUAN, CHI) and diagnostic pressure levels (FU, ZHONG, CHEN).

16.

Anisotropic tactile sensors for health surveillance. (a) Changes in electrical signals produced by tactile sensors in response to throat vibration and phonation. Reproduced with permission from ref . Copyright 2024, Elsevier. (b) Triboelectric tactile sensors for the monitoring of joint movements of the shoulder and neck. Reproduced with permission from ref . Copyright 2021, AAAS. (c) Three-dimensional imaging maps and pulse signal of volunteer IV before and after exercise. Reproduced with permission from ref . Copyright 2024, Wiley-VCH.

6.3. Object Recognition

In domains such as daily life, industry, and agriculture, many tasks involving repetitive human actions, such as food preparation, heavy lifting, and fruit and vegetable harvesting, are efficiently performed by machines. Over the years, researchers have developed numerous artificial intelligence devices and robotic systems capable of recognizing and handling objects. These robots employ combination of tactile and optical sensors to accurately identify objects. While optical sensors are commonly used, tactile sensors offer distinct advantages through precise detection of an object’s weight and texture, enabling the differentiation of object types. As a result, there has been growing interest in tactile sensors with three-dimensional tactile perception. , Various smart flexible sensor devices, including robotic hands and smart gloves, have been proposed to enhance object recognition. For instance, Yuan et al. developed a novel ultra-lightweight multifunctional tactile aerogel sensor (MTAS). This sensor, fabricated from layered graphene with a wavy ordered structure created through a unique directional freezing and annealing, exhibits excellent piezoresistive properties. The wavy graphene layers come into contact under pressure, and due to graphene’s high thermal conductivity, the integrated tactile system, when combined with a neural network, detects pressure, temperature, and object types with an accuracy of 94.63% in kitchen scenarios involving common food items. As shown in Figure a, the sensor successfully distinguishes 18 types of vegetables, fruits, and meats, producing distinct electrical signals for each. Additionally, inspired by the bionic design of human skin hair, Chen et al. developed a flexible tactile sensor based on magnetic cilia arrays (Figure b). When deflected by an external force, the magnetic cilia arrays detect changes in the magnetic field, enabling the measurement of both the magnitude and direction of the force. This sensor achieves a resolution of 0.2 mN, operates within a range of 0–19.5 mN, and can accurately determine the direction of external forces. Training a machine learning algorithm on the signals produced by a tactile sensor in contact with seven different items results in a 97% recognition accuracy for the intelligent system (Figure c). In other work, Shen et al. developed a soft tactile sensor featuring self-decoupling and super-resolution capabilities, constructed with a sinusoidally magnetized flexible film just 0.5 mm thick (Figure d). The application of external force changes the magnetic flux beneath the film, which, detected by the embedded Hall sensors in the intermediate layer, is converted into an electrical signal by the underlying hard printed circuit board. This allows for precise measurement of both normal and shear forces. When deployed on a robot handling an egg, this sensor precisely indicates the necessary grip by evaluating the egg’s friction and required clamping force. Integrating robots with a human-like somatosensory network is critical for enabling them to adapt their grip based on the object’s softness and texture. Qiu et al. created a bimodal tactile sensor capable of three-dimensional force detection. The sensor integrates piezoelectric and piezoresistive layers with a unique raised surface in the form of a prism. Due to this structure, the sensor has a multidimensional force detection capability (Figure e). This design mimics human skin’s ability to sense multidimensional stimuli, static and dynamic, while measuring softness and texture. The sensor also identifies eight materials types, as shown in Figure f, by distinguishing the coefficients of friction of different materials. As shown in Figure g, this 3D force sensor deciphers pressing angles (15°) and adapts accurately to categorize softness, enabling the secure grip of fresh grapes with a 0.5 N force. Additionally, when applied to identifying the ripeness of white strawberriesa task difficult even with visual assessmentthe robot equipped with this sensor attained an 82.18% success rate in evaluating the softness and ripeness of 101 strawberries (Figure h).

17.

Anisotropic tactile sensors for object recognition. (a) Structure of an aerogel sensor and its ability to recognize objects. Reproduced with permission from ref . Copyright 2023, Springer Singapore. (b) Structure of a magnetic tactile sensor modeled on human cilia and (c) its ability to recognize different objects. Reproduced with permission from ref . Copyright 2024, Wiley-VCH. (d) Working principle of a magnetic sensor film and its detection of egg friction. Reproduced with permission from ref . Copyright 2021, AAAS. (e) Schematic diagram of a tactile sensor with three-dimensional force detection capability. (f) Comparison of friction coefficients of sensors after exposure to eight substances. (g) Comparative experiments of unidirectional force and multidirectional force sensing feedback based robotic arm gripping grapes. (h) Process of picking strawberries by a sensor-wearing robot. Reproduced with permission from ref . Copyright 2024, AAAS.

6.4. Human–Machine Interaction

Human–machine interaction (HMI) refers to the interaction and communication between humans and systems, including machines, computer systems, and software. Users typically input commands through peripherals such as keyboards and mice, while computers provide feedback via screens and speakers. , Traditionally, HMI has relied on visual sensors to capture human motion signals for interaction but has rarely utilized the command potential of human skin. The use of anisotropic tactile sensors, which can be attached to joints or fingers, introduces a novel approach that converts electrical signals generated during movement into computer commands, thereby enhancing the HMI experience. , For example, Kim et al. designed a highly sensitive multidirectional sensor composed of aligned carbon nanotube (CNT) arrays with functionally distinct anisotropic layers (Figure a). This sensor exhibits significant structural anisotropy, in which transversely aligned CNTs bridge the longitudinal microcracks in the lower layer. Such a design prevents expansion in unintended directions when the sensor is highly stretched, offering exceptional sensitivity across a wide linear operating range of up to 100% strain, with a strain coefficient of 287.6. When stretched over a joint, the normalized resistance change (ΔR/R ₀) along the X-axis increases with bending, while that along the Y-axis change remains nearly constant. The sensor demonstrates exceptional performance in gesture recognition by converting gesture signals into electrical signals that can be translated into 3D character model animations via an intelligent system, thereby realizing HMI. Furthermore, Shen et al. developed a full-skin bionic (FSB) inspired by the structure of human millimeter hair and the layered composition of epidermis, dermis, and subcutaneous tissue. The FSB e-skin features a dual interlocking lamellar micron-sized cone structure combined with a supercapacitive ionic-electronic effect, achieving an ultrahigh sensitivity of 8053.1 kPa^–1 (<1 kPa), a linear sensitivity of 3103.5 kPa^–1 (1–34 kPa), and a rapid response/recovery time of <5.6 ms. For speech-impaired individuals, interpreting thoughts expressed through sign language can be difficult for the untrained. In this context, tactile sensors have proven invaluable for HMI when paired with robotic hands outputting information. By training a sensor array using neural networks on sign language signals, the intelligent system was able to recognize American Sign Language with high accuracy (Figure b). After repetitive gestures, 1200 data sets were collected, with 1080 used for training, resulting in a gesture recognition rate of 90.83% for the FSB. When worn on the palm, the e-skin enables a connected robotic hand to mirror the wearer’s gestures. Additionally, Ma et al. developed an anisotropic strain transducer (LCT) by combining conductive ink and liquid metal. The LCT exhibited maximum strain coefficients of 158.9 and 1.9 in the conductive ink and liquid metal directions, respectively, while maintaining dynamic stability, frequency independence, and durability over 3000 cycles. Applying different strains to the LCT alters its relative resistance, enabling precise control over the position of a target object. When stretching the sensor along the x-axis, only the horizontal sensor deforms significantly, causing the target object to move along the horizontal. In contrast, stretching the sensor along the y-axis, only the vertical sensor produces significant deformation, causing the target object to move in the vertical direction (Figure c).

18.

Anisotropic tactile sensors for human–machine interaction: (a) structure and control circuit system of the sensors, and the application in human-machine interactions. Reproduced with permission from ref . Copyright 2021, Springer Singapore. (b) Control system, software of the e-skin, and the electrical signals corresponding to different actions. Reproduced with permission from ref . Copyright 2022, Wiley-VCH. (c) Two sensors stacked on top of each other, and the application in human-machine interactions. Reproduced with permission from ref . Copyright 2024, Elsevier.

7. Conclusions

To enhance the development of advanced detection capabilities in electronic skin, significant emphasis has been placed on constructing anisotropic flexible tactile sensors that can detect forces from multiple directions. Progress in this area has been accelerated by innovative design concepts and the introduction of new materials. Substantial research continues to broaden the design possibilities for anisotropic tactile sensors. This review offers a comprehensive examination of the evolution and current state of anisotropic flexible tactile sensors over the past decade. Initially, the paper categorizes and evaluates various types of flexible sensors, including resistive, capacitive, piezoelectric, triboelectric, magnetic, and optical sensors. Key performance metrics include flexibility, sensitivity, linearity, operational range, drift, response/recovery time, stability, resolution, and multidirectional force detection capability. The review also examines common substrate materials and conductive fillers used in anisotropic tactile sensor design, such as organic substrates, metallic fillers, and carbon-based fillers. Additionally, the paper outlines design strategies for anisotropic structures in different sensor types, including integrated sensors and sensor arrays, internal orientation structural design, and surface microstructural design, demonstrating anisotropic functionalities. Finally, the paper explores the intelligent applications of anisotropic tactile sensors in fields such as sports training, health monitoring, object recognition, and human–computer interaction. Although numerous anisotropic tactile sensors have been created, research in this field is still relatively nascent, and progress is slow, primarily due to limited design approaches for anisotropic structures and insufficient recognition of their significance.

The development and application of anisotropic tactile sensors continue to encounter several challenges. These include issues related to materials and synthesis methods, the ability to detect forces from multiple directions, and the fabrication of smart devices as well as algorithm design. The design of future flexible anisotropic tactile sensors must take into account a range of factors:

• 1.Materials: The choice of biosolvents and conductive fillers is a critical factor in the design and synthesis of flexible tactile sensors. Materials such as vinyl amide (classified as a carcinogen) and carbon nanotubes present partial toxicity concerns. Residual organic compounds from the synthesis process and inadequate encapsulation can pose health risks in applications such as medical devices, motion detection, and human–computer interaction. To address these issues, the synthesis of tactile sensors should prioritize biocompatible and flexible materials. Prolonged use of tactile sensors may lead to sweating and wound infections, highlighting the importance of breathable and comfortable sensor designs. While strategies such as structural innovations and fabric-based sensors are common, further advancements in material development are required. Additionally, sensors should be environmentally sustainable, with long-term devices incorporating extended life cycles or self-healing properties and short-term devices utilizing biodegradable materials. Future efforts should prioritize the discovery of novel materials that are biocompatible, breathable, comfortable, and eco-friendly for tactile sensor applications. In addition, selecting materials that are compatible with industrial integration and large-scale demand is also a key focus of future research. Current research is generally concentrated on graphene, carbon nanotubes, MXene, and liquid metals, but these conductive materials are generally quite expensive. In the future, these expensive conductive materials can be combined with relatively inexpensive silver paste and carbon black to reduce costs.
• 2.Multidirectional Force Detection Capability: This paper reviews various anisotropic tactile sensors, highlighting their ability to exhibit significant sensitivity differences across different directions and unique mechanical decoupling mechanisms. However, most sensors are limited to detecting signals in two directions and cannot effectively sense complex three-dimensional forces. While anisotropic tactile sensors often demonstrate enhanced sensitivity to pressure magnitude and the direction of shear force motion, some rely on arrays to determine motion direction. Microstructured anisotropic tactile sensors can measure normal and shear force magnitudes but often compromise sensitivity for stability and durability. To improve multidirectional force detection, it is crucial to develop sensors with high sensitivity in all directions, distinct directional sensitivity, or explore more robust structures and wear-resistant materials to enhance durability and sensitivity.
• 3.Algorithm and Control: The integration of sensors with artificial intelligence (AI) has become increasingly prevalent. Large data sets collected from sensors during activities such as object contact, gesture perception, texture perception, and object recognition are analyzed using machine learning to extract diverse information from various signals. However, the effectiveness of AI algorithms depends heavily on the quality and quantity of data. Flexible tactile sensors face challenges in rapidly gathering large data sets and maintaining consistent performance after prolonged training. Developing comprehensive databases and advanced algorithms can reduce dependence on extensive training. Additionally, while combining AI with machines to create intelligent devices capable of detecting diverse signals is a common practice, many current AI devices lack appropriate sizing and compatibility. Future sensor designs should prioritize integrated circuit design and control to develop lightweight, comfortable, and efficient smart devices with tactile sensors.
• 4.Industrial Scalability: Applications for object recognition and friction force recognition can not only be used in miniaturized devices but also expanded into industrial environments to broaden their scope. However, in current industrial applications, sensors suitable for heavy-load handling are scarce, and rigid sensors are often used instead. Additionally, these operations require manual intervention, failing to meet future demands for unmanned and intelligent factories. For heavy-load gripping, flexible sensors must possess the ability to recover after undergoing significant deformation. Designing anisotropic tactile sensors with a high Young’s modulus can meet the requirements for detection capability and good recovery. For example, some elastomers feature internal microstructures and ultrahigh-strength polyurethane. For miniaturized applications such as feeding small raw material quantities, selecting sensors with multiple electrical signal output capabilities is more appropriate. For instance, combining triboelectric and piezoresistive sensors, which can identify raw materials and select appropriate gripping forces, significantly reduces feeding errors. In the future, selecting tactile sensors with different Young’s moduli and sensing modes tailored for specific application scenarios can further reduce manual labor.

In summary, flexible anisotropic tactile sensors remain in the early stages of development, with significant progress required to achieve practical applications. Future research should focus on designing sensors that combine high sensitivity, stability, low noise, durability, ease of production, and a wide detection range. Integrating these sensors with AI enables developing lighter, more efficient, and user-friendly smart devices. As AI technology continues to advance, anisotropic tactile sensors hold significant potential for applications in healthcare, human–machine interaction, augmented and virtual reality (AR/VR) technologies, and motion monitoring.

The authors declare no competing financial interest.