Dahiya, R. (2019) E-skin: from humanoids to humans. Proceedings of the IEEE, 107(2), pp. 247-252. (doi:10.1109/JPROC.2018.2890729).

This is the author’s final accepted version.

There may be differences between this version and the published version. You are advised to consult the publisher’s version if you wish to cite from it.

http://eprints.gla.ac.uk/183481/

Deposited on: 05 April 2019

Enlighten – Research publications by members of the University of Glasgow http://eprints.gla.ac.uk

> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 1

E-Skin: From Humanoids to Humans BY RAVINDER DAHIYA BEST Group, School of Engineering, University of Glasgow, UK

I. WHAT IS E- SKIN With robots starting to enter our lives feedback as well as decoding user’s tears) or the physiological parameters in a number of ways (e.g. social, intentions in real time), (e) high- (e.g. pulse rate, blood pressure etc.) in assistive, and surgery, etc.) the electronic performance (e.g. fast response) low- real-time [7, 8] For health and medical skin (e-skin) is increasingly becoming power electronics, and (f) sufficient applications there are additional important. The capability of detecting energy for operation of touch sensors and requirements such as substrates should subtle pressure or temperature changes, associated electronics, particularly for be disposable, dissolvable, bioresorbable makes the e-skin an essential component autonomous robots. With these features, and biocompatible etc. [9, 10]. Such e- of a robot’s body or an artificial limb [1, the e-skin has been sought to resemble skin patches could either be placed 2]. This is because the tactile feedback the human skin. directly on body surface or on daily enabled by e-skin plays a fundamental With this background, the e-skin can wearables such as clothes, watch or role in providing action-related be defined as a multisensory patch or jewellery etc. In fact, in the latter case, information such as slip during system (Fig. 1), having a set of sensors much wider variety of sensors can be manipulation/control tasks such as (e.g. touch, temperature, gas, display, integrated on e-skin patch. The face grasping, and estimation of contact energy scavengers, and electrochemical masks with gas sensor, for example. parameters (e.g. force, soft contact, sensors etc.) and the associated Fig. 2 shows some of the scenarios and hardness, texture, temperature etc. electronics either integrated on flexible/ use cases for e-skin in robotics, during exploration [3]). It is critical for bendable/stretchable substrates or prosthetics, wearables and health the safe robotic interaction – be it as a co- embedded into soft substrates [6]. The monitoring [11-13]. The e-skin, as worker in the futuristic industry 4.0 parameters to be measured, and hence the described above, could enable advances setting or to assist the elderly at home. types of sensors, depend on the target in areas such as electronics In context with robotics and application. For example, in the case of manufacturing, mobile health and prosthetics, the e-skin is also referred to robots and prosthesis, human-like tactile robotics etc. and will open interesting as tactile skin. To be an effective feelings can be attained by measuring avenues for applications ranging from component of a robot’s body, the e-skin sensing parameters such as contact force, wearable systems for individual-centric should have a complex mix of functional temperature, pain and slip, etc. With health monitoring or self-health and morphological features such as: (a) these sensors the e-skin can advance the management, to instrumented smart multiple types of sensors distributed over capabilities of robots by allowing them to objects for internet of things and surgical large areas (entire body) to measure exploit area contacts and complement or tools etc. New applications could also multiple touch sensing parameters (e.g. replace internal sensors (e.g. emerge because of the new form of human skin has about 45K simultaneous use of tactile and electronics needed for the development touch/pressure sensitive receptors in ~1.5 proprioceptive sensing). A rudimentary of e-skin. m2 area) [4, 5]; (b) appropriate placement illustration of e-skin can be seen in touch of sensors to obtain varying degree of screens in wide use today. Touch screens sensitivities over the entire body; (c) detect contact location in the sensors (and associated electronics) manner of a simple switch, i.e., integrated on or embedded into soft and ‘contact’ or ‘no contact.’ stretchable substrates to allow When applied to healthcare conformability to 3D surfaces (for applications such as real time superior object handing, improved user monitoring of chronic diseases, comfort and reliable data acquisition); the e-skin (also referred to as (d) capability to handle large data “second-skin”) should have generated by the sensors through local sensors to measure the processing or neural computing and variations in the composition of extracting useful information (e.g. analytes or biomarkers present Figure 1: The e-skin concept with multiple functionalities collecting the data for critical tactile in the body fluids (e.g. sweat or integrated over ultra-thin flexible substrates.

> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 2

materials, and manufacturing methods are currently the areas of intensive investigation. This is also evident from the recent surge in the number of publications (Figure 3) related to e-skin and flexible electronics. The major technological approaches that are being explored for the development of e-skin are briefly described below. A. E-Skin with Off-the-Shelf Components To meet the immediate need of tactile feedback in various robotic tasks, various e-skin alternatives have been explored, with off-the-shelf sensing/electronic components soldered on to flexible printed circuit boards (FPCB) [11] or

Figure 2: The need to cover 3D surface of robot’s body (a-b) and prosthetic limbs (c) with large number of stitched to the flexible surfaces [14]. In touch sensor has been key driver application for flexible and conformable e-skin. The e-skin is now being these cases, the e-skin is made of used to develop wearable patches (or ‘second skin’) with sensors for health monitoring (d). mechanically integrated but otherwise magnetic etc.) [3, 5, 11]. A large number distinct and stiff sub-circuit islands of II. TECHNOLOGICAL EVOLUTION of experimental devices and prototypes off-the-shelf components connected with Technological advances have reported in the literature show a good wires or stretchable metal interconnects. motivated numerous multidisciplinary diversity among the types of sensors that This approach has been explored by investigations leading to the were developed in the 1980s. Particular various research groups. These semi- development of artificial organs such as attention was given to the development rigid FPCB based skin patches conform electronic nose, ear and bionic eyes. of tactile sensing arrays for the object to surfaces with large curvature such as Development of these artificial organs recognition. The creation of multi finger arms of a humanoid robots (Fig 2) and was challenging, but they also have the robotic hands, in late 1980s, increased have served some of the urgent needs the interest in tactile sensing for robotic advantage in terms of their centralized such as tactile feedback from whole body manipulation and the works utilizing locations (i.e. unlike skin they are not or large parts of the body. In fact, such tactile sensing in real-time control of distributed over the whole body). large area implementations of e-skin manipulation started to appear [14]. Further, the number of sensory have changed the robotics research parameters they are required to acquire Likewise, the multi finger prosthetics hands, in 1990s, increased the interest in paradigm - from hand-based are much lower than the skin. As a result, manipulation to exploiting multiple it was possible to use a single tactile feedback, although the major focus was on methods such as EMG contact points or areas contact to plan technological solution to develop these and execute robotic tasks/movements. artificial organs. For example, CMOS based control, and this continues till date. Further improvements in the FPCB or (Complementary Metal Oxide These early solutions for tactile sensing stitching based e-skin can be made by Semiconductor) imagers for high- are nowhere close to the complexity of e- including more functional components performance cameras and bionic eye. To skin system defined earlier. The a greater extent better understanding of anticipated use of robotics in tasks such such as local memory. However, any the centralized sensory modalities such as human-robot working side by side in new addition of non-bendable off-the- as vision in humans also contributed to the emerging Industry 4.0 setting, shelf components would severely the successful development of artificial exploiting whole body contacts for tasks constrain the e-skin in terms of organs such as electronic nose, ear and in an unstructured environment, and bendability or conformability. bionic eyes. On the contrary, the details assistive/rehabilitation tasks etc., has about the working of human sense of brought to the fore the importance of touch are still emerging and being large area tactile skin i.e. over the whole debated [5]. Nonetheless, the advances in body of robotic/prosthetic limbs [2]. electronics technology are helping to Likewise, using e-skin as ‘second skin’ bridge the gap. in human health monitoring is opening From historical perspective, tactile new opportunities (Figure 2). These new sensing began to develop in the 1970s scenarios demand e-skin to have several and the early works focused on the functional elements in addition to the developing sensors by exploring various challenging requirement of being transduction mechanisms (e.g. resistive, conformable, stretchable and light capacitive, optical, piezoelectric, weight etc. Accordingly, new designs, > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 3

The use of FPCB based approach can electronics, such as silicon (Si), are not neuron-like processing [4], it will be also be seen in the application such as flexible and often devices from them possible to obtain next generation health monitoring and wearable systems. require high-temperature processing neuromorphic e-skin or tactile skin. For example, FPCB based wearable e- steps. In this regard, the innovative The e-skin with printed high-mobility skin like patches to monitor chronic methods such as transfer printing and semiconductor nanowires is likely to conditions such as diabetes [8, 15]. A contact printing are attractive as they to lead us to the high-performance flexible wide variety of wearable gadgets for help overcome such issues related to electronics at low-fabrication cost. With wellness applications are also available conventional electronics technologies. this unique combination of high- in market today and the use of FPCB Transfer printing involves picking a performance and the low-cost fabrication based sensing patches can also be seen in set of basic building blocks, such as this transfer printing and contact printing fashion. nanowires and ribbons from Si, approach offers an attractive alternative The additive manufacturing explored Graphene etc., from a mother or growth to organic semiconductor-based recently for 3D PCBs could offer substrate to the flexible receiver approach [24, 25]. advances in the FPCB based approach. substrates using elastomeric stamp such C. Ultra-thin Flexible Chips For example, printed sensors embedded as PDMS. Transfer printed micro- in 3D printed artificial limbs will lead to /nanostructures form an electronic grade The relevance of ultra-thin chips robust and cost-effective robotic ultra-thin layer on receiver substrates, (UTCs) to e-skin is in context with the platforms with inherent tactile sensing which is ultimately used to develop the bridging technology needed to realize a (or structurally integrated sensing), active/passive electronic components. fully flexible high-performance system. which is also free from the traditional With this approach the high-temperature For example, with compact electronics wear and tear issue associated with e- processing steps (e.g. high-quality oxide on flexible chips, the UTCs (thickness skin. However, for this the 3D printing deposition etc.) are carried out while the <50 µm) offer solution for efficient techniques must evolve from current micro-/nanowires are still on the wafer driving or output unit, sensor readout and single material (either metal, polymers and the remaining low-temperature signal conditioning and on-site etc.) printing to simultaneous processing steps such as metallization processing of the raw tactile data multimaterial printing. Some recent are carried out after completion of gathered from e-skin. The UTC based works are already hinting these advances transfer process [19, 20]. Among several approach can also be followed to develop in 3D printing [16]. When it comes to e-skin patches that have been developed skin patches for body parts requiring artificial limbs, new designs will also be using this approach, a few notable ones high-density of touch sensors. For needed to allow routing of wires etc. are conformable tattoo like e-skin example, POSFET (Piezoelectric Oxide within the 3D printed structures. In fact, patches for health monitoring [13], Semiconductor Field Effect Transistor) the approach can be extended to any 3D thermoelectric energy harvesting [21] devices-based touch sensors and flexible structure where inherent sensing is and graphene based transparent energy chip [26, 27] for body parts such as desired and can open new avenues for autonomous e-skin [12]. fingertips. Further, owing to the electronics packaging and The contact-printing involves the bendability and excellent form factor it is manufacturing. directional sliding of a donor substrate, easy to integrate the UTCs on flexible consisting of free-standing nanowires, on substrates in comparison with the B. E-Skin with printed top of a receiver substrate. During the conventional thick chips and hence they Electronic/Sensing Components sliding step, the nanowires tend to be could be used to advance the FPCB based Another approach for obtaining e-skin aligned and combed due to the sliding components or patch involves printing of 12000 shear force. Then, nanostructures are Flexible electronics E-skin for wearable electronics and healthcare sensors on flexible substrates[17, 18]. In 10000 detached from the donor substrate due to E-skin for robotics fact, printing is widely explored for next the accumulation of structural strain, and 8000 generation electronics. Several tactile finally are anchored to the receiver 6000 skin patches with printed (e.g. screen substrate by the Van der Waals printing, inkjet printing etc.) sensors interactions. Unlike transfer printing, the 4000 using composites such as conductive

contact printing does not require the publications of Number 2000 fillers in PDMS matrix etc. have been elastomeric transfer substrate. The reported [17]. In these cases, the method has been used to develop flexible 0 sensitive materials are directly printed on e-skin patches using nanowires made 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 Year of publication flexible and soft substrates. from both the bottom-up [22] and top- The sensors readout on e-skin as well Figure 3: These e-skin applications have also down approaches [23]. Owing to contributed significantly to the growth of the field as some sensing devices (e.g. transistors miniaturized dimensions the micro- of flexible electronics as evident from rapidly increasing number of publications on e-skin in or solid-state sensors) require high- /nanowires are highly flexible and the e- performance electronics on flexible and robotics and health monitoring. The trend matches skin with this approach is highly flexible. with the growth in the field of flexible electronics. conformable surfaces. The conventional Extending this approach with multiple The data for these plots was taken from Web of materials for high-performance Science by using relevant keywords (e.g. tactile gate Si nanowire, demonstrated for skin, e-Skin, flexible electronics etc.). > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 4 large area tactile skin discussed earlier. E. Neural e-Skin like algorithms could be better Further, due to reduced package volume Efficient ways are needed to process alternative. A few hardware and lower parasitic capacitance, they the sensory data, especially in the case of neuromorphic implementations reported have better high-frequency large area e-skin where large number of in literature, although not in context with performances, lower power consumption touch sensors are needed. As an example, e-skin, use spin-logic [34], memristor, and stable electronic response for a a human inspired e-skin will require neuron MOSFET, analog circuit-based particular bending state [28]. They hold about 45K mechanoreceptors in ~1.5 m2 neurons, and field programmable gate the potential to open up new avenues for area [4, 5]. With whole body tactile array etc. Although none of these have heterogeneous integration of various sensing, the tactile data will increase made major impact on robotic e-skin, semiconductor materials (e.g. organic rapidly and so will be the challenges they could be potential alternative and inorganic) based electronics, which related to compilation and processing. despite challenges in terms of is characteristic of an e-skin system. Currently, limited solutions are available complexity, scalability, speed, With these features UTCs can underpin to deal with large data generated in tactile reliability, non-bendability, power advances in several emerging skin, let alone for the resulting consumption etc. The neural nanowire applications such as robotics, wearable perception. field effect transistors-based approach is systems, m-Health, smart cities, Internet New techniques for management of another option for hardware neural of Things, body area network, body-dust, the tactile data will add significant value network (HNN) [4]. and neural interfaces, etc. to the e-skin research. One option is to III. CHALLENGES AND OPPORTUNITIES D. Energy Autonomous e-Skin develop e-skin with local processing of sensory data i.e. instead of sensing all Energy autonomy is key to the next With technological advances over the raw data to central unit for processing generation portable and wearable past few years, particularly in the field of and decision making, sending only the flexible and soft electronics, we have systems for several applications. The partially processed data and allowing high density of multiple types of gotten closer to mimicking some of the central unit to take higher level decisions. electronic components (e.g. sensors, abilities and morphology of real skin The neuroscience studies on human with sensors and electronics embedded actuators, electronics, etc.) required in e- touch sensing suggests that such on-site skin, and the need to power them without in soft substrates. However, just copying distributed computing may be occurring adding heavy batteries, have fuelled the skin morphology or capturing few in humans. For example, the ensemble of development of compact flexible energy parameters that we experience as touch is tactile data from peripheral neurons is systems to realize self-powered or not enough. The challenge lies in considered to indicate both the contact energy-autonomous e-skin. The compact reproducing the functions rather than the force and its direction. Therefore, a and wearable energy systems consisting shape while accepting the fact that the neuron-like inference to handle the of energy harvesters, energy storage shape (morphology), at the micro and tactile data early on can be helpful. A few devices, low-power electronics and macro levels, affects the functions. works have recently focused on such a efficient/wireless power transfer-based Therefore, we must focus on the processing at sensor level to detect force, functionalities as well and in this regard, technologies, are expected to pain etc. [30-32]. Such approaches must revolutionize the market for wearable there is also a need to find the ways to be scaled up for large area e-skin, even if systems and in particular for e-skin. We extract the information from sensor data. this is not an easy task. A meaningful e-skin, for a broad range have discussed in a recent review article The software based neural networks [29], a wide range of solutions such as of applications discussed here, requires a (NN) approaches have also been light-weight e-skin with wearable energy holistic approach starting from the way explored for tasks such as object harvesters, (e.g. photovoltaics, thermo- data is acquired, encoded and eventually recognition via texture or materials [33]. electric, piezoelectric and triboelectric, acted upon. The way ahead lies in However, the software-based approach etc.) and energy storage devices (e.g. multidisciplinary team-work: on one still requires all the data to be transferred flexible batteries, supercapacitors, etc.). side, neuroscientists and clinicians to the central processing unit. The NN for Among various energy harvesters the analyzing the multi-technological bases inference from data gathered by e-skin photovoltaics generate the most and has tactile encoding or clinical validity of could work for health monitoring of good potential for energy autonomous e- sensors on e-skin, on the other, engineers population in a region. The application of and technologists synthesizing the skin – as demonstrated recently through software-based approach for e-skin in graphene based transparent e-skin artificial systems, not only as “living” robotics require attention as the time integrated on photovoltaic cells [12]. proof-of-concept but also scaling up for between data gathering and reflex action real use. Of course, there are several Wireless powering is another attractive could be very short. The software-based alternative, which is possibly more challenges in the way. Major challenges NN approaches are slower and less suitable for e-skin application in health related to comfortability, signal energy-efficient due to the lack of large- monitoring as user comfort is important. acquisition and transmission and energy scale parallel processing. Instead the autonomy etc. have been discussed in hardware implemented neuromorphic previous sections along with the way tactile data processing along with NNs technology is being developed to > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 5 overcome some of them. Since various Energy Harvesters," Adv. Mater., vol. 26, pp. [20] R. Dahiya, et al., "PDMS residues-free technologies that are being used to 149-162, 2014. micro/macrostructures on flexible [2] R. Dahiya, et al., "Developing Electronic substrates," Microelectr. Engg., vol. 136, pp. develop e-skin are themselves at various Skin with the Sense of Touch," Information 57-62, 2015. stages of development, it may take Display, vol. 31, p. 5, 2015. [21] S. Khan, et al., "Flexible thermoelectric couple more years for widespread use of [3] R. S. Dahiya, et al., "Tactile Sensing—From generator based on transfer printed Si Humans to Humanoids," IEEE Trans. microwires," in 44th European Solid State e-skin. Robotics, vol. 26, pp. 1-20, 2010. Device Research Conf., 2014, pp. 86-89. Challenges often come with [4] W. Taube Navaraj, et al., "Nanowire FET [22] K. Takei, et al., "Nanowire active-matrix opportunities. The opportunity for e-skin Based Neural Element for Robotic Tactile circuitry for low-voltage macroscale artificial Sensing Skin," Frontiers in Neuroscience, skin," Nat. mater., vol. 9, p. 821, 2010. in various application areas have been vol. 11, 2017. [23] C. G. Núñez, et al., "Heterogeneous already discussed. In addition, the [5] R. S. Dahiya, et al., Robotic tactile sensing: integration of contact-printed semiconductor development of e-skin is also acting as technologies and system: Springer Science & nanowires for high-performance devices on Business Media, 2012. large areas," Microsystems & catalyst for advances in other areas and is [6] R. Dahiya, "Electronic Skin," in 2015 XVIII Nanoengineering, vol. 4, p. 22, 2018. opening up new opportunities in those AISEM Annual Conference, 2015, pp. 1-4. [24] T. Someya, et al., "The rise of plastic areas. For example, with requirement of [7] A. Nathan, et al., "Flexible Electronics: The bioelectronics," Nature, vol. 540, p. 379, Next Ubiquitous Platform," Proc. of the 2016. bendability in e-skin, the conformable IEEE, vol. 100, pp. 1486-1517, 2012. [25] M. Drack, et al., "An Imperceptible Plastic electronics has also advanced and have [8] W. Dang, et al., "Stretchable wireless system Electronic Wrap," Adv.Mater., vol. 27, pp. opened opportunities for developing for sweat pH monitoring," Biosensors and 34-40, 2015. predictive, and accurate models for Bioelectronics vol. 107, pp. 192-202, 2018. [26] A. Adami, et al., "POSFET touch sensor with [9] X. Huang. (2017). Bioresorbable Materials CMOS integrated signal conditioning flexible devices, which could be and Their Application in Electronics. electronics," Sensors and Actuators A: exploited to predict the bending state or Available: Physical, vol. 188, pp. 75-81, 2012. stress patterns. This is because device https://www.cambridge.org/core/elements/bi [27] S. Gupta, et al., "Temperature Compensated oresorbable-materials-and-their-application- Tactile Sensing using MOSFET with P response changes constructively of in- (VDF-TrFE)/BaTiO3 Capacitor as Extended destructively as a result of bending electronics/C3077400C469CD7740ABAB35 Gate," IEEE Sensors J., Vol. 19 (2), pp 435- induced stresses [35-37]. 0B17E5CA 442, 2019. [10] S.-K. Kang, et al., "Advanced Materials and [28] S. Gupta, et al., "Ultra-thin chips for high- The e-skin or tactile skin in robotics Devices for Bioresorbable Electronics," performance flexible electronics," npj and “second skin” in wearable Accounts of Chemical Research, vol. 51, pp. Flexible Electronics, vol. 2, p. 8, 2018. applications are closely linked through 988-998, 2018. [29] C. G. Nunez, et al., "Energy Autonomous [11] R. S. Dahiya, et al., "Directions toward Electronic Skin," NPJ Flexible Electronics, flexible electronics technology. effective utilization of tactile skin: A review," vol. x, 2019 (DOI: 10.1038/s41528-018- Altogether these areas offer IEEE Sensors J. vol. 13, pp. 4121-4138, 2013. 0045-x). unprecedented opportunities for tackling [12] C. G. Núñez, et al., "Energy-Autonomous, [30] B. C.-K. Tee, et al., "A skin-inspired organic Flexible, and Transparent Tactile Skin," Adv. digital mechanoreceptor," Science, vol. 350, several pressing societal challenges. For Funct. Mater., vol. 27, p. 1606287, 2017. pp. 313-316, 2015. example, when wrapped around surgical [13] D.-H. Kim, et al., "Epidermal Electronics," [31] L. E. Osborn, et al., "Prosthesis with tools (e.g. tool used in pin-hole surgery) Science, vol. 333, pp. 838-843, 2011. neuromorphic multilayered e-dermis the flexible sensory patches could allow [14] V. J. Lumelsky, et al., "Sensitive skin," IEEE perceives touch and pain," Science Robotics, Sensors J., vol. 1, pp. 41-51, 2001. vol. 3, p. eaat3818, 2018. surgeons the feel the tissues or palpate [15] W. Gao, et al., "Fully integrated wearable [32] C. Bartolozzi, "Neuromorphic circuits impart the internal body parts. The disposable sensor arrays for multiplexed in situ a sense of touch," Science, vol. 360, pp. 966- and wearable solutions enabled by perspiration analysis," Nature, vol. 529, p. 967, 2018. 509, 2016. [33] S. Decherchi, et al., "Tactile Sensing Data advances in e-skin are aiding positive [16] H. Ota, et al., "Application of 3D printing for Classification by Computational changes in health monitoring practices at smart objects with embedded electronic Intelligence," IEEE Tran. Robotics, vol. 34, the global scale. E-skin type patches are sensors and systems," Adv. Mater. Techn., pp. 1-5, 2011. vol. 1, p. 1600013, 2016. [34] J. Grollier, et al., "Spintronic Nanodevices for also being explored as smart tags for real [17] S. Khan, et al., "Technologies for printing Bioinspired Computing," Proceedings of the time monitoring of food and drugs. sensors and electronics over large flexible IEEE, vol. 104, pp. 2024-2039, 2016. Coupled with advances in the Internet of substrates: a review," IEEE Sensors J., vol. [35] H. Heidari, et al., "Bending induced electrical 15, pp. 3164-3185, 2015. response variations in ultra-thin flexible chips Things (IoT), e-textiles, virtual reality, [18] C. García Núñez, et al. (2018). Integration and device modeling," Applied Physics augmented reality, Industry 4.0, organ on Techniques for Micro/Nanostructure-Based Reviews, vol. 4, p. 031101, 2017. chips, and artificial intelligence/deep Large-Area Electronics. Available: [36] S. Gupta, et al., "Device Modelling for https://www.cambridge.org/core/elements/in Bendable Piezoelectric FET-Based Touch learning, the e-skin technologies hold the tegration-techniques-for- Sensing System," IEEE Trans. Circuits and promise of achieving a new level of micronanostructurebased-largearea- Systems I: Regular Papers vol. 63, pp. 2200- human connectedness. electronics/F83E7CAF7CDBE93E69C7434 2208, 2016. C3EE29DED [37] A. Vilouras, et al., "Modeling of CMOS [19] D.-H. Kim, et al., "Stretchable and foldable devices and circuits on flexible ultrathin REFERENCES silicon integrated circuits," Science, vol. 320, chips," IEEE Trans. Electron Devices, vol. [1] S. Bauer, et al., "25th Anniversary Article: A pp. 507-511, 2008. 64, pp. 2038-2046, 2017. Soft Future: From Robots and Sensor Skin to