1 Introduction

After years of academic research and commercial promotion on Internet of things (IoTs), various applications (e.g., smartphones, intelligent monitoring, home security systems, wearable electronic devices) have greatly improved human life in multidiscipline aspects. The concept of IoTs, defined as things (or people) connecting to the Internet through functional nodes, has rapidly expanded to various fields of intelligent transportation, smart environment, urban construction, industrial manufacturing, augmented reality (AR), virtual reality (VR), etc. [1]. Wireless sensor network is the core of IoTs, which is commonly composed of more than one billion sensors and electronic devices [58, 59]) have been developed to increase the charge density. And ultra-high energy conversion efficiency of 48% could be achieved by utilizing a novel voltage balance bar design in liquid lubrication and charge space-accumulation effect [60]. These studies are significant to promote the commercialization of TENG-based self-powered systems. The other is that TENG acts as an energy harvester to provide electrical energy for the sensor [61]. Here, we will summarize the application of TENGs in the intelligent IoTs from five aspects: agriculture, industry, city, emergency monitoring, and ML-assisted AI applications (Fig. 2).

Fig. 2
figure 2

Multi-discipline applications of triboelectric nanogenerators for the intelligent era of Internet of things [29, 46, 68, 75, 80, 82, 129, 149,

Fig. 4
figure 4

Copyright 2021 Elsevier Ltd. b Schematic diagram of TENG tilt sensor applied to ship attitude sensing and experimental apparatus, and circuit diagrams and design of the self-powered system [81]. Copyright 2021 American Chemical Society. c Application of the smart chemical protective suit in biological movement energy harvesting and self-powered safety monitoring system [82]. Copyright 2021 Wiley–VCH GmbH. d Schematic diagram of seawater self-powered electrolysis and TENG test system [86]. The structure of PSW-TENG is composed of pump TENG and main TENG, which are two typical freestanding rotating TENGs. (Pump TENG is used to improve the performance.) To obtain self-powered wireless sensing information, by integrating the inductance coil with PSW-TENG, the pulse signal from PSW-TENG is modulated into a resonant signal by LC oscillating circuit for wireless transmission. Due to the influence of metal on the resonant frequency of oscillation circuit, they proposed a self-powered metal composition detection/monitoring system combining PSW-TENG with the electromagnetic sensor, which can be used in factory pipeline for checking the types of tin and calculating the quantity of products. It can detect the height and horizontal position of the metal parts (which pass under the sensor) to identify the products and count the number of products, which is very suitable for the management of the production line.

The excellent electrical output of TENG can also be used for seawater electrolysis. Zhu et al. designed a pulsed TENG that could continuously harvest energy from water and wind for self-powered seawater electrolysis [101, 102]. With the rapid development of modern industrial technology, higher requirements are put forward for the speed and service life of rolling bearings. However, the skidding phenomenon easily occurs at high speed, which is one of the most common failure forms of rolling bearings. It may cause scratches, wear, reduce the rotation accuracy of bearing, and significantly affect the working performance and service life of bearings [103, 104]. In order to avoid serious equipment damage and consequent major economic losses, it is very essential to monitor the skidding rate of rolling bearing. Therefore, ** detection, and speeding capture for the traffic management system realized by the CN-STS and real-time charge density output signals through Raspberry Pi [134]. Copyright 2021 Elsevier Ltd

TENG applications in smart transportation: condition monitoring. a Working principle of self-powered vehicle monitoring system fixed on the main axis of automatic vehicles and BS-TENG rolling independent mode [129].

Full size image

In Fig. 9c, Jiang et al. showed a self-powered hydrogen sensor based on an impedance adjustable windmill-like TENG [49]. The self-powered H2 sensing system consists of three main parts: TENG, H2 sensor, and alarm LED. When the H2 concentration changes, the output signal will be influenced, which will lead to the brightness change of the alarm LED. With the development of new energy vehicles, real-time monitoring of hydrogen leakage is of great significance during vehicle operation. Moreover, in order to further promote the development of intelligent transportation systems, Tang et al. proposed a TENG-based wireless monitoring system to bicycles, electric bicycles, and motorcycles, which can monitor them illegally entering the sidewalk, speed measurement, and the driving direction on non-motorized lanes [133]. Yang et al. developed a self-powered triboelectric sensor made of electrospun composite nanofibers for intelligent transportation monitoring and management [134]. To meet the requirements of fast response and high sensitivity for the smart traffic management, the transferred charge density was adopted as the sensing signal which could perfectly record the subtle differences and was more suitable for dynamic traffic monitoring (compared to the voltage or current signal). The adoption of carbon nanotubes doped into PVDF nanofibers can greatly improve the electrical output performance and pressure sensitivity (0.0406 C m−2 kPa−1 at low pressure range and 0.0032 μC m−2 kPa−1 at higher pressure range). Notably, using charge amplifiers in power management to process the charge input signals, the vehicle signal and human walking signal can also be clearly distinguished. By connecting to cloud IoTs service, the functions of traffic flow management, overlap** and speeding vehicle capturing, and plate number recognition were realized by the self-powered TENG sensor arrays with a compensation circuit (Fig. 9d). In contrast to this, Shan et al. designed a novel inverting TENG to realize the conversion from DC to AC signal based on the coupling of triboelectrification and air breakdown [135]. The inverting-TENG exhibited unique characteristics that both the pulse width ratio and amplitude ratio of AC signals can be controlled by tuning the distributed width and electronegativity difference of two opposite materials. Benefiting from these characteristics, the authors designed a real-time computer-simulated displacement and direction controller for a car, which demonstrates its potential applications for real-time cars position monitoring and car parking management.

The trains and high-speed rails running on the track will generate enormous wind energy due to their fast speed. In order to collect this wasted wind energy, as shown in Fig. 10a, Zhang et al. fabricated an elastic rotary TENG (ER-TENG) to harvest the wind energy generated by high-speed trains and supply the power to relevant sensing devices [136]. The transferred charge (QSC), short-circuit current (ISC), and open-circuit (VOC) obtained by ER-TENG are 0.9 μC, 120 μA, and 600 V, respectively. When the load resistance is 50 MΩ, the obtained peak power is 29.1 mW. Du et al. designed a multi-mode TENG (MM-TENG), which is composed of multilayer floating sliding components and multilayer corrugated contact-separation components [137]. It is used to harvest environmental mechanical energy at the junction of freight cars and supply power to the sensors for monitoring the train status (Fig. 10b). The multilayer waveform contact-separated TENG provides charge excitation to the entire TENG, which greatly improves the overall output performance of the MM-TENG. Similarly, ** et al. developed a magnetic levitation porous nanogenerator (MPNG), which effectively combines TENG and EMG (Fig. 10c) [138]. The device is used to harvest the vibration energy of the train and continuously supply power to the wireless smart sensors. At 20 Hz, the peak VOC of TENG is 43.8 V and the ISC is 1.39 μA, and the maximum VOC of EMG is 7.7 V and ISC is 4.1 mA. By connecting with external resistors, TENG provides a peak power density of 0.34 mW g−1 at 50 MΩ, while EMG delivers a maximum power density of 0.12 mW g−1 at 700 Ω. MPNGs have the potential of wireless monitoring without an external power source, especially in the train monitoring systems.

Fig. 10
figure 10

Copyright 2021 American Chemical Society. b Multiple-mode TENG with charge excitation collects environmental mechanical energy for self-powered freight train monitoring [137]. Copyright 2021 Elsevier Ltd. c Self-powered wireless smart sensor based on magnetic levitation porous nanogenerator for a train monitoring system [138]. Copyright 2017 Elsevier Ltd

TENG applications in smart transportation: railway monitoring. a Structure design and output power of elastic rotation TENG [136].

Full size image

Like cars and trains, ships are driven by large power sources, but the distributed sensors on-ship require continuous low-power driving. If these sensors can be driven by the harvested wave vibration energy, it will be a very attractive application. On this basis, Zhang et al. proposed a hollow-ball buoy-assisted triboelectric ocean-wave spectrum sensor (TOSS) [139]. The TOSS can monitor wave height (H), period (T), frequency (f), wave speed (v), wavelength (L), and wave steepness (δ) in real time. Due to the low cost and simple structure of sensors, large sensor arrays can be used to detect waves in a wide range of sea areas. Based on TENG, a magnetic flip-plate dual-function sensor (MFTDS) was developed to detect pneumatic flow and liquid level [140], which could detect the pneumatic flow rates ranging from 10 to 200 L min−1 and had a flow resolution of 2 L min−1. As shown in Fig. 11a, Ren et al. proposed a hybrid nanogenerator with TENG and EMG as power source, which harvested wave energy for long-distance (1.5 km) wireless transmission [141]. The integration of TENG and EMG for hybridized wave energy harvesting can combine their complementary advantages. Based on this, they developed a self-powered route avoidance warning system for navigation at sea to ensure the navigation safety of ships in all weather conditions. An et al. combined the advantages of sensors based on TENG and slug assemblies to construct self-powered slug mechatronics (BIM) panel [142]. With the help of magnetic floats, the BIM panel can be used as a reliable mechanical indicator to directly reflect the induced liquid level and flow characteristics through the visible color change of the flap and can also be used as an electronic sensor for automatic control and remote wireless monitoring to ensure the safe transportation of cruise ships. Figure 11b shows a robust and self-powered tilt sensor for ship attitude sensing [143]. Wang et al. designed an annular PTFE tube with copper electrodes segment disposed on the surface, and the internal liquid was encapsulated in PTFE tube without bubbles. Based on this, they developed a tilt monitoring system to indicate the ship’s attitude in real time. When the inclination occurs, the internal fluid (pure water) flows through different electrode segments with the LED lighting up to indicate corresponding angles. Then, the inclination direction of the ship can be judged by the positive and negative value of VOC/dt. When the inclination of the ship exceeds the dangerous value, the visualization system can also trigger the alarm device.

Fig. 11
figure 11

Copyright 2021 Wiley–VCH GmbH. b Schematic diagram of TENG tilt sensor applied to ship attitude sensing and experimental apparatus [143]. Copyright 2020 Elsevier B.V. c Schematic and experimental structure of a fully self-powered vibration monitoring sensor driven by AC/DC-TENG [145]. Copyright 2020 American Chemical Society. d Schematic diagram of the sensor working in the tunnel, and the enlarged view of the self-power sensing system [146]. Copyright 2016 American Chemical Society

TENG applications in smart transportation: navigation monitoring. a Schematic of hybrid wave energy harvesting nanogenerator (HW-NG) network distributed in the waters adjacent to coral reefs, which is used to harvest wave energy, and the developed automatic RAW system for ocean navigation [141].

Full size image

As aforementioned, monitoring the condition of bridges and tunnels is also important for smart transportation. Lin et al. proposed and implemented a self-powered, flexible, timbo-like triboelectric force and bending sensor, which could effectively monitor the rockfalls and landslides near roads [167, 173, 175, 176]. As shown in Fig. 14a, Wu et al. developed a smart keyboard based on TENG for dynamic monitoring of keystrokes [177]. The multi-channel keyboard array consists of 16 highly flexible and malleable silicon-based keys, and each key consists of a CE mode TENG and a shield electrode. The TENG is used to capture the electrical signal of the ty** behavior, while the shielding electrode helps to reduce the environmental interference. Different voltage signals can be extracted from the keystroke-related features, such as ty** delay, hold time, and voltage signal amplitude. After the principal component analysis (PCA), the multi-class SVM classifier is used for authentication and identification. The accuracy is as high as 98.7%, which shows that the authentication system based on keystroke dynamics is feasible for practical applications. Besides, Maharjan et al. proposed a keystroke biometric authentication system based on hybrid electromagnetic–triboelectric sensors assisted by artificial neural network (ANN) [173]. DL (as a subfield of ML) can provide an effective way to automatically learning the representative features from the collected raw signals by training end-to-end neural network. By acquiring the keystroke information of users, the DL neural network continuously learns and adapts to the keystroke behavior of users, such as ty** force, hold time, flight time, and interval, thus providing dual security for user authentication. Compared with other ML platforms, the neural network learns and adapts in the system to provide more sophisticated personal security systems. This battery-free keystroke sensing system and the software system based on neural network can accurately differentiate and verify different users with 99% accuracy throughout their individual ty** habits.

Fig. 14
figure 14

Copyright 2018 Elsevier Ltd. b Experimental platform and the framework of the proposed method [171]. Copyright 2020 Elsevier Ltd. c Artificial intelligence toilet (AI-Toilet) for integrated health monitoring system and t-SNE diagram of sitting posture dataset recorded from the pressure sensor array [178]. Copyright 2021 Elsevier Ltd

AI applications of TENG with machine learning. a Schematic and exploded view of a single triboelectric key. The system overview includes the training process and the authentication/identification process, and the difference matrix between user inputs of different feature types combinations [177].

Full size image

In addition to the keystroke dynamics, driving behavior-based monitoring and recognition is another promising technology in biometric authentication applications. With the help of AI technology, the driver’s steering action can be recognized by analyzing the driver’s steering operation to establish a steering wheel motion model. Zhang et al. investigated and verified the possibility of using TENG to detect the steering wheel driving behavior [171]. To compare the response speed of different sensors (driving simulator, camera, and TENG) and evaluate the accuracy of steering action detected by TENG, two experiments based on driving simulators are designed, and the experimental framework is shown in Fig. 14b. At the same time, a supervised ML algorithm is developed to detect the driver’s steering actions, and the results show that TENG has the fastest response speed statistically. The algorithm trained with the random forest (RF) classifier has the highest accuracy of 92.0% in the test dataset, which demonstrates the potential application of TENG as the sensors for drivers’ steering actions.

Recently, the combination of TENG pressure sensor and DL method has been proposed for biometrics recognition of artificial intelligence of toilet (AI-toilet). Zhang et al. demonstrated an AI-toilet system equipped with multiple functions for an integrated health monitoring system [178]. Figure 14c shows the schematics of the AI-toilet, and it is configured with an array of triboelectric pressure sensors for biometric recognition and a commercial image sensor for medical monitoring. Ten textile-based TENG sensors are placed on the toilet seat to detect the corresponding pressure changes when the user sits down. The design of the friction layer with spacers and frustum structure can improve the sensitivity and detection range. AI-toilet has the advantages of high privacy, low cost, and easy preparation. With the help of DL method, the biometrics information of 6 users sitting on the toilet seat was successfully identified with an accuracy of 90%. The AI-toilet can not only process the data collected from the hardware side, but also process the multimodal data interpreted from the software side. Finally, it can be combined with IoTs to embed the AI-IoT system into smart home.

7.2 AI Application of TENG with AR/VR Technology

With the commercialization of AR and VR technology, the virtual world has become a reality. The rapid development of AR and VR technology has laid the foundation for diversified applications in social media, industrial production simulation, surgical training, games, and so on. Similar to the interaction between human and virtual world, AR and VR systems rely on HMI sensors to interact with the virtual world. In addition to audio and visual feedback, this immersive experience may involve other types of tactile feedback. Therefore, the wearable sensors based on TENG have the advantages of ultra-thin, ultra-soft, conformal, and imperceptible, which help to provide a more extreme experience for AR/VR technology [179,180,181,182,183]. The identity recognition based on gait analysis is also a promising biometric technology [166, 184]. On the one hand, with the help of AI, information about user identity and real-time activities can be transmitted by analyzing the physical signals obtained from the sensing systems of the floor or socks. On the other hand, the collected physical signals can be mapped into virtual space, which can be used to build a digital human body system for motion monitoring, healthcare, identification, and future smart home applications. Lee’s group has proposed a textile-based TENG sock to assist DL techniques for gait analysis [32]. As shown in Fig. 15a, this TENG consists of the silicone rubber film with patterned frustum structures, the nitrile film, and conductive textiles, which are used as negative triboelectric layer, positive triboelectric layer, and output electrodes, respectively. The textile-based TENG integrated into the smart sock for gait monitoring shows high pressure sensitivity of 0.4 V kPa−1 and large sensing range of > 200 kPa. Using the optimized four-layer one-dimensional (1D) convolutional neural network (CNN) model to automatically extract features from the walking spectrum, the recognition accuracy of 13 participants was 93.54%, and five different human activities were detected simultaneously with an accuracy of 96.67%. In a practical scenario, the authors demonstrated a VR fitness game with smart socks as the control interface, which had great prospects in the development of digital people. After this work, they further developed a textile-based triboelectric sensory system for gait analysis and waist motion capture. Compared with their previous work, it added robotic assistance for lower limb and low back rehabilitation [185]. They integrated smart insoles and harnesses into a lower extremity rehabilitation robot, demonstrating the system’s promise for user identification, motion monitoring, robot manipulation, and game-enhanced training. In addition to smart socks and insoles in the wearable field, Lee’s group also proposed a smart glove assisted with ML method, which uses self-powered conductive and superhydrophobic triboelectric textile for the gesture recognition in VR/AR applications [48]. As shown in Fig. 15b, the pristine textile is modified to be superhydrophobic by using a facile, scalable, and cost-effective coating method. On the one hand, superhydrophobic textiles are used to harvest biomechanical energy from human motion and as active sensors to monitor human motion. On the other hand, the integration of superhydrophobic textiles, the use of glove-based HMI, and the training of finger motion signals by ML are synergetic to realize complex gesture recognition. Due to the superhydrophobic capability of the device, the recognition accuracy under sweat conditions can still be maintained at 96.9%. Toward practical applications, the author demonstrated gesture recognition based on glove-style HMIs, which can realize 3D VR/AR scenarios of shooting, baseball pitching, and flower arrangement.

Fig. 15
figure 15

Copyright 2020, Copyright The Author(s), Published by Springer Nature. b Demonstration of shooting game, which is based on the amplitude of output signals. And the AR demonstration of flower arrangement based on machine learning for complex gesture recognition [48]. Copyright 2020 The Authors. Published by WILEY–VCH Verlag GmbH & Co. KGaA, Weinheim

AI applications of TENG with machine learning. a Process from sensory information collection to the real-time prediction in VR fitness games and the confusion map of deep learning results [32].

Full size image

7.3 AI application of TENG with Digital Twin Technology

Digital twin is a technology to make full use of physical model/signal update/operation history and integrate multi-disciplinary/multi-physical/multi-scale/multi-probability simulation process to complete the map** in virtual space or Meta universe, which is possible to reflect the full life of corresponding entity equipment cycle process. In actual application scenarios, the digital twin system needs to integrate AI, ML, neural networks, and other methods to continuously learn and update the input information and adjust relevant operation mode. The combination of the digital twin and other related technologies (such as the IoTs) has provided favorable cyber-physical interaction and data integration. Recently, Lee’s group introduced several digital twin applications based on TENG, involving smart homes, intelligent manufacturing, virtual shop, etc. [33, 186, 187]

As shown in Fig. 16a, Shi et al. developed DL-enabled smart mats based on the triboelectric mechanism to realize an intelligent, low-cost, and highly scalable floor monitoring system [186]. The system was achieved by integrating the triboelectric floor mat array (with minimum electrode layout) with advanced DL-based data analysis. When the user walks through, the smart mat can obtain the differential signals. Through the integrated data analysis based on DL, the CNN model can be used to extract the identity information associated with walking gait patterns from the output signals. Based on this, they demonstrated an intelligent floor monitoring system that enabled real-time position sensing and recognition. The positioning information of each step can be used to control the light at designated position. Analyzing the complete walking signal can help to confirm whether the person is an effective user of the room, thereby automatically controlling the access control.

Fig. 16
figure 16

Copyright 2020, The Author(s), Published by Springer Nature. b Schematic of the low-cost TENG for soft gripper and its digital twin applications. The process flows from sensory information collection to ML training and real-time prediction in digital twin system [33]. Copyright 2020, The Author(s), Published by Springer Nature. c System architecture of the digital-twin-based virtual shop and ML-enabled automatic grasped objects recognition system [187]. 2021 Copyright The Authors. Advanced Science published by Wiley–VCH GmbH

AI application of TENG with digital twin technology. a Smart floor monitoring system based on the deep learning-enabled smart mats (DLES-mats) [186].

Full size image

One of the digital twin applications in intelligent manufacturing is the classification of components in the workshop and the assembly of complex products. In the sorting lines of unmanned factories, sensors are built into the manipulators to identify the shape, size, and type of components. TENG-based sensors are more compatible with soft robots because the Young’s modulus of soft materials typically used in TENG sensors is at the same level with silicone rubber and thermoplastic polyurethane rubber (TPU) [48, 16b, the scenes of digital twin are then established to replicate the digital information of the above manipulation in VR environment. By realizing real-time object recognition in a replicated virtual environment, digital twins can be further applied to the production control management of next-generation smart industry [187]. Another important scenario for digital twin applications is unmanned stores. Based on the enhanced soft manipulator, Lee’s team integrated PVDF pyroelectric sensors to achieve more complex sensing functions. To show the potential of the proposed intelligent manipulator for future online shop** and unmanned shop applications, they proposed a virtual shop system based on the digital twin model. As shown in Fig. 16c, through this system, users can select items in the digital twin virtual store, and the smart manipulator in the real unmanned store will simultaneously make corresponding actions based on the signals collected by the TENG sensors.

In the future, the combination of AI technology and IoTs technology will bring people a new living, working, and manufacturing environment. Imaging the future prospect under 5G and IoT infrastructure by utilizing the self-powered sensory interaction system (with the features of facile design, low cost, and high compatibility, etc.) together with AI techniques, a smarter society can be established toward intelligent industrial and automation, smart city modernization, smart agricultural mechanization, and normalization of emergency monitoring.

8 Summary and Perspective

With a wide range of material choices and diverse structural designs, the smart sensing technology based on TENG has been widely studied because of its great potential in the construction of IoTs-related smart applications in this 5G era. In this review, we systematically summarize the application progress of TENG in multi-discipline application scenarios of IoTs, such as smart agriculture, smart industries, smart cities, emergency monitoring, and AI applications assisted by ML method. With the assistance of ML, AI technology has been introduced into the IoTs, which provides a promising research direction for the development of IoT technology.

In the future, the gradual promotion of IoT technology will greatly improve people’s lifestyles. Although the applications based on TENG (whether for energy harvesting or self-powered sensing) have made significant progress in the past period, there are still challenges to be solved in the practical application of TENG, such as power management and energy storage, service life, packaging technology, and large-scale sensor integration. The further development of the smart sensing applications of IoTs based on TENG needs to be considered from the following aspects (Fig. 17).

  1. 1.

    Whether in industry, agriculture, or smart city sensing applications, sensors mainly have two working modes. First, the device itself plays the role of energy harvesting. The AC signal is converted into DC, which can be directly used through the power management circuit to provide the required energy for subsequent commercial sensors [190, 191]. Paired power management is quite necessary for the application of TENGs because the harvested energy from surroundings is time-dependent, unstable, and susceptive to environmental changes. Meanwhile, corresponding energy conversion efficiency (an important technical index in energy harvesting systems) needs to be elaborately considered due to the energy consumption by some active electronic components during the AC/DC conversion [23, 192, 2.

    Another typical application of TENG is to use it as an active self-powered sensor, in which the generated electrical signal is a direct response to the external environment. In this case, it requires the TENG active sensors with good sensitivity, fast response time, and wide detection range. Therefore, develo** sophisticated structural design and effective working mode is particularly important. In practical application, the performance of TENG-based self-powered sensor may degrade to reduce the detection accuracy and service life. Therefore, the packaging of the self-powered device/system is also important. Researchers should make greater efforts in packaging materials and packaging technology to improve the stability and durability of the sensors. For example, researchers can select materials with excellent elasticity and mechanical properties (such as PDMS) to encapsulate the sensory devices to enhance durability or select appropriate encapsulation materials according to application scenarios (e.g., considering the biocompatibility of materials in the process of intelligent medical application). For multiple sensory devices, under the framework of IoTs, a large number of sensors need to be deployed for real-time sensing and data management/analysis. Therefore, it is very necessary to develop the integration of other modules in self-powered sensory systems to achieve favorable operation of the whole wireless sensory nodes (including the supporting components).

  2. 3.

    The combination of AI technology with IoTs technology will bring a new living, working, and manufacturing environment. In the construction of IoTs, the development of intelligent systems is the core issue. The overall architecture of IoTs intelligent system is composed of three layers: data collection layer, data communication layer, and data processing and analysis layer. In the self-powered IoTs, a large number of sensory nodes based on TENGs act as the data collection layer. Wireless networks are typical data communication layers. The cloud platform is used for data processing and analysis, and ML technology is a typical AI technology widely used in data processing [200]. The future research on AI-related smart in IoTs sensing application of TENG should be considered from the following aspects. First, people need to develop more advanced TENG sensors with embedded AI technology. When large numbers of TENG sensors are used as data sources for AI, the random characteristics of ambient mechanical energy (or information) may raise unavoidable problems for AI processing. For example, when the mechanical energy in the environment is insufficient, the obtained data by TENG sensor are random and the accuracy is low. AI technology needs to improve the incomplete information and improve the accuracy of information by learning the characteristics of complete information. Second, necessary power management for TENG sensory networks will also be a new research direction for the AI-related applications, which is helpful to solve the power consumption problem of traditional sensory systems. Third, most of the AI applications based on TENGs are in the early stage. For example, the data collection and data analysis in the process of machine learning often use offline analysis methods. The development of the auxiliary circuits between the sensor and the computer is lack of investigation, which limits the communication and practical application of the TENG-based IoTs sensory system. Therefore, in addition to the continuous development on TENG-based sensors, the development and design of auxiliary circuits, communication equipment, and relevant hardware should also be considered to pursue its future industrialization. Fourth, machine learning is a widely used AI technology in data processing, and the frontier of ML is deep learning, which can be used to classify and process a large amount of sensing data. The large amount and high randomness of sensing data captured by TENG sensor network put forward higher requirements for the accuracy of ML. New ML algorithm design, optimization, and training method need to be further developed to extract more feature information and capture the imperceptible sensing information. Finally, researchers need to devote to expand the application of AI such as digital twin. By combining TENG sensors with digital twins, the application scope of TENG will be highly expanded in smart manufacturing and engineering construction.

  3. 4.

    The application of TENG in IoTs keeps expanding in various scenarios. Previous studies have demonstrated the feasibility of TENG in smart agriculture, smart industry, smart city, emergency monitoring, and other fields. As an important part in the new-generation information technology, the IoTs have been widely used and are still under develo**. Continuous research can explore more potential applications of TENG in logistics warehousing, positioning and navigation, intelligent buildings, public safety, enemy detection, etc.

Fig. 17
figure 17

Challenges and perspectives of the IoTs application based on TENG sensors

Full size image

Overall, as an energy harvester, TENG can collect a wide range of sustainable energy in the environment, which is very efficient and environmentally friendly. Besides, TENG itself can also be used as a self-powered sensor for various applications. Moreover, the applications of TENG sensors in smart IoTs have always been the main subject. The studies on the combination of AI technology based on TENG and IoTs are still in the early stage and need to be given with further attention. Although challenges still exist, the continuous research and exploration of multi-discipline intelligent applications based on TENGs with the assistance of AI technology will definitely shed light on the harmonious coexistence of humans and machines in the era of IoTs, as well as the immersive and efficient interaction in many scenes.