In the era of fifth-generation (5G) wireless communications where devices are heavily used, due to the great advances in the Internet of Things (IoT) and artificial intelligence (AI), perceivable devices with miniaturization and flexible regulations are facilitated for managing the wireless channels, and many interesting applications are therefore developed. Hence, in the physical level, the wireless channel and electromagnetic (EM) link between the transmitter and receiver with complex propagation internally play an important role in the future wireless communications.

On the other hand, surface plasmon polariton (SPP) in the optical frequency region¹ is a special kind of surface wave restricted to the interface between two media with opposite signs of permittivity. In order to realize spoof SPPs (SSPPs) in the terahertz and microwave bands, many efforts have been made^2,3. In particular, SSPPs transmission lines (TLs) with ultrathin substrate and metallic structures have been presented and attracted many interests^4,5. In recent years, lots of planar SSPPs devices have been proposed, such as low-loss transmission lines^6,7, filters^8,9, couplers^10,11, amplifier^12,13, backward phase matching¹⁴ and others¹⁵⁻¹⁷. The planar SSPPs have the characteristics of outstanding field-confinement, flexible dispersion regulation, low pass nature, wide band, low cost, and lightweight and compact properties, and therefore have great potentials in future wireless communication systems. Furthermore, some ultra-thin spoof metasurface, due to their modulation of the dispersion of structured light on the subwavelength scale, give rise to beam deflector¹⁸ and integrated imaging^19,20.

Specifically, many excellent characteristics of SSPP electromagnetic materials such as flexible design, greater slow-wave, EM coupling, low cost, low loss, low profile, etc., make it more conducive to the design of antennas, which can control the near field by the control unit arrangement to change the radiation effect in the far field. For example, as one type of beam-scanning antennas (BSA), leaky-wave antennas (LWA) with periodic radiation elements or continuous apertures^21,22 are widely used in radar and wireless communication systems for their low profile and easy integration into the surface of carriers²³⁻²⁶. Compared with the conventional feeding structures, the LWA based on the SSPP structure has the characteristics of simple and flexible design, more flexible bandwidth design, and more consistent gain in the frequency band, which can achieve the advantages of large scanning angle, wide working band and high gain^27,28. At present, most of the SSPP-LWAs could be divided into two types according to the radiation schemes: periodic modulation and EM coupling. For the first type, SSPP-LWA is modulated periodically in the waveguide structure to excite an infinite number of space harmonics and thus radiate^29,30. In contrast, the SSPP antennas based on EM coupling³¹⁻³³ are generally in the form of periodic radiation cells located close to the SSPP TL for energy coupling and radiation. Compared to the first type, the independent design of the feed SSPP TL and the radiation units greatly improves the degree of freedom of design, and can be designed with shorter antenna lengths to achieve higher gain.

Hence, the reconfigurable SSPP-LWA is very suitable for developing a new type of smart antenna because of the outstanding ability of dispersion regulation and high freedom of wave control. The field of smart antenna is constantly evolving, generally including the following three categories. First, beamforming antennas³⁴⁻³⁷ focus the signal by dynamically adjusting the antennas’ orientation, providing more powerful signal coverage and quality. However, complex hardware and signal processing requirements are required, which increases the system cost. Also, adaptability to dynamic environments may present some challenges. The second category, adaptive antennas^38,39, adjusts antenna parameters in real time to improve signal quality and combat signal interference, but the hardware and algorithms are more complex and require a lot of computing resources to support. The third one, multi-functional metasurfaces⁴⁰⁻⁴³ and reconfigurable intelligence surfaces⁴⁴⁻⁴⁷. The former achieves full-space functions by flexibly adjusting the phase or polarization of the unit, enabling capabilities like transmission-reflection-integrated^40,43, tunable holograms⁴², and optical sensing⁴¹. The latter is to optimize signal channels by adjusting surface reflection properties, and have great potentials to improve signal coverage and capacity. However, complex control and communication systems are usually required in this case. Recently, metasurfaces and sensors have been combined to enable a lot of work in intelligent EM control system⁴⁸⁻⁵¹. For example, metasurfaces combined with computer vision (CV) relies on cameras, data and algorithms to receive instructions or take actions such as beam tracking^52,53, remote sensing image⁵⁴, traffic surveillance system^55,56. Among them, intelligent metasurface systems⁵³ based on CV have been developed for EM tracking and real-time wireless communication of moving targets. EM characteristics of each unit cell in the metasurface is electrically tuned for the realization of real-time beams forming with high accuracy and flexibility. However, due to limitations in manufacturing complexity and power consumption, there exist some technical challenges in the intelligent metasurface systems. In view of the above, SSPP-LWA with self-reconfigurable operating frequency and auto switch of radiation state is especially required in spectrum analysis and wide-band communication applications. The SSPP antenna possess much less unit cells than the metasurface, and therefore is relatively simple to manufacture and easy to modulate.

In this paper, we propose a new approach combining reconfigurable SSPP-LWA with CV as a smart antenna system to realize the global management of wireless channels in the process of transmission, and automatically meet the needs of users in dynamic environment. By adding PIN diodes on the radiating element and adjusting the bias voltage of the PIN diode, the working state (e.g., the radiating state and the non-radiating state) is controlled dynamically within the frequency band. Experiments show that in the “radiating” state, the beam scanning range of the reconfigurable SSPP-LWA is −43°~5° in the frequency band from 8.5 to 13 GHz. Next, visual information is employed to assist SSPP-LWA in adjusting beam scanning. The proposed vision-based scheme utilizes advanced CV technology to acquire the position information of the user relative to the reconfigurable SSPP-LWA, sends the position information to the computer, and then feedbacks instructions to the FPGA and signal generator. We provide two experiment scenarios. In the first one, the camera is used to detect whether there is a target in the specified range, and the adaptive real-time switching between the radiation state and the non-radiation state is realized accordingly. In the second scenario, the SSPP-LWA provides feedback on the base station side according to the target position, and supplies the desired frequency instruction in line with the direction towards the selected target, thereby realizing beam tracking.

The proposed reconfigurable SSPP-LWA combined with CV can perceive the moving targets through target detection, and then dynamically adjust the use of spectrum resources. In wide-band communication, the SSPP-LWA can adjust the operating frequency according to this information to avoid spectrum conflict and improve spectrum utilization. In the optimization of energy efficiency, combined with the real-time perception of CV, the smart SSPP-LWA can more intelligently select the operating mode and frequency in wide-band communication to optimize the energy efficiency of the communication system, reduce power consumption. Also, the smart SSPP-LWA can switch the antenna working state according to the presence of the target and realize target beam tracking in real time, enhance physical layer communication security.

The schematic of the proposed intelligent system is presented in Figs. 1 and 2, which is composed of the reconfigurable SSPP-LWA and an RGB-Depth camera. The moving target is represented by a model car or the target user (e.g., here a broadband horn antenna is used as the target user for signal reception) moving along a certain path. The images of the car (or the broadband horn antenna) are taken by an Intel RealSense Depth Camera D435i (RS-Camera) located on the SSPP-LWA at the rate of 40 FPS (frames per second), and each image is selected by the convolutional neural network (CNN) based on YOLOv4-tiny. The original image is firstly scaled to [608, 608, 3] when reasoning, and then input to the CSPDarknet53-tiny network for the feature extraction. After passing through five groups of convolution and pooling layers, the feature maps with three different dimensions are obtained, which are fused by the network and YOLO layers. We optimize the network to not only detect the target, but also collect its position and its elevation and azimuth angles in the coordinates of the RS-Camera.

The moving object is tracked dynamically and its position information is refreshed in real-time by the RS-Camera, with each refresh followed by a voltage control sequence that feeds the FPGA or by the corresponding frequency signal that feeds the signal generator connected to the reconfigurable SSPP-LWA. The input voltage and frequency of the reconfigurable SSPP-LWA is obtained through the computer using the feedback information, and the instruction are constantly refreshed. In the first experimental scenario, by the voltage that feedback from detecting whether there are external objects within a specific range, the smart SSPPs LWA can realize the adaptive real-time switching between the radiation state and the non-radiation state. In the second experimental scenario, the smart SSPP-LWA is carefully designed to transmit a frequency-dependent scanning beam towards the moving target with regard to its changing positions. Among them, the coordinate systems of the camera and the reconfigurable SSPP-LWA (as they are closely located) are unified to ensure the accuracy of the position.

The theoretical analysis in this section includes two parts, one is reconfiguration of state switching in the operating frequency band, and the other is frequency-controlled beam scanning of the SSPP-LWA when PIN diode is turned on.

Firstly, the radiation principle of the SSPP-LWA is analyzed. The dispersion curve of SSPPs TL is taken as the starting point, as shown in Fig. 3. It can be seen that the dispersion curve for the fundamental mode of SSPPs TL (the red curve with n=0) is located on the right side of the light line, that is, SSPPs present a slow-wave property and cannot be radiated directly. Radiation of EM waves to free space requires two steps: the first step is coupling from the SSPPs TL to the radiation patch, and the second step is radiating through the patch unit into free space. Theoretical calculations of the radiation pattern are given in Supplementary information Section 1 for details. According to the Floquet theory analysis, periodic placement of radiation unit cells near SSPPs TLs can excite an infinite number of spatial harmonics:

$k_n=k_{\text{spp}}+\frac{2n\pi }{d}\mathrm{,}n=\mathrm{0,}\pm \mathrm{1,}\pm 2\cdots$(1)

where k_spp is the propagation constant of the n = 0 harmonic (the fundamental mode), and d is the distance between two adjacent radiation unit cells. In Fig. 3, n represents the n-th harmonic. Partly harmonics with n=−1 and n=−2 are located within the fast wave region of the optical cone and can be radiated. To be noted, if n≤−2 space harmonics are designed to radiate fast waves, then −1 spatial harmonics can also radiate⁵⁷. In order to obtain pure mode radiation, in practice −1-th harmonic is usually selected for radiation. Here the radiation frequency band is between the frequency points of A and B in Fig. 3. According to Eq. (1), it is found that the radiation frequency band is closely related to the distance between the two radiation unit cells, and hence one can calculate the distance d between the two radiation unit cells according to the required working frequency band. A specific frequency point f₀ is designed as the normal radiant point, at this time k₋₁ is 0, so $d=2\pi /\left(k_{\text{spp}}|f_0\right)$, where $\left(k_{\text{spp}}|f_0\right)$ represents the propagation constant of the fundamental wave at the specific frequency f₀, it can be see that the period between the radiation unit is consistent with the wavelength of the SSPPs at the specific frequency f₀, and we can obtain $\left(k_{\text{spp}}|f_{\text{0}}\right)$ through Fig. 3, thus the distance between two radiation unit cells can be designed.

The detailed structure of reconfigurable SSPP-LWA is shown in Fig. 4. An SSPPs TL with miniaturization, strong field binding and high EM energy density is expected. Therefore, an SSPPs TL with “T”-shaped units was designed through modifying the unit’s height. The rectangular structures on both sides of the commonly used H-shaped SSPP units were bent to both sides respectively, and the “T”-shaped unit structure was finally determined, as shown in Fig. 4(c). For validations, an SSPPs TL is fabricated, and its performance has been verified, maintaining a higher transmission coefficient and lower reflection coefficient. Please see Supplementary information Section 2 for details.

The switchable radiating/non-radiating working state is caused by the change of the state of the PIN diodes embedded in the radiating patch, which are mounted on either side of the ELC resonant structure, as shown in Fig. 4(c). The topological circuit of the PIN diode, that is, how to position the PIN diode, is critical for the working bandwidth and gain isolation of the two working states. In addition, as shown in Fig. 4(d), the structure of the middle capacitor is thicker than usual, which helps to optimize the influence between the two operating states. For simulated transmission coefficient of the reconfigurable SSPP-LWA with different number of diodes, please see Supplementary information Section 3 for details. The PIN diodes are embedded in the approximate middle of both sides of the radiating element and are connected in series, in which the distance p_d between the two diodes is 1.1 mm. In addition, the via hole on the radiation patch close to the TL are connected to the positive pole of on the back of the dielectric substrate, and the side length of the square copper plate is 2 mm. The via hole at the other end of the radiation patch is connected to the ground. Four 1 nH inductors are placed at the ground terminal, with a distance of 0.5 mm between each inductor, and finally connected to a 2.55 mm × 1.5 mm copper sheet for feeding.

The switching of the working state of the antenna is controlled by the bias voltage of the PIN diode. When the PIN diode is turned on, the antenna is in the radiating state; and when it is turned off, the antenna is in the transmission (non-radiating) state. In the topology design of the PIN diode of the patch, the isolation of switching states and the working bandwidth should be weighed. Series switches have the advantage of low transmission loss in broadband but at the price of low isolation. Parallel switches are usually used in combination with quarter-wavelength TLs, which are narrow in nature, but have higher isolation than series switches. In addition, using two or more PIN diodes in series mode can improve the isolation of one PIN diode, and the same bias current can be used in series connection to save power. In this work, in order to realize broadband performance, high isolation and low-power dissipation, two PIN diodes are designed in series. (Please see Supplementary information Section 4 for the equivalent circuit of the PIN diode used).

The reconfigurable SSPP-LWA redistributes the current distribution on the antenna by switching the diode between “ON” and “OFF” states, and so as to control the coupling state between the radiation units and the SSPPs TL. The antenna adopts the SSPP TL based on “T”-shaped unit structure as the feed, uses periodically loads radiation units on both sides of the TL to stimulate the higher harmonic for radiation, and selects the −1 order harmonic for single mode radiation. The less-lossy dielectric substrate of the antenna is Taconic TLY-5 (ε_r = 2.2, tanδ = 0.0009) with a thickness of 0.762 mm, and the copper thickness is 0.018 mm. The overall size of the designed SSPP-LWA is 269×63.24 mm².

The commercial software, Computer Simulation Technology (CST) Microwave Studio was used to conduct time-domain full-wave simulation of the above antenna. Figure 5(a) plots the simulated reflection coefficient (S₁₁) and transmission coefficient (S₂₁). The impedance matching performance of the proposed SSPP-LWA is measured by the Agilent vector network analyzer (VNA) in Fig. 5(b). It can be seen that when the PIN diode is interrupted (OFF of PIN diodes), S₁₁ is low and S₂₁ is high from 8.5 GHz to 13 GHz, and the “transmission” state is realized. In contrast, when the PIN diode is switched ON, the simulated and measured S₂₁ decreased in 9–13 GHz and S₁₁ remains below −10 dB. This result is due to the fact that in this case most energy is successfully coupled to patches and radiated to free space, and the radiation units and the SSPPs TL as a whole work in the “radiation” state. The observed return loss and insertion loss are quite low, indicating a good matching and a good radiation efficiency. To be noted, there is an error between the test result and the simulation result, which is related to the loss of dielectric and active devices, and the insertion loss of cables and SMA connectors in the high-frequency test environment.

To verify the proposed model, we plot the simulated local current distribution of the coupling part and radiation unit in Fig. 5(c, d). When the radiation unit is placed near the TL of SSPPs, there is a current distribution on each radiation unit due to EM coupling. When the PIN diode is switched on, the radiation unit corresponds to the position where the magnetic field of the TL is strong, and the current distribution on the radiation unit forms a closed loop with the corresponding part of the TL, forming a magnetic coupling, as shown in Fig. 5(c). At this time, the current distribution of the overall structure of the radiation unit is along the Y-axis, indicating that the radiation unit is equivalent to an electric dipole and radiates EM waves to the free space. On the other hand, when the PIN diode is disconnected, the current distribution of the radiation unit and its corresponding part of the TL does not flow in one direction as a whole, but presents two closed-loop flows in opposite directions, which is equivalent to two magnetic dipoles whose energy cancels each other out. At the same time, there will be four effective pores on the radiation unit. The current distribution is mainly concentrated in the two effective pores below and their current distribution direction is opposite, which is equivalent to the magnetic dipole in the opposite direction, and the coupled electromagnetic energy basically cancels each other, as shown in Fig. 5(d).

There are also some similar studies on the principles of switching or modulating radiation and transmission of waves. Firstly, metasurfaces can convert spatially incident CP beams into spoof surface plasmons (SSP) with the same wave vector through phase gradients in the THz regime⁵⁸. Regarding LWAs, a millimeter-wave beam steerable slot array antenna using an Inter-Digitated Capacitor (IDC) based corrugated substrate integrated waveguide (CSIW) acts as a normal waveguide when the PIN diodes are unbiased, but when a segment’s PIN diodes of IDC-CSIW are activated, it tends to divert the field toward inductive fingers and away from its adjacent longitudinal slot, which then reduces the power radiated from that slot, hence achieving beam reconfigurability⁵⁹. Other modulation types also include adjusting the period length of the antenna by toggling the operation states of the patches between “ON” and “OFF” via PIN diodes biasing to achieve the beam-scanning antenna⁶⁰⁻⁶².

In view of this, when the PIN diode is disconnected, the EM energy in the radiation unit does not radiate out, and most of the energy is transmitted, that is, the antenna changes from the state of leakage wave radiation to the state of transmission. In Fig. 5(e−l), the distributions of simulated and measured near electric field test results are plotted. It is worth emphasizing that because of the constraints imposed by the actual test conditions, the phase of the output signal received by the VNA cannot be calibrated based on the input signal generator during the scanning process, resulting in only amplitude values of the electric field being measured. Here, we combine simulated results with the measured ones to present the reconfigurable SSPP-LWA field distribution in the radiation and the non-radiation states. It can be observed that when the field is distributed in a radiating state at 9 GHz and 12 GHz, the energy gradually decreases along the waveguide of the reconfigurable SSPP-LWA. In contrast, in the non-radiating state, the field distribution is clearly transmitted along the TL, which intuitively demonstrates the field variation in the two states. For field distribution at frequencies other than those mentioned above, please see Supplementary information Section 5 for details. In the design, the loss from active components (the diode) are quite noticeable, followed by substrate loss, connectors, and test environment. The designed sample reflects the scenario after accounting for most influencing factors. Our future goal is to further refine the design so as to achieve a lower total loss.

For more visualized perception of the beam scanning capability, Fig. 6(a−c) shows the simulated 3D radiation patterns in different beam scanning angles. When the frequency increases, the beam scans towards different angles, and a backward to forward beam scanning can be achieved. The amplitudes of electric fields at different frequencies of the structure are also simulated by CST Microwave Studio. It can be clearly observed from Fig. 6(a−c) that, whether it is the three-dimensional far-field pattern or the amplitudes of electric fields at different frequencies of the antenna, it can be seen that as the frequency increases, the beam is directed at different angles, and backward to forward beam scanning can be achieved. At the same time, from near-field simulation results, it can be seen that the amplitude of the EM wave transmitted along the SSPP TL is significantly lower, indicating the leakage of energy.

To gain a more vivid understanding of the beam scanning capability, simulated results for the frequency scanning of reconfigurable SSPP-LWA with two working states are shown in Fig. 6(d, e). The polar plot shows that the beam scans from −43° (at 8.5 GHz) to 5° (at 13 GHz) in the relative central coordinate system. Fig. 6(f) shows that the realized gain of the reconfigurable SSPP-LWA remains in the range of 11.2–7.293 dBi throughout the working frequency band of 8.5 GHz to 13 GHz. The average gain in the radiated state is 7.3 dBi higher than that in the non-radiated state. Through the comparison of the total radiated power in the two states of reconfigurable SSPP-LWA, we can also see that the two states have excellent switching effect in Fig. 6(g).

To verify the above theoretical analysis and demonstrate the good performance, the prototype of the proposed SSPP-LWA is fabricated and measured. The radiation patterns are tested in an anechoic chamber environment, as shown in Fig. 7(a). The front and back enlarged photos of the prototype are shown in Fig. 7(b, c). Figure 7(d−g) shows the measured radiation patterns of the proposed beam-scanning antenna from 8.5 to 13 GHz with co-polarization and cross-polarization under two working states. The measured beam angle scans from −44° to 4.7°, whereas the measured gain ranges from 8.36 to 4.4 dBi. It can be seen that the beam steers from backward to forward with scanning ranges of −44° at 8.5 GHz, 34° at 9 GHz, −12° at 11 GHz, and 4.7° at 13 GHz, respectively. The measured results are given as a validation of the design. These gain drops are mainly caused by the error of the fabrication and measurement system, especially, the coaxial feeder and active device loss and radiation effects from the turntable base lead to amplitude drop. The gain curves of the co-polarization and cross-polarization measurements are given in Fig. 7(h, i). The radiation gain is lower when the diode is OFF (transmission state), and in the case of cross polarization. The average gain of the ON and OFF operating states is 6.9 dBi and 1.18 dBi, respectively. And the gain of each frequency point can be read in the Fig. 7(h). The radiation pattern results and gain curves of two states also show that the designed SSPP-LWA has good state switching and frequency beam scanning characteristics.

In wireless communication, radar system can detect and track targets in the complex EM environment. However, due to the high cost, large size and complex design, other technologies are urgently needed to make up for it. Fortunately, with the fast development of computer vision technology, intuitive, reliable, and cost-effective target detection and tracking become possible in many application scenarios. Moreover, AI-enabled computer vision (CV) technology is evolving rapidly, which can solve more complicated problems and serve as an aid to intelligent communications. Here, it is designed as a perceptible innovative technology, with the proposed smart SSPP-LWA system to realize an intelligent perceptible platform that combines perception and communication.

The externally perceptive platform is shown in Fig. 8(a), which consists of an Intel RealSense Depth Camera D435i (RS-Camera), control system (a computer) and Field-Programmable Gate Array (FPGA). First, the images of the toy car are taken by a RS-Camera located on the reconfigurable SSPP-LWA at the rate of 40 FPS (frames per second), and each image is selected by the convolutional neural network (CNN) based on YOLOv4-tiny. Here, we use the toy car to represent the detection target of the scene. The RS-Camera collects moving targets, and the control system connects with the camera, feeds back the position of the target, and then transmits the corresponding voltage sequence to the FPGA.

In the test scene, we use a toy car instead of a detector or jammer as the target, and the field of view in Fig. 8(a) that can be captured by the camera is the area where the target is moving, the field of view is the range of angles the camera can see, including depth information. When we move the car into or out of the target area, the antenna can quickly change the non-radiation/radiation state accordingly. In addition, the smart SSPP-LWA is connected to Agilent Vector Network Analyzer N5230C (VNA), and the state of the smart SSPP-LWA can be observed through the S-parameter in VNA. The process of the car appearing and disappearing from the field of view is shown in Fig. 8(c, d). The lower right corner of each figure is the S-parameter curve shown in VNA. When the car moves to the blue area that can be captured by the RS-camera, the car will be marked. At this time, the FPGA output level is low, the PIN diode on the smart SSPP-LWA is disconnected and the antenna is in the transmission (non-radiation) state. The FPGA in the system has a 20-channel 3.3 V analog output port, each driving four diodes per radiation patch. Outputs can be switched via a serial port tool, allowing independent power supply for each radiation unit. In far-field testing without active control, a DC power source under 5 V and a current limit of 40 mA can be used as an alternative and can operate stably for a long time. The state of the antenna can be verified through the S-parameter feedback in VNA, which corresponds to the transmission state. When there is no toy car appears in the field of view, the FPGA outputs high level at this time, and the PIN diode on the smart SSPP-LWA is switched on, the antenna is in the radiation state, accompanied by low S₁₁ and S₂₁ in VNA. Please see Supplementary Movie 1 for details. The working states are switched by the SSPP-LWA under the control of the real-time varying coding sequence to keep tracking the target at a sampling speed of 0.2 s (see Supplementary information Section 8 for details). The speed limit lies in the sampling speed of the RS-Camera.

It can be observed that the S-parameter is very consistent with the measured results in Fig. 5(b). Therefore, in the external perceptive platform of this scene, the smart SSPP-LWA realizes automatically control its working state, which can achieve wireless communication security in the airspace to a certain extent.

We have carried out the intelligent system experiment of beam tracking. By detecting the location of the receiver (a horn antenna) with a RS-Camera, we can send the frequency instruction of the corresponding beam direction to the signal generator (the SSPP LWA), so as to realize the closed-loop operation of the system without human instruction. The system consists of a signal generator (SA, Keysight E8257D), a signal analyzer (SG), the SSPP-LWA, a RS-Camera, a computer and a receiver (a dual-ridged horn antenna, 750 MHz to 18 GHz). The SG is connected to the LWA, the receiver is connected to the SA, and the computer controls the RS-Camera and SG respectively, as shown in Fig. 9(a, b). In order to more accurately show the power acceptance of the SSPP-LWA at a certain distance, we have carried out the near-field test of the SSPP-LWA, please see Supplementary information Section 7 for details.

Firstly, we tested the accepted energy values for all frequencies without the camera working. It can be seen that the receiving antenna is fixed on a slide rail and bisects ten moving points in a horizontal movement of 1.5 m. Each point position measures the received energy at different emission frequencies. The signal generator transmits 9–13 GHz EM waves and records them every 0.2 GHz. The task of the vision-aided SSPP-LWA is to accurately radiate the EM waves to the horn antenna at the receiver side based on the visual information. In test progress, as for the vision-based beam tracking scheme, we use object detection and CV technology to obtain the position information of the horn antennas with respect to the SSPP-LWA. Then beam tracking is implemented in two stages. In the first stage, since we have actually tested the radiation status of SSPP-LWA, we can determine the correspondence of the angle with regard to the frequency. In the second stage, according to the coordinates of the target user, SSPP-LWA can obtain the desired beam direction and select the corresponding codeword in the precoder to achieve beam tracking. We can get the direction of the predicted bounding box of the target user at a certain moment through the visual algorithm, as shown in Fig. 9(a).

During testing, we move the target user within coverage areas that include SSPP-LWA beamforming. To highlight the effectiveness of vision, we plot in one diagram with visual aids and in all situations. When there is no visual aid, the traditional frequency scanning method, in which the feedback procedure is ignored, traverses the received power at each frequency point at all locations. When there is visual assistance, the strategy of beam tracking is: first use RS-Camera to feedback the position of the target user in the field of view, and when the user is outside the beam scanning range, the computer does not send instructions to the SG. And when the user is within the coverage range of SSPP-LWA beamforming, the computer sends the frequency instruction of the response to SG according to the returned angle information. The SSPP-LWA’s operating frequency is adjusted from 8.5 GHz to 13.1 GHz to allow the platform to operate at a sufficient scanning range. It can be seen that the blue five-star mark records the receiving power value of different positions in Fig. 10, and through the change of frequency, it is very stable to maintain a high receiving power, ensuring the high quality of communication. The range of the test is beyond the beam-scanning area, but SSPP-LWA still retains the scattered beam with low gain in the region of more than 5° (Moving more than 100 cm), where SSPP-LWA still executes the command, but the sensitivity decreases due to the decreased gain value.

At the same time, there is no additional feedback process and beam training overhead throughout the communication. For traditional passive devices, it is basically difficult to achieve beam tracking and dynamic switching. The smart system adjusts the operating frequency according to the information of target detection in the process of beam scanning to avoid spectrum conflict, improve spectrum utilization, and reduce overhead.

In this paper, we propose a smart SSPP-LWA with external perceptibility, which can realize the adaptive real-time switching between the radiation state and the non-radiation state and beam tracking by regulating the coupling unit. In two intelligent communication scenarios, the SSPP-LWA obtains the real-time location information of the external jammer or target user through the CV, switches the working states accordingly, and realizes automatic tracking of moving target by adjusting the input frequency. Leveraging the capabilities of beam reconfiguration and external perception, the proposed SSPP-LWA serves as a prime contender for smart systems, offering the ability to autonomously select optimal operating modes and intelligent communication in dynamic, complex environments.