Roadbed tribological energy harvester

Abstract

Roadbed tribological energy (RTE) is a promising recoverable resource with an estimated potential on the terawatt scale, generated annually by the interaction between tires and road surfaces. However, RTE remains underutilized due to the lack of effective energy harvesting technologies that can address its high-entropy characteristics. Here, we present a revolutionary harvester formed by a freestanding layer triboelectric nanogenerator array embedded in the road. The harvester effectively converts low-grade vibratory RTE into electrical energy. It demonstrates the potential to achieve a peak power of 16.409 milliwatts and an average power of 2.2 milliwatts from a compact 78–square centimeter road area under a single tire impact, with a triboelectric conversion efficiency of 11.723%. In addition, we developed a self-powered intelligent and connected transportation system (SP-ICTS), integrating a five-in-one harvesting array. Experimental findings show that the system can meet the SP-ICTS’s electricity requirements along a 1-kilometer segment with a 50-meter harvester.

Large amounts of electrical energy are harvested from the roadbed with the triboelectric mechanical coupling.

INTRODUCTION

The development and use of energy are critical to human progress and the basis of civilization. The increase in productivity and the advancement of civilization are closely tied to new energy sources and innovative technology. As a recoverable energy resource, sustainable roadbed tribological energy (RTE), which spans multiple disciplines, including mechanics, thermodynamics, materials science, and electrical engineering, has enormous potential and reserves. It is calculated that annual energy generation from RTE is about 0.3 TW via the interaction of tires and road surfaces. However, the enormous energy of RTE remains not to be fully used. The reasons for this consist of two aspects: RTE is difficult to capture energy due to its high entropy, with irregular and discontinuous character (1–4); the collection of RTE needs materials such as exceptional durability, abrasion resistance, slip resistance, thermal stability, and low-temperature cracking resistance. These prevent the use of energy of RTE, implying that the need for the development of efficient conversion technologies is urgent. Several energy harvesting technologies, such as piezoelectric, thermoelectric, electromagnetic, and photovoltaic methods, have been explored for converting RTE into electrical energy. But their effectiveness also needs to be proven. The most important one is the low conversion efficiency (0.5 to 10% for the piezoelectric, 1 to 10% for the thermoelectric, 1 to 15% for the electromagnetic, and 15 to 25% for the photovoltaic) (5–8). At the same time, these technologies also face various limitations in the application process: Piezoelectric devices are expensive (~$700/m²), thermoelectric devices require large temperature gradients (around 200°C), and photovoltaic systems rely heavily on sunlight (varies with season and latitude). As a result, existing energy harvesting technologies still cannot effectively solve the great challenges in the use of RTE, especially in road environments.

Although the acquisition of RTE is challenging, its potential remains undeniable, especially in road infrastructure. RTE plays a crucial role in advancing the future of self-powered intelligent and connected transportation systems (SP-ICTS). The primary goal driving the development of SP-ICTS is to establish a distributed, very microgrid across roadways through in-road energy production. This innovative local very microgrid network will greatly improve the efficiency of energy capture and distribution and can revolutionize the way we view road infrastructure, transforming roads from passive infrastructure to intelligent systems that generate energy and actively share data. Combining RTE with the very microgrid network will not only provide a solution to the current energy sustainability challenges but will also additionally improve the intelligence of the transport system. Roads will no longer serve just as channels for traffic; they will evolve into carriers of both energy and data. By capturing the interaction between the tire and the road surface, RTE can not only support sustainable energy targets but also provide real-time mass data support for traffic safety operation and management. These data, including information such as vehicle load, speed, and movement patterns, offer valuable insights into road conditions and traffic behavior. Efficient capture of energy and analysis of the data will allow SP-ICTS to evolve into an integrated five-in-one system that encompasses self-power, sensing, positioning, communication, and guidance. These functions will enhance the overall ability of SP-ICTS, providing authorities with the tools to more effectively manage traffic flow, monitor road conditions, and respond to emergencies. This represents a crucial step toward building a fully connected, intelligent transportation system that will thrive in the artificial intelligence–driven future.

Triboelectric nanogenerators (TENGs; invented in 2012) are an efficient energy harvesting technology that captures high-entropy mechanical energy from a variety of sources, such as wind, airflow, noise, vibration, water waves, and biodynamic energy. TENGs are particularly effective at converting low-grade mechanical energy into electrical power (9–13), especially when it comes to harvesting the high-entropy vibration energy generated between the tire and the road surface. With advantages such as low cost, a wide selection of materials, and high conversion efficiency at low frequencies, TENGs are ideally suited for large-scale deployment on roadways. In addition, TENGs have excellent vibration sensing capabilities (14–17), enabling them to accurately detect small changes in road and vehicle conditions (18). By capturing the vibration data generated by the interaction of the tire with the road surface, TENGs can collect extensive information about abnormal driving behaviors such as speeding, overloading, and illegal parking. These data, moreover, enhance the functionality of SP-ICTS. With their efficient sensing capabilities, TENGs enable SP-ICTS to not only monitor traffic conditions in real time but also elevate the intelligence of traffic management systems. As a result, TENGs offer innovative opportunities for RTE applications and hold great potential in driving smart, sustainable transportation systems, particularly in enabling SP-ICTS to achieve full functionality.

This paper presents an RTE harvester that embeds a freestanding layer TENG array in the roadbed. It effectively converts low-grade RTE into electrical energy, achieving a maximum triboelectric conversion efficiency of 11.723%. It generates a peak power of 16.409 mW and an average power of 2.2 mW from a compact surface area of 78 cm² with a single tire impact. The harvester shows high sensitivity to vibrations in the 0- to 50-Hz range, with a vibration sensitivity of 28.28 V/mm. These advantages are attributed to its unique multilayer structure, which enhances both energy conversion efficiency and sensing sensitivity. Leveraging the RTE harvester, an SP-ICTS system integrated with self-powering, positioning, communication, sensing, and guidance functionalities was developed. The system includes an energy harvesting unit formed by eight RTE harvesters to establish a very microgrid, along with two RTE harvesters for the sensing unit, a power management circuit, an energy storage unit, a data collection and processing unit, and a communication module. Notably, the multiscale SP-ICTS system can be programmed to expand according to specific requirements, enabling the deployment of the appropriate number of RTE harvesters to meet application needs. Experimental results indicate that the system can cover the electricity demand of SP-ICTS within a 1-km range when the 50-m highway is equipped with the RTE harvester. The SP-ICTS holds important promise for advancing vehicle-to-infrastructure and vehicle-to-everything connectivity, marking a major step toward the full integration of vehicles, roads, and cloud computing systems and contributing to the development of intelligent transportation systems.

RESULTS

RTE harvester

The world now has more than 21 million kilometers of roads, with each continent and country maintaining its interconnected road network (19), drawing result that the distribution map of global road network using the statistical data is shown in Fig. 1A. Figure 1B (data sources in table S1) shows a consistent increase in total road mileage over the past 20 years, and we expect this upward trend to continue in the future. These expanding networks include highways, urban roads, and rural lanes. Roads are categorized into four primary classifications based on their scale: total mileage, road density, unpaved roads, and paved roads (Fig. 1, C to F). Notably, countries such as the United States (7.04 × 10⁶ km), Monaco (3850 km/100 km²), India (2.2 × 10⁶ km), and China (5.35 × 10⁶ km) lead in the extent of their road networks (data sources in table S2). At the same time, global gasoline consumption continues to rise annually (20), although a notable portion of the energy in gasoline is wasted (Fig. 1G), with only about 15% being used to drive the wheels (21). On busy urban roads and highways, frequent vehicle motion leads to substantial interaction between the tires and the road surface, generating major forms of energy, such as vibration, friction, and heat, as shown in Fig. 1H. This results in the formation of a large high-entropy energy source between the tires and the road surface. It is estimated that the annual exploitable energy capacity generated by the tire-road interaction is ~0.3 TW (data estimation details in text S1).

Fig. 1. RTE characteristic and RTE harvester.

(A) Drawing the distribution of the global road network using statistical data. (B) Road growth changes in selected countries over the past 20 years. (C) Statistics of total road coverage. (D) Statistics of road density. (E) Statistics of unpaved roads. (F) Statistics of paved roads. (G) Gasoline energy distribution map. (H) Schematic diagram of tire-road interaction. (I) Current status of RTE harvesting technology and applications. PENG, piezoelectric nanogenerator; EMG, electromagnetic generator; TEG, thermoelectric generator; PV, photovoltaic. (J) Schematic diagram of RTE harvester installation. (K) Exploded view of RTE harvester.

Because of the notable potential of this energy source yet the lack of suitable capture technologies, we reviewed the literature on the inventions and applications of roadbed energy harvesting devices over the past decade (text S2 provides further details). On the basis of the analyses of literature data, the adoption rate of TENG technology for high-entropy energy capture from roadbeds has reached 37.3% (Fig. 1I). A high utilization rate means that, in practice, the feasibility of using TENG for RTE collection is promising. Therefore, this paper is dedicated to capturing the high-entropy energy (vibrational energy and frictional energy) generated during the interaction between vehicles and road surfaces through TENG technology and presents a design for an RTE harvester based on TENG technology, with overall dimensions of 100 mm by 140 mm (diameter by height). Figure 1J positions the device between the road’s surface layer and subgrade. The harvester consists primarily of three components: a force-bearing hemisphere, a vibration amplifier, and a generator, as illustrated in Fig. 1K, with additional details in figs. S1 and S2. The spherical design at the top allows it to absorb impact forces from all directions. When combined with a spring, this design efficiently converts external impacts into internal vibrations (22). The fiber-reinforced nylon casing ensures that the device retains sufficient strength and durability under road and environmental conditions. The components are connected using threaded joints, facilitating component replacement and large-scale production. The internal spherical structure of the TENG amplifies the vibrations, causing the polyoxymethylene (POM) balls to alternate between contact and separation in the upper and lower stages (23, 24). The multilayer structure further enhances the internal charge density (25). The upper and lower electrodes, together with the POM balls, form a freestanding layer model, converting the bouncing of the balls into electrical energy. The output performance is shown in fig. S3. Compared to other road devices, this system offers lower manufacturing complexity, simpler process steps, zero energy consumption, and easier large-scale deployment.

The internal TENG comprises a thin polyfluorinated ethylene propylene (FEP) film, copper (Cu) foil electrodes, fiberglass, and POM balls, with a schematic diagram provided in Fig. 2A. The materials used for triboelectric generation offer a wide range of options, with their selection based on the difference in their ability to gain or lose electrons. The larger the gap in electron transfer capabilities, the stronger the charge transfer and, consequently, the higher the electrical output performance. The POM used in this study has a strong electron-donating ability (26) and is easy to process, cost-effective, and suitable for large-scale production. The FEP film used here has a strong electron-accepting ability and is widely used in TENGs (27, 28), with the performance of the generators it constitutes being extensively validated. The charge transfer mechanism is depicted in Fig. 2B, where the positively charged POM balls undergo reciprocating motion between two layers of FEP. The simultaneous effects of triboelectric charging and electrostatic induction result in the generation of equal quantities of positive and negative charges on the contact surfaces, establishing a potential difference as the balls alternately come into contact and detach. This motion drives the cycling of charges between the upper and lower stages. Both simulated and experimental results of the generated potential are in close agreement (fig. S4). In the multilayer configuration, the top of all layers is interconnected to form a single electrode, while the bottom of each layer is similarly connected to create the opposing electrode, with the POM balls moving between them (fig. S5). Figure 2C illustrates the theoretical model of ball motion between electrodes and presents a schematic of the optimized parameters, where L denotes the electrode separation distance, D represents the diameter of the POM balls, G indicates the gravitational force acting on the balls, and Δh refers to the vertical displacement of the balls.

Fig. 2. Structure and performances of RTE harvester.

(A) Structure of internal RTE harvester. (B) Charge transfer mechanism. (C) Schematic diagram of the structural model. (D) Electrical output performance of different materials. V_OC, Open-circuit voltage; I_SC, short-circuit current; Q_SC, short-circuit transfer charge. (E) Electrical output performance at different Δh values. (F) Electrical output performance at different layer numbers. (G) Transferred charge as a function of ball diameter D and interlayer spacing L. (H) Charge density per unit height for different ball diameter D and interlayer spacing L. (I) Energy conversion efficiency for different ball diameter D and interlayer spacing L. (J) Capacitor charging curve. (K) Output comparison after 90 days. (L) Transferred charge under different temperature and humidity conditions.

The friction layers of both the upper and lower electrodes are constructed from FEP, a material with a high propensity for electron gain (29, 30). When various materials with differing electron affinities are used for the friction layers, the resulting energy output is shown in Fig. 2D, comparing five types of small balls that are prone to losing electrons, including aluminum, copper, polypropylene, polyamide, and POM. Among these materials, POM balls exhibit superior performance compared to others and are easy to process, cost-effective, and suitable for large-scale production, which led to their selection as the friction layer material for this study. As roughness increases, the output performance continuously decreases (fig. S6A), with untreated FEP films exhibiting the highest performance. This is because the structure differs from the traditional independent layer model, where the generation of electric potential primarily relies on the sequential contact between the POM balls and the top and bottom surfaces. Therefore, a smooth surface maximizes the contact area, leading to better performance (fig. S6B). Both the diameter of the POM balls (D) and the electrode separation distance (L) exert a direct influence on the vertical displacement (Δh) and the output of the RTE harvester (31). Notably, when the difference between L and D is 2 mm, the generated electrical energy reaches its peak, with optimal performance achieved at D = 4 mm, as demonstrated in Fig. 2E. This configuration strikes a balance, optimizing both output and sensing accuracy. Furthermore, as illustrated in Fig. 2F, an increase in the number of layers results in a corresponding increase in the RTE harvester’s output, revealing a positive linear correlation between the output performance and the number of layers (additional test results for varying D and L are presented in figs. S7 and S8).

To further optimize the parameters for this structure, we evaluated the output performance for varying values of D and L. The results demonstrate that, for a fixed value of L, the transferred charge initially increases and subsequently decreases as D increases, as shown in Fig. 2G. This behavior is influenced by both gravitational (G) and electrostatic forces. As the particle size decreases, the contact area between the balls and electrodes increases during movement, thereby enhancing the output. However, smaller balls have reduced mass and gravitational force, making them more susceptible to electrostatic attraction by one of the electrodes, which inhibits further charge transfer and consequently diminishes the output. Conversely, for a constant particle size, the transferred charge also exhibits an initial increase, followed by a decrease as the interlayer distance (L) increases. This can be attributed to the fact that as L increases, the separation between the electrodes becomes more pronounced, improving electrical performance (32). However, if L becomes excessively large, then inadequate contact with the friction layers occurs, thereby reducing efficiency. Consequently, optimizing both L and D is critical for maximizing output. The peak output performance is achieved when L = 6 mm and D = 4 mm, with a single layer transferring 70 nC of charge. To enhance spatial efficiency, the charge generated per unit height is shown in Fig. 2H, where the charge density is maximized at D = 4 mm and L = 6 mm, yielding a value of 11.6 nC/mm. In this process, the mechanical energy transferred to the balls is denoted as E_P, with the electric potentials at the top and bottom of the balls represented by φa and φb, respectively. The energy conversion efficiency (η) is given by the following expression

$$\mathrm{\eta }=\frac{W}{E_{\mathrm{P}}}=\frac{\int (\mathrm{\phi }a-\mathrm{\phi }b)\mathit{dQ}}{\mathit{mg}∆h}$$ (1)

As shown in Fig. 2I, the energy conversion efficiency reaches a peak value of 11.723% when the parameters are optimized to D = 4 mm and L = 6 mm. Through the study of the transfer charge, charge density, and energy conversion efficiency, we determined the optimal parameters of this type of structure (D = 4 mm, Δh = 2 mm) and used this parameter in the following contents. The energy storage performance is evaluated using the charging curve of a capacitor, as depicted in Fig. 2J, where the charging behavior of a single RTE harvester is tested at a frequency of 5 Hz for capacitors with varying capacitance values. A 1000-μF capacitor is charged to 5 V in ~130 s. Figure 2K compares the output of an RTE harvester before and after 3 months of storage in a laboratory environment. In the case of vibration with a frequency of 10 Hz and an amplitude of 5 mm, the change in the amount of transfer charge in the laboratory for 90 days is negligible. The RTE harvester’s performance remained stable even after 100,000 cycles (f = 10 Hz, A = 5 mm), and no observable morphological changes were observed on the surface of the FEP layer (fig. S9). The output stability before and after immersion is shown in the figure (fig. S10). The results demonstrate the excellent durability and long-term operational stability of the device, even when deployed on road surfaces for extended durations.

Furthermore, recognizing the variation in natural environmental conditions across regions, the output performance of the RTE harvester was tested under various temperature conditions (ranging from −40° to 60°C) and humidity levels (ranging from 10 to 90%). As illustrated in Fig. 2L, the output remained relatively stable across these diverse temperature and humidity conditions, with only minor fluctuations observed. Additional test results for temperatures below 0°C are provided in fig. S11. These findings indicate that the RTE harvester maintains high stability under varying environmental conditions and is capable of withstanding the climate variations encountered in different regions.

Principle and properties of RTE harvester

Figure 3A presents a schematic illustrating the motion process of the ball. It shows the displacement in the vertical direction as a function of time. The vibration trajectory of the ball can be modeled as a damped oscillation, and its motion is governed by the following approximate equation

$$\frac{d^{2}x}{d{t}^2}+2\mathrm{\beta }\frac{\mathit{dx}}{\mathit{dt}}+{\mathrm{\omega }}_0^{2}x=0$$ (2)

Fig. 3. Characteristics and principles of RTE harvester.

(A) Schematic diagram of the motion model. (B) Simulation of motion displacement and photos captured by a high-speed camera. (C) Displacement variation results from small balls and glass fibers in different regions. (D) Transferred charge and peak splitting in a single cycle. (E) Impedance matching curve of RTE harvester. (F) Transferred current at 1 megohm and amplified single-cycle graph. (G) Voltage-charge (V-Q) curve for a single cycle.

The damping is primarily attributed to air resistance and thermal losses, with β < ω₀. Consequently, the solution can be approximated as follows

$$x=A{e}^{-\mathrm{\beta }t}\text{cos}(\mathrm{\omega }t+\mathrm{\phi })$$ (3)

The detailed process is outlined in text S3. The motion of the ball was simulated, and its movement was recorded using a high-speed camera (Fig. 3B and movies S1 and S2). In the absence of external excitation, the ball remains in contact with the lower electrode, leading to the transfer of electrons from the POM ball to the lower FEP layer. Upon the application of external impact, the spherical structure and spring system work synergistically to convert and transfer the shock energy to the fiberglass layers within the TENG, resulting in minor displacements and deformations, as shown in Fig. 3B-II. It is important to note that the deformation in the central region is more pronounced than at the edges. As a result, the balls in the center absorb more strain energy, leading to more vigorous motion in subsequent cycles. The displacement variations of the fiberglass and POM balls in both the central and edge regions are depicted in Fig. 3C. Subsequently, the internal balls are induced to bounce, as shown in Fig. 3B-III, generating a potential difference and driving charge transfer between the two electrodes. Figure 3B also captures the actual motion of the POM spheres using a high-speed camera. The displacement of the central spheres is notably greater than that of the peripheral spheres, although both regions exhibit similar motion trends and maintain consistent directional movement. The simulated motion process is illustrated in Fig. 3 (B-II to B-IX). Furthermore, the RTE harvester has a minimal impact on passing vehicles, as evidenced by the pressure simulation results on the wheels, which are provided in movie S3.

When the ball contacts the upper electrode, charge transfer occurs, as illustrated in Fig. 3D. The charge transferred in a single cycle reaches 0.8 μC. The transferred charge shows periodic increases and decreases. This is due to the resonance of the internal balls, with a resonance frequency of ~3 Hz. Near the 3-Hz frequency, the motion of the balls tends to synchronize, causing the output to exhibit periodic variation. A magnified view of the cycle reveals that the initial contact with the upper electrode results in a charge transfer of Q₁. Subsequently, the ball moves downward, making contact with the lower electrode before bouncing upward once more, generating another charge transfer, Q₂. As potential energy is gradually converted into electrical energy, the ball fails to reach the upper electrode again. Instead, after reaching a certain height, it returns to the lower electrode, repeating the cycle until it comes to rest. The subsequent charge transfers are denoted as Q₃ and Q₄. The relationships between these charge transfers are as follows

$${\mathrm{Q}}_1={\mathrm{Q}}_2+{\mathrm{Q}}_3+{\mathrm{Q}}_4$$ (4)

Because the motion process governs the conversion between mechanical and electrical energy and each vibration cycle involves a single contact with the upper electrode but multiple contacts with the lower electrode, asymmetric charge transfer occurs between the two electrodes, resulting in several transfer peaks. Each peak corresponds to a contact event between the sphere and an electrode. The asymmetry in charge transfer also leads to variations in the electrical energy generated during the upward and downward contacts (33, 34). Figure 3E presents the impedance matching results for the RTE harvester. When the internal resistance of the generator matches the external resistance, the output power reaches its peak. Although the internal resistance remains ~1 megohm regardless of whether the upper or lower electrode is contacted, the external output power differs substantially. The current generated upon contact with the upper electrode is higher, yielding a peak output power of 16.409 mW, whereas the current generated during contact with the lower electrode is lower, with a peak output power of only 1.458 mW (average power of 2.2 mW, as detailed in fig. S12).

To further explore the current characteristics, the RTE harvester was connected to a 1-megohm load at a frequency of 5 Hz. The current directions were found to be asymmetric, with the peak positive current reaching ~120 μA, while the peak negative current is around 40 μA. Expanding a single cycle, as shown in Fig. 3F, the current generated during contact with the upper electrode is large but brief, whereas the current during contact with the lower electrode is smaller but persists longer. This further supports the asymmetry in charge transfer, with a high, transient current peak observed during contact with the upper electrode and a series of decreasing, sustained current peaks occurring during contact with the lower electrode. The energy generated by the RTE harvester was calculated using the charge-voltage (Q-V) curve (35, 36). Figure 3G presents the energy generated during the contact process between the sphere and the upper and lower electrodes in a single cycle (additional output data for various external resistances are provided in fig. S13). The results show that under a 1-megohm external load, the energy generated during contact with the upper electrode is 49.722 μJ, accounting for 78.3% of the total output energy, while only 13.762 μJ is generated during contact with the lower electrode. Across different external loads, energy release during contact with the upper electrode consistently remains the dominant contributor to the total energy output.

RTE harvester array

The RTE harvester demonstrates high sensitivity in detecting various vibration conditions. Figure 4A shows the voltage signal generated by the RTE harvester at different amplitudes. As the amplitude increases, the output also increases proportionally (figs. S14 and S15). The linear relationship between amplitude and voltage is consistent across both low- and high-amplitude scenarios, with the sensing sensitivity being most pronounced at intermediate frequencies, as illustrated in Fig. 4B (fig. S16). Figure 4C presents the short-circuit current at different frequencies (1 to 50 Hz), demonstrating that the RTE harvester can effectively distinguish vibrations within this frequency range (figs. S17 and S18). Notable differences are observed in both the number of peaks and their respective peak values. It is noteworthy that the rate of change in the positive and negative currents varies. Both currents increase with frequency, but at lower frequencies, the increase in positive current is more pronounced, as shown in Fig. 4D. As the frequency increases, the number of contacts between the sphere and the lower surface decreases, causing the contact frequency with the lower surface to approach that of the upper surface. Consequently, at higher frequencies, the positive and negative currents tend to converge. Both amplitude and frequency increases result in higher output, as shown in Fig. 4E. For example, when the frequency is 26 Hz with an amplitude of 5 mm and at 5 Hz with an amplitude of 10 mm, the output currents are comparable. To mitigate this interference, we introduce two RTE harvesters in the subsequent sensing system for mutual correction, thereby ensuring sensing accuracy.

Fig. 4. Output and sensing performances of the RTE harvester.

(A) Open-circuit voltage at different amplitudes. (B) Relationship between amplitude and open-circuit voltage. (C) Short-circuit current at different frequencies. (D) Relationship between frequency and short-circuit current. (E) Short-circuit current curves under different frequencies and amplitudes. (F) Open-circuit voltage at different tilt angles. (G) RTE harvester powering ~2000 LEDs. (H) Comparison of output before and after 48 hours of immersion. h, hours. (I) Short-circuit current under different impact energies. (J) Management circuit and demonstration photos. SCR, Silicon Controlled Rectifier. (K) Effective energy collected by the RTE harvester array. (L) Charge-discharge cycle curve. R², coefficient of determination.

Figure 4F presents the output results at different tilt angles. The RTE harvester exhibits high sensitivity to angle variations, accurately detecting tilt angles as small as 5° (fig. S19). The vibration feedback allows for the detection of road damage caused by events such as landslides, seismic collapses, and traffic accidents, enabling timely maintenance to prevent further casualties and economic losses (3). In addition, the RTE harvester demonstrates exceptional energy storage performance. When excited by a shaker (frequency = 5 Hz, amplitude = 5 mm), a single RTE harvester can illuminate 1920 light-emitting diode (LED) lights, as shown in Fig. 4G (the LED connection method is detailed in fig. S20). Figure 4H compares the output before and after the RTE harvester is submerged in water for 48 hours. The outputs remain nearly unchanged, indicating the harvester’s strong environmental adaptability and stable operation in harsh weather conditions, such as heavy rain and flooding. Furthermore, the energy output of the RTE harvester under varying impact forces was tested using a drop hammer experiment. Figure 4I shows the short-circuit current at different impact energies, which increases substantially with external impact energy, reaching a maximum of 280 μA, thereby demonstrating excellent output performance (fig. S21).

Subsequently, we fabricated an RTE harvester array consisting of 10 units and integrated a management circuit to enhance energy utilization efficiency (The SCR stands for Silicon Controlled Rectifier) (37), thereby forming a self-powered intelligent system (Fig. 4J). As the number of units in the array increases, the effective power stored in the capacitor increases linearly, reaching 2.676 mW with 10 devices, as shown in Fig. 4K. The stored energy is used for real-time signal detection and wireless transmission between mobile devices. To simulate the impact caused by a vehicle passing over, we struck the top of the spherical shell twice with a heavy object, completing the discharge and stop-discharge process (Fig. 4L). The generated electrical signals are transmitted to a smartphone, with the experimental process demonstrated in movie S4. In addition, the effective energy stored in a capacitor was tested on a vibration platform (30 Hz). A 10-mF capacitor reaches 3 V after 2500 vibration cycles (fig. S22). The vibration frequency of the road surface is related to the actual road conditions and vehicle status, but it is generally within 50 Hz (38). Therefore, we selected two frequencies for the experiment: a lower frequency of 5 Hz and a higher frequency of 30 Hz.

Self-powered intelligent and connected transportation system

The SP-ICTS consists of an array of 12 RTE harvesters, integrated with a chip that transmits the collected data to the cloud, as shown in Fig. 5A. The circuit diagram and working principle are detailed in text S4 and fig. S23. Installing RTE harvesters into the roadway (text S5 and fig. S24), the first and last RTE harvesters are used for sensing, providing distinct signal feedback for different vehicles, while the eight central harvesters are dedicated to RTE collection. These central harvesters form a localized very microgrid, supplying energy to sensors along the road segment and ensuring self-sufficiency in energy supply for the road section. The SP-ICTS, based on RTE harvesters, can implement a five-in-one intelligent traffic network system, which includes self-powering, localization, perception, communication, and guidance, thereby achieving the integration of vehicles, roads, and cloud technologies (Fig. 5B).

Fig. 5. Self-powered intelligent and connected transportation system.

(A) Schematic diagram of the composition and relationships within the SP-ICTS. (B) Schematic diagram of the principle of SP-ICTS. (C) Comparison of the signal of SP-ICTS and laser vibrometer. (D) Current at different vehicle weights and speeds. (E) Signal results are generated at different vehicle speeds. (F) Energy values are generated under different traffic volumes based on the layout model. (G) Comparison with other vibration sensors. (H) Actual test results, model diagrams, and physical images, including schematic diagrams and actual vibration observations.

The device’s compressive strength reached 8.88 MPa, with a maximum load-bearing capacity of 70.7 kN (fig. S25A), which is generally higher than the maximum requirements for road construction worldwide (39–42). The corresponding compressive stress-strain curve is shown in fig. S25B. In addition, impact resistance is also a crucial factor affecting functionality in practical applications. Therefore, we conducted impact testing on the shell material at different temperatures. The results indicate that the device’s shell has an impact strength of 32.33 kJ/m² at room temperature (fig. S25C). Some impact strength can be maintained at low temperatures (14.66 kJ/m²) (fig. S25D), and the impact strength is substantially improved at higher temperatures (206.08 kJ/m²) (fig. S25E), which is higher than that of conventional road materials (43–45). To further validate the mechanical robustness, we subjected the device to 5000 compression cycles under a load of 5 MPa, as illustrated in fig. S25F. These additional experiments demonstrate that the device has sufficient mechanical durability, making it well suited for deployment in road surface environments.

We used a laser Doppler vibrometer for real-time vibration detection. The results show that vibration generation and the corresponding electrical signal are almost simultaneously produced (Fig. 5C), demonstrating that the RTE harvester has minimal time lag in its feedback signals. In addition, the SP-ICTS is capable of identifying vehicle speed, weight, and other parameters. As shown in Fig. 5D, the electrical signals generated by different vehicle speeds and weights exhibit marked differences, indicating that the SP-ICTS can provide timely feedback on abnormal driving behaviors, such as overloading or speeding.

To further assess the sensing stability and sensitivity of the RTE harvester, we conducted 100 drop-weight impact tests under varying impact energies (fig. S26A), followed by the same tests after immersing the device in water (fig. S26B). The resulting linear response curves were nearly identical to those obtained 1 year prior (Fig. 4I and fig. S26C), indicating exceptional long-term linearity and minimal performance degradation. These results confirm that the device maintains a stable linear output even after a year, with negligible influence from moisture or water exposure. To evaluate its real-world applicability, we constructed a simulated road testing environment (fig. S26D) and monitored the generated electrical signals in real time (fig. S26E and movie S5). To replicate diverse road conditions, we performed tests under ambient temperature, at 60°C (fig. S26F) and submerged them in water (fig. S26G and movie S6) and in dry ice environments (fig. S26H and movie S7). Figure S26I illustrates the variation in temperature and humidity throughout a day, alongside the signals produced at different times, vehicle speeds, and environmental settings. The data show a clear distinction in signals across different times of day (i.e., different environmental conditions) and vehicle speeds, while signals at the same speed remained highly consistent. This demonstrates the harvester’s high sensing sensitivity and excellent stability, with minimal susceptibility to environmental fluctuations such as temperature and humidity. In addition, a total of 150 vehicular passes were performed, and regression analysis of the resulting signals revealed a strong linear correlation between vehicle speed and output current (fig. S26J), with an estimated sensitivity of ~40 μA/km per hour. Even after 300 total passes (including both front and rear wheels), the surface of the device showed virtually no visible wear (fig. S26J). These findings highlight the harvester’s adaptability to road environments while maintaining high sensitivity and stable sensing performance.

However, to account for potential measurement errors—such as a fast and light vehicle producing a signal similar to that of a slow and heavy vehicle—we introduce two RTE harvesters for mutual correction of these discrepancies. Furthermore, the distance between the two sensors can be used to calculate the vehicle’s wheelbase, providing a more comprehensive reflection of the vehicle’s characteristics (e.g., model and size). As illustrated in Fig. 5E, this process generates three distinct signal outputs. The first signal is generated when the vehicle’s front wheel first contacts the first RTE harvester. The second signal is triggered when the front wheel contacts the last RTE harvester and the rear wheel contacts the first RTE harvester. Last, the third signal occurs when the rear wheel contacts the second RTE harvester. Here, ΔT represents the time elapsed from when the vehicle first contacts and then leaves the RTE harvester, while Δt represents the time the vehicle’s front and rear wheels take to contact the RTE harvesters. The calculations for the vehicle’s wheelbase and speed are as follows

$$v=\frac{S}{∆t}$$ (5)

$$L=v∆T-S$$ (6)

In this context, v represents the vehicle speed, S is the distance between the two sensors, and L refers to the vehicle’s wheelbase. Through mutual correction between the two sensors, precise measurements of the vehicle’s speed, weight, and wheelbase can be obtained, enabling the detection of road anomalies and illegal driving behaviors.

To estimate the energy generated under varying traffic flows, we developed a large-scale deployment model. They are closely arranged on the road with a spacing of 250 mm, covering a total length of 50 m (for a road with a width of 4 m, a total of 3200 units are laid over the 50-m length). This arrangement ensures that the wheels of the vehicle can pass over the RTE harvester, thereby ensuring the accuracy of the sensing (details of the model are provided in fig. S27). As shown in Fig. 5F, for a highway equipped with a 50-m RTE harvester, the model predicts that with a daily traffic flow of one vehicle, a 0.3-mW hygrometer can operate continuously for 1.8 hours. For a daily traffic flow of 10,000 vehicles, a 24-W camera can operate for 84 s (further details are available in table S3). A dense deployment of 50-m sections can support facilities such as Global Positioning System/Beidou navigation systems, traffic lights, and cameras within a 1-km range. This highlights the potential of RTE harvesters for providing a distributed energy supply along roadways. The RTE harvester demonstrates exceptional performance in both vibration energy harvesting and sensor sensitivity, as shown in Fig. 5G (46–59). In addition, the RTE harvester is easy to manufacture and cost-effective, as detailed in Table 1, making it well suited for industrial production and large-scale deployment.

Table 1. Comparison of the cost of different subgrade energy harvesting devices.

TEM, transmission electron microscopy.

Time	Type	Model	Cost (dollar/m2)
2020	TENG	OT-TENG (60)	318.78
2021	PV	SDPTES (61)	376.59
2022	TEG	RTEGS (62)	163.24
2022	PENG	URPG (63)	658
2022	PV	PIPVT (64)	476
2023	PENG	PSN-PZT (65)	1013.44
2024	PV	CPSP (66)	459.28
2024	PENG	PVDF/PZT/CNTs (67)	442.48
2024	TENG	EVA-PZT-PANI (68)	399.37
	This work	RTE harvester	71.3

Furthermore, we demonstrate the feasibility of the SP-ICTS by deploying a set of SP-ICTS units on an actual roadway. The experimental video is available in movie S8. The signals generated after vehicles pass over the RTE harvesters are monitored in real time using an electrostatic meter and transmitted to a mobile phone for signal reception. To observe the actual motion of the RTE harvesters following vehicle passage, we prepared a transparent casing and installed both RTE harvesters and cameras inside it (fig. S28). As a result (Fig. 5H), we captured the movement of the ball under various conditions: when the vehicle did not make contact, when the front wheels made contact, and when the rear wheels made contact (movie S9). In summary, the RTE harvester has the potential to fundamentally transform our understanding of road infrastructure, converting roads from passive elements into active contributors to energy ecosystems. This shift is poised to play an important role in advancing sustainable energy goals and fostering intelligent, eco-friendly, and green development.

DISCUSSION

We propose an innovative RTE harvester embedded in road infrastructure, using the freestanding layer mode TENG array that effectively converts low-grade RTE into electrical power. The RTE harvester demonstrates exceptional performance, achieving a peak power output of 16.409 mW and an average power of 2.2 mW from a compact 78-cm² road area under a single tire impact, with a maximum energy conversion efficiency of 11.723%. It exhibits high sensitivity to vibrations in the 0- to 50-Hz range, ensuring reliable energy generation across diverse traffic conditions. The RTE harvester array forms a very microgrid, with substantial potential to power distributed infrastructure such as sensors, cameras, and navigation systems. In addition, we have developed the SP-ICTS, an integrated system comprising RTE harvester arrays. The SP-ICTS offers five core functions: self-powering, positioning, communication, sensing, and guidance, and is capable of meeting the energy demands of the system over a 1-km stretch of highway. This system also facilitates real-time traffic monitoring and anomaly detection, proving effective in precise vehicle speed, weight, and wheelbase measurement. This work introduces an approach to efficient RTE harvesting and distributed power supply, paving the way for the integration of vehicle-to-road, vehicle-to-infrastructure, and vehicle-to-everything systems.

MATERIALS AND METHODS

Preparation of RTE harvester

The housing was fabricated using fiber-reinforced nylon via computer numerical control machining. Other components include a top wave spring (LISPRING, LM45-H1), durable rubber, and POM balls (4 mm in diameter). The connection parts are secured using threaded fasteners and adhesive, with overall dimensions of 100 mm by 140 mm (diameter by height). The internal TENG consists of a polychlorinated biphenyl circuit board with a fiberglass substrate, onto which a 50-μm-thick FEP film is applied. The board is thoroughly cleaned and dried using anhydrous ethanol before assembly. After assembly, the nodes are welded to connect the circuit and ensure structural integrity. In the multilayer structure, the bottom surface of each layer is interconnected to form one electrode, while the top surface of each layer is connected to form the opposing electrode. The internal TENG has dimensions of 85 mm by 80 mm (diameter by height).

Characterization and measurement

The surface morphology after 100,000 cycles was characterized using a field-emission scanning electron microscope (Hitachi, SU 8020). To assess its basic output performance, the TENG was subjected to vertical vibrations using a modal vibration table (YMC VT-500), driven by a linear power amplifier (YMC LA800). A sine wave signal was generated by a Stanford low distortion function generator (SRS Stanford, DS360). The typical electrical output from the TENG was measured using a programmable electrometer (Keithley, 6514). Real-time vibration measurements were performed using a miniature laser Doppler vibrometer (OmniSensing Photonics MotionGo, MV-HW-TR-U). Surface roughness measurements were performed using a step meter (DektakXT). Fatigue resistance measurements were performed using a hydraulic fatigue tester (USA-INSTRON8801). Compressive strength measurements were performed using an electronic universal testing machine (USA-MTS-E44.305). Impact strength measurements were performed using an impact tester (Zwick Zwick/Roell Amsler, Germany, HIT2000F).

Supplementary Materials

The PDF file includes:

Texts S1 to S5

Figs. S1 to S28

Tables S1 to S3

Legends for movies S1 to S9

Other Supplementary Material for this manuscript includes the following:

Movies S1 to S9