Low-power dynamic coverage optimization method and system based on distributed intelligent metasurface

Abstract

The application discloses a low-power dynamic coverage optimization method and system based on distributed intelligent metasurface. The system comprises three low-power 2-bit intelligent metasurfaces and their control boards, a binocular camera and a matching host computer control system. The system uses the binocular camera to capture image data in real time, uses a deep learning target detection algorithm to identify human targets in the image and calculate the relative coordinates of the targets in three-dimensional space. According to the grid cell coordinates of the targets, the corresponding intelligent metasurface code sequence optimized by a conditional sample mean algorithm is indexed out, and the code sequence is converted into a voltage control signal of the intelligent metasurface through the control board with an FPGA as a main controller chip, so that the distributed intelligent metasurface realizes real-time dynamic enhancement of wireless signal coverage in the area where the targets are located. The system uses visual perception and distributed coding optimization technology to realize real-time positioning of target users and dynamic coverage optimization of wireless signals, and ensures the acceptance signal quality of the target users.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of signal coverage enhancement system design based on intelligent metasurfaces, which mainly includes a low-power dynamic coverage optimization method and system based on distributed intelligent metasurfaces. BACKGROUND

[0002] With the deployment of 5G networks and the popularity of Internet of Things applications, users have higher expectations for the coverage and quality of wireless communication. However, in various environments such as urban areas with high-rise buildings, vast rural fields, complex indoor or underground spaces, etc., it is still a difficult problem to achieve comprehensive and high-quality network coverage. Physical obstacles often cause signal attenuation or blockage, forming "blind areas" that existing base station networks cannot cover. In addition, with the increasing scarcity of spectrum resources, operators are forced to use higher frequency bands, such as millimeter waves. The penetration and coverage capabilities of these high-frequency signals are significantly lower than those of traditional frequency bands, exacerbating the signal coverage problem. Therefore, how to expand the signal coverage range while maintaining network performance has become a key problem in the field of wireless communication that needs to be solved.

[0003] To address the challenges in the field of wireless communication, researchers have explored various deployment strategies for small base stations and heterogeneous networks. By densely deploying small base stations in urban environments, researchers can effectively fill in the blind areas of network coverage and significantly improve the spatial utilization efficiency of the network. At the algorithm level, various signal attenuation and coverage optimization algorithms have been developed based on genetic algorithms, particle swarm optimization (PSO), deep learning, etc. These algorithms aim to optimize network configuration and resource allocation, striving to achieve the best balance between signal coverage and performance.

[0004] However, although the individual cost of small base stations is lower than that of traditional base stations, the total cost of large-scale deployment and subsequent maintenance is still relatively high. In addition, although optimization algorithms can achieve optimal network performance, they often require a large amount of computing resources, especially in real-time processing or large-scale applications. Therefore, there is an urgent need for an economical and efficient solution to address these challenges.

[0005] In recent years, intelligent metasurface technology has shown great application potential due to its unique advantages. Metasurfaces are artificial surfaces composed of numerous sub-wavelength scattering units. By adjusting the active devices embedded in these units, intelligent metasurfaces can intelligently manipulate incident electromagnetic waves, including changing their reflection direction, focusing, and diffusion. Reasonable use of intelligent metasurface technology can effectively enhance signal penetration and coverage in complex environments, thereby solving the problem of signal blind areas and providing a novel solution to expand network coverage. However, due to the inherent Brewster angle limitation of metasurfaces, a single metasurface may not be able to cover the entire space in some complex non-line-of-sight scenarios.

[0006] To overcome this limitation, this invention introduces a distributed smart metasurface strategy. By strategically deploying multiple smart metasurfaces, full-area signal coverage can be optimized in complex environments. In this setup, signal coverage no longer relies solely on the beam control capabilities of a single smart metasurface. The collaborative operation between smart metasurfaces not only effectively extends signal coverage but also enhances the network's anti-interference capabilities. However, the deployment of multiple smart metasurfaces inevitably leads to a significant increase in power consumption. Therefore, effectively reducing the power consumption of the distributed deployment scheme becomes a critical consideration.

[0007] Building upon this foundation, a low-power dynamic coverage optimization system based on distributed intelligent metasurfaces is proposed. To address the low-power requirements of distributed deployment, this scheme designs a metasurface unit with operating power consumption in the microwatt range. Then, the conditional sample mean algorithm is used to optimize the blind beamforming codebook of the intelligent metasurface within a given space, utilizing different coding sequences of the codebook to enhance signal coverage in corresponding regions. Finally, visual tracking technology is combined to achieve dynamic coverage optimization of spatial signals. Summary of the Invention

[0008] The purpose of this invention is to design a low-power dynamic coverage optimization system based on a distributed intelligent metasurface, which features low power consumption, high gain, wide coverage, and real-time signal enhancement.

[0009] To solve the above-mentioned technical problems, the specific technical solution of the present invention includes the following four steps:

[0010] Step 1: Design a low-power smart metasurface unit with 2-bit phase response and beamforming effect in the 5GHz band, determine the composition architecture of its array, and rationally design the FPGA driver control board and its control strategy for the smart metasurface.

[0011] Step 2: Use the three smart metasurfaces to build a distributed smart metasurface system, and use the conditional sample mean codebook optimization algorithm to calculate the optimal code for each region in this scenario.

[0012] Step 3: Build a visual perception system to recognize human targets in conjunction with the host computer. At the same time, convert the target location information into the corresponding code of the area in real time and transmit it to the FPGA driver control board to regulate the intelligent metasurface.

[0013] Step 4: Build a video wireless transmission platform using Software-Defined Radio (USRP) to visually demonstrate the system's real-time signal coverage enhancement for the target user in non-line-of-sight scenarios through video transmission.

[0014] Furthermore, step 1 specifically includes the following steps:

[0015] The designed low-power smart metasurface consists of 8×12 metasurface cells with 2-bit quantization phase. The basic pattern of each cell is composed of an irregular metal patch and two feed rectangular patches. To improve the oblique incidence stability of the cell, a row of metal vias is placed on each side of the cell along the polarization orthogonal direction. The stack-up structure of the cell is relatively simple, including a top metal pattern layer, a bottom all-metal backplane layer, and an F4B dielectric substrate layer in between.

[0016] To achieve dynamic control of the reflected electromagnetic wave phase, active tunable electronic components need to be embedded in the basic structure of the unit. Commonly used tunable electronic components include PIN diodes and Schottky diodes, but these devices typically require 5mA of current to conduct, resulting in high power consumption when forming large-scale arrays. The metasurface unit designed in this invention reduces power consumption by embedding a CMOS-based single-pole single-throw (SPST) switch chip in a carefully designed location. To achieve higher beam control accuracy, a unit with 2-bit quantized phase is designed. By using two chips with different state combinations, four reflection phase states with a 90-degree phase difference are achieved, defined as "00", "01", "10", and "11". At the center frequency of 5GHz, the reflection coefficient amplitude of the four coded states is less than -1.5dB, ensuring good reflection efficiency of the metasurface. At 5GHz, the reflection phases of the four coded states differ by 90 degrees sequentially, achieving four discrete phase states within a 360-degree range.

[0017] Furthermore, step 2 specifically includes the following steps:

[0018] The conditional sample mean codebook optimization algorithm used in this invention differs from traditional beam pattern coding optimization schemes in that it is based on the statistical characteristics of a large number of measured received signal energies. It performs coding optimization for specific environments without channel estimation, also known as blind beamforming. The following section provides a fundamental theoretical analysis of the mathematical model of a controlled channel on a smart metasurface under general conditions. Typically, the transmitter-to-receiver channel coefficient is defined as...

[0019]

[0020] in

[0021] β n ∈(0,1),α n ∈[0,2π) (2)

[0022] β n α represents the amplitude of the channel coefficient. n This represents the phase of the channel coefficients. Specifically, h0∈C represents the sum of other channels from transmission to reception that do not pass through the smart metasurface; this can also be called the background channel.n ∈C, n=1,2,...,N then represents the cascaded channel from emission to reception for the nth metasurface unit. Define vector

[0023] θ=(θ1,...,θ N (3)

[0024] θ N The set of possible values ​​for the reflection phase of the Nth metasurface unit is:

[0025] Φ K ={ω,2ω,...,Kω}, (4)

[0026] in

[0027]

[0028] ω represents the minimum interval of phase changes in the set, and K is the number of elements in the set. Let X∈C represent the transmitted signal with an average power of P, i.e.

[0029] E[|X| 2 ]=P, (6)

[0030] The received signal Y∈C can then be calculated as

[0031]

[0032] Where Z~CN(0,σ) 2 ) is used to simulate additive background noise, σ 2 Let represent the variance of the random distribution. Therefore, the accepted signal-to-noise ratio model can be expressed as:

[0033]

[0034] Specifically, the reference signal-to-noise ratio is defined as

[0035]

[0036] Therefore, the improvement in signal-to-noise ratio can be expressed as follows:

[0037]

[0038] At this point, the optimization objective is clear: to find the vector θ that maximizes the improvement in signal-to-noise ratio.

[0039]

[0040] However, solving the above problem presents two challenges. First, numerically optimizing the discrete variable θ is very complex; second, from a practical perspective, obtaining the channel coefficients {h0,...,h...} is challenging. NThe cost of this approach is high. The conditional sample mean training method cleverly avoids these two difficulties by calculating the mean of randomly encoded conditional samples to obtain the θ that satisfies the target. This method first generates a set of T randomly encoded data samples, letting... It indicates that it is a subset of him, in which

[0041] Q nk ={t:θ nt =kω} (12)

[0042] Let kω be the index of the nth metasurface unit. Then, calculate the conditional sample mean of the reflection phase for each metasurface unit, using the following formula:

[0043]

[0044] This conditional sample mean essentially reflects the performance of each metasurface unit at θ. n The average performance when θ = kω, therefore it is obvious that each θ can be chosen based on this value. n ,Right now

[0045]

[0046] This invention employs a blind beamforming coding method primarily for two reasons: First, in the distributed smart metasurface deployment scenario discussed in this invention, significant multipath effects and complex channel models are often present. In such an environment, the blind beamforming coding optimization method can fully utilize electromagnetic waves along various propagation paths in space, converging sufficient energy at a designated location. Second, regarding the joint control method of distributed smart metasurfaces, as the number of smart metasurfaces increases, traditional beamforming methods require multiple coordinate transformations between the smart metasurfaces and the target to obtain the electromagnetic wave manipulation angle for each smart metasurface, resulting in high computational complexity and high requirements for positional accuracy. Therefore, using conditional sample mean (SRM) to simultaneously optimize the coding of multiple smart metasurfaces is more convenient and efficient than individual beamforming. Research on various blind beamforming coding algorithms reveals that when the number of quantized phases of the smart metasurfaces is no greater than 4, the SRM algorithm performs well in both optimization speed and signal-to-noise ratio improvement. The number of smart metasurfaces used in this invention precisely meets the application conditions of the SRM algorithm; therefore, this method is chosen as the basis for dynamic coverage optimization.

[0047] Furthermore, step 3 specifically includes the following steps:

[0048] Currently, the hardware foundation for distributed intelligent metasurfaces and the conditional sample mean algorithm for optimizing blind beam coding exist. However, to achieve dynamic signal coverage optimization, a target tracking platform capable of real-time feedback of target position information is still needed. This invention constructs a platform based on mature visual perception technology. The main hardware device of the platform is a binocular camera, and the software algorithm is based on the YOLO-v5 framework. This device uses dual cameras to simulate the human visual system, capturing image information. By comparing slightly offset images captured by the two cameras, it perceives three-dimensional structural information. By continuously processing each frame of image information, it can locate the three-dimensional coordinate position of the target user in real time. Finally, using the codebook pre-trained in step 3, the encoding of the corresponding position is indexed and sent to the driver board of each metasurface unit to update the status, realizing dynamic signal coverage optimization of the target position, simplifying the process as follows: Figure 5 As shown. In this process, the host computer generates an encoding matrix, which is synchronously distributed to the FPGA control board of each metasurface unit through a switch. The control board maps the received encoding to the voltage driving each metasurface unit, thereby jointly regulating the complex electromagnetic waves in space.

[0049] Furthermore, step 4 specifically includes the following steps:

[0050] To visually demonstrate the capabilities of the proposed low-power dynamic coverage optimization system, this invention constructs a software-defined radio-based wireless video transmission platform. The quality of video transmission clearly reflects the real-time received signal strength in the corresponding area, effectively proving the effectiveness of the proposed system. Three smart metasurfaces, a binocular camera, a transmitting antenna, and an omnidirectional receiving antenna are placed in fixed positions. Placing the transmitting antenna outside the laboratory wall demonstrates that the system functions even in NLoS scenarios. In this experiment, the receiving antenna remains stationary while the target moves through different areas of space, and changes in the communication quality at the receiving end are observed. Two typical locations are selected for illustration. When the target within the green box moves to the receiving antenna area, the signal quality in that area significantly improves, video transmission is smooth, and the symbols in the constellation diagram cluster together. However, when the target moves out of the receiving antenna area, the received signal-to-noise ratio at the antenna location becomes too low due to the distributed smart metasurface system directing the main electromagnetic energy to the target location, resulting in a high bit error rate and degraded video transmission quality. Furthermore, experimental results show that the proposed system has the characteristic of tracking a specified target, which can include multiple targets. This characteristic has great potential and scalability in future multi-target, multi-smart metasurface scenarios. Attached Figure Description

[0051] Figure 1The structural design and performance of the metasurface unit in an embodiment of the present invention are as follows: (a) exploded view; (b) top view; (c) side view; (d) reflection amplitude of each coded state; (e) reflection phase of each coded state.

[0052] Figure 2 Experimental prototypes and measurements for embodiments of the present invention: (a) measurement environment; (b) fabricated metasurface array prototype; (c) power consumption measurement; (d) normalized reflection amplitude; (e) normalized reflection phase; (f) far-field scattering patterns for different beam directions.

[0053] Figure 3 This is a schematic diagram of blind beam coding training in a distributed intelligent metasurface deployment, which is an embodiment of the present invention.

[0054] Figure 4 The following are examples of blind beam coding training performance under distributed smart metasurface deployment in the implementation of this invention: (a) the relationship between the number of smart metasurfaces and the received power; (b) the relationship between the number of training samples and the optimization effect; and (ce) the received power heatmaps for no smart metasurface, one smart metasurface, and three smart metasurfaces.

[0055] Figure 5 This is an example of the operation process of a signal tracking platform deployed in a distributed intelligent metasurface environment, which is an implementation case of the present invention.

[0056] Figure 6 Wireless communication experiments for embodiments of the present invention: (a) experimental scenario; (b) communication performance when the target is within the receiving area; (c) communication performance when the target is outside the receiving area. Detailed Implementation

[0057] To better understand the purpose, structure, and function of this invention, the following detailed description of the low-power dynamic coverage optimization system based on distributed intelligent metasurfaces is provided in conjunction with the accompanying drawings.

[0058] To solve the above-mentioned technical problems, the specific technical solution of the present invention includes the following four steps:

[0059] Step 1: Design a low-power smart metasurface unit with 2-bit phase response and beamforming effect in the 5GHz band, determine the composition architecture of its array, and rationally design the FPGA driver control board and its control strategy for the smart metasurface.

[0060] Step 2: Use the three smart metasurfaces to build a distributed smart metasurface system, and use the conditional sample mean codebook optimization algorithm to calculate the optimal code for each region in this scenario.

[0061] Step 3: Build a visual perception system to recognize human targets in conjunction with the host computer. At the same time, convert the target location information into the corresponding code of the area in real time and transmit it to the FPGA driver control board to regulate the intelligent metasurface.

[0062] Step 4: Build a video wireless transmission platform using USRP, and use a low-power dynamic coverage optimization system based on distributed intelligent metasurfaces to achieve real-time signal coverage enhancement for targets in non-line-of-sight scenarios.

[0063] Furthermore, step 1 specifically includes the following steps:

[0064] The designed low-power smart metasurface consists of 8×12 metasurface units with 2-bit quantization phase, such as... Figure 1 As shown in (b), the basic pattern of each cell consists of an irregular metal patch and two feed rectangular patches, with the following geometric parameters: L1 = 20mm, L2 = 0.89mm, L3 = 11.32mm, W1 = 20mm, W2 = 3.8mm, W3 = 2.9mm, W4 = 2.47mm, W5 = 6.2mm, P1 = 6.62mm, P2 = 6.25mm, H = 5mm, R = 0.6mm. To improve the oblique incidence stability of the cell, a row of metal vias is placed on each side of the cell along the polarization orthogonal direction. The stack-up structure of this cell is relatively simple, consisting of a top metal pattern layer, a bottom all-metal backplate layer, and an F4B dielectric substrate layer (ε) between them. r =2.65andtanδ=0.001).

[0065] To achieve dynamic control of the reflected electromagnetic wave phase, active tunable electronic components need to be embedded in the basic structure of the unit. Commonly used tunable electronic components include PIN diodes and Schottky diodes, but devices such as PIN diodes typically require 5mA of current when turned on, resulting in high power consumption when forming large-scale arrays. This study designs a metasurface unit that reduces power consumption by embedding a CMOS-based single-pole single-throw (SPST) switch chip in a carefully designed location. This switch chip has a bias voltage of +2.5V in the on state and -2.5V in the off state. Furthermore, the chip consumes only 10uA of current in the on state, with each chip consuming only 25uW. Compared to PIN diodes, power consumption is reduced by three orders of magnitude. To achieve higher beam control accuracy, this invention designs a unit with 2-bit quantized phase. By using chips with two different state combinations, we achieve four reflection phase states with a 90-degree phase difference, defined as "00", "01", "10", and "11".

[0066] This invention underwent full-wave simulation in CST Microwave Studio. By jointly simulating the passive structure of the unit and the S-parameter file of the switching chip, the reflection amplitude and phase of the unit under different encoding states were obtained. Figure 1 (d) It can be seen that at the center frequency of 5 GHz, the reflection coefficient amplitudes of the four coding states are all less than -1.5 dB, which ensures the good reflection efficiency of the metasurface. Figure 1 As shown in (e), at 5 GHz, the reflection phases of the four coding states are successively 90 degrees apart, realizing four discrete phase states within a 360-degree range.

[0067] To verify the reliability of the designed metasurface performance, a low-power metasurface array was fabricated using standard printed circuit board (PCB) technology, with an effective size of 160 × 240 mm. 2 Furthermore, a custom driver circuit board with an FPGA as the main controller was developed to ensure flexible beam manipulation. The reflection amplitude, phase, and far-field beam pattern of the metasurface were measured using a vector network analyzer (Agilent N5245A) in a microwave anechoic chamber. Figure 2 As shown in (a), the transmitting antenna is placed in the far-field region of the metasurface to achieve approximate plane-wave excitation when measuring the far-field scattering pattern. The fabricated sample image and element details are as follows. Figure 2 As shown in (b). Because the interface between the metasurface array and the driver circuit board was considered in the design layout, the two boards can be easily integrated for use, facilitating experiments. Furthermore, the power consumption of the metasurface array under full-power operation was measured, such as... Figure 2 As shown in (c), the multimeter current readings indicate that when all chips on the metasurface array are turned on, the total current consumption is 2.27 mA, and the average current consumption per low-power chip is 11 µA, consistent with the theoretical design. This low power consumption ensures the feasibility of deploying a distributed intelligent metasurface in the future.

[0068] Figure 2 (d) and Figure 2(e) shows the measured amplitude and phase of the metasurface reflection coefficient, normalized using measurements from a metal plate of the same size as the metasurface array. The curves indicate that at 5 GHz, the measured amplitude of the designed metasurface reflection coefficient is approximately 2 dB lower than the simulation results. This difference is attributed to potential differences in dielectric constant and substrate loss during manufacturing. Based on the measured phase data, the four coded states maintain a 2-bit quantized phase within the operating bandwidth. Overall, the measured amplitude and phase of the metasurface reflection coefficient are substantially consistent with the simulation results. Furthermore, this invention designs several phase codes for different beamforming directions, covering angles from 10° to 40°, measured at 10° intervals. The measured far-field scattering patterns are shown below. Figure 2 As shown in (f), the excellent beam control capability of the smart metasurface is demonstrated. However, it can be observed that the limited array size during fabrication results in some sidelobes in the reflected beam. Overall, the direction of maximum beam gain confirms the reliability of the designed metasurface and lays the foundation for further use of distributed smart metasurfaces for dynamic signal coverage.

[0069] Furthermore, step 2 specifically includes the following steps:

[0070] The conditional sample mean codebook optimization algorithm used in this invention differs from traditional beam pattern coding optimization schemes in that it is based on the statistical characteristics of a large number of measured received signal energies. It performs coding optimization for specific environments without channel estimation, also known as blind beamforming. The following section provides a fundamental theoretical analysis of the mathematical model of a controlled channel on a smart metasurface under general conditions. Typically, the transmitter-to-receiver channel coefficient is defined as...

[0071]

[0072] in

[0073] β n ∈(0,1),α n ∈[0,2π) (2)

[0074] β n α represents the amplitude of the channel coefficient. n This represents the phase of the channel coefficients. Specifically, h0∈C represents the sum of other channels from transmission to reception that do not pass through the smart metasurface; this can also be called the background channel. n ∈C, n=1,2,...,N then represents the cascaded channel from emission to reception for the nth metasurface unit. Define vector

[0075] θ=(θ1,...,θ N (3)

[0076] θ NThe set of possible values ​​for the reflection phase of the Nth metasurface unit is:

[0077] Φ K ={ω,2ω,...,Kω}, (4)

[0078] in

[0079]

[0080] ω represents the minimum interval of phase change in the set. Let X∈C represent the transmitted signal with average power P, i.e.

[0081] E[|X| 2 ]=P, (6)

[0082] The received signal Y∈C can then be calculated as

[0083]

[0084] Where Z~CN(0,σ) 2 ) is used to simulate additive background noise, σ 2 Let represent the variance of the random distribution. Therefore, the accepted signal-to-noise ratio model can be expressed as:

[0085]

[0086] Specifically, the reference signal-to-noise ratio is defined as

[0087]

[0088] Therefore, the improvement in signal-to-noise ratio can be expressed as follows:

[0089]

[0090] At this point, the optimization objective is clear: to find the vector θ that maximizes the improvement in signal-to-noise ratio.

[0091]

[0092] However, solving the above problem presents two challenges. First, numerically optimizing the discrete variable θ is very complex; second, from a practical perspective, obtaining the channel coefficients {h0,...,h...} is challenging. N The cost of this method is very high. The conditional sample mean training method cleverly avoids these two difficulties by calculating the mean of randomly encoded conditional samples to obtain the θ that satisfies the target. This method first generates a set of T random encoded data samples, letting... It indicates that it is a subset of him, in which

[0093] Q nk ={t:θ nt=kω} (12)

[0094] Let kω be the index of the nth metasurface unit. Then, calculate the conditional sample mean of the reflection phase for each metasurface unit, using the following formula:

[0095]

[0096] This conditional sample mean essentially reflects the performance of each metasurface unit at θ. n The average performance when θ = kω, therefore it is obvious that each θ can be chosen based on this value. n ,Right now

[0097]

[0098] This invention employs a blind beamforming method primarily for two reasons: First, in the distributed smart metasurface deployment scenario discussed in this invention, significant multipath effects and complex channel models are often present. In such an environment, the blind beamforming coding optimization method can fully utilize electromagnetic waves along various propagation paths in space, converging sufficient energy at a designated location. Second, regarding the joint control method of distributed smart metasurfaces, as the number of smart metasurfaces increases, traditional beamforming methods require multiple coordinate transformations between the smart metasurfaces and the target to obtain the electromagnetic wave manipulation angle for each smart metasurface, resulting in high computational complexity and high requirements for positional accuracy. Therefore, using conditional sample mean (SRM) to simultaneously optimize the coding of multiple smart metasurfaces is more convenient and efficient than beamforming alone. Through research on various blind beamforming algorithms, it can be found that when the number of quantized phases of the smart metasurfaces is no greater than 4, the SRM algorithm performs well in terms of optimization speed and signal-to-noise ratio improvement. The number of smart metasurfaces used in this invention precisely meets the application conditions of the SRM algorithm; therefore, this method is chosen as the basis for dynamic coverage optimization.

[0099] A schematic diagram of the distributed intelligent metasurface blind beam coding training experiment is shown below. Figure 3 As shown, the experimental space was divided into 20 equal-area regions, each numbered, and smart metasurfaces were deployed at multiple locations in the environment. To accurately measure the received power at the antenna location, a horizontally polarized omnidirectional antenna with a center operating frequency of 5 GHz was designed as the receiving antenna. The antenna was placed at the designated location to receive controlled electromagnetic waves emitted from all directions in the space. Then, a spectrum analyzer was used to feed the signal power value back to the host computer in real time, thereby obtaining a set of sample data for the designated location. By repeating the above steps a predetermined number of times, a fairly large dataset was obtained. Then, using this dataset, a blind beamforming code suitable for that location was calculated using the conditional sample mean algorithm.

[0100] The received power of the same region without any smart metasurfaces, with one smart metasurface deployed, and with three smart metasurfaces deployed was measured and compared. Figure 4 As shown in (a), measurement data indicates that three smart metasurfaces are more effective than one smart metasurface in enhancing the received power at a given location. Of course, deploying a single smart metasurface also improves the received power compared to having no smart metasurface. Furthermore, this invention analyzes the impact of different dataset sizes on system performance improvements. Figure 4 The fitted curve in (b) shows that when the sample size is less than 150, the received power increases significantly with the increase of the sample size. However, when the sample size exceeds 150, the improvement effect tends to stabilize. Figure 4 (c) through 4(e) show the received power heatmaps for 20 divided regions under three different deployment schemes. The overall color change allows for a more intuitive observation of the significant improvement in the electromagnetic environment resulting from the deployment of distributed smart metasurfaces.

[0101] Furthermore, step 3 specifically includes the following steps:

[0102] We now have the hardware foundation for distributed intelligent metasurfaces and the conditional sample mean algorithm for optimizing blind beam coding codebooks, but to achieve dynamic signal coverage optimization, we still need a target tracking platform that can provide real-time feedback on target location information.

[0103] This invention constructs a signal tracking platform based on mature visual perception technology. The platform's main hardware device is a binocular camera, and the software algorithm is based on the YOLO-v5 framework. This device uses dual cameras to simulate the human visual system, capturing image information. By comparing slightly offset images captured by the two cameras, it perceives depth and three-dimensional structural information, continuously processing each frame of image information to locate the target's three-dimensional coordinates in real time. Finally, using the pre-trained codebook from the previous section, the encoding at the corresponding positions is indexed and sent to the driver boards of each metasurface unit to update the state, achieving dynamic signal coverage optimization at the target location, such as... Figure 5 The simplified process is shown below. In this process, the host computer generates an encoding matrix, which is synchronously distributed to the FPGA control boards of each metasurface unit through a switch. The control boards map the received encoding onto the voltage driving each metasurface unit, thereby jointly regulating the complex electromagnetic waves in space.

[0104] Furthermore, step 4 specifically includes the following steps:

[0105] To visually demonstrate the capabilities of the proposed low-power dynamic coverage optimization system, this invention constructs a software-defined radio-based wireless video transmission platform. The quality of the video transmission clearly reflects the real-time received signal strength in the corresponding area, thus effectively proving the effectiveness of the proposed system. Three smart metasurfaces, a binocular camera, a transmitting antenna, and an omnidirectional receiving antenna are placed in fixed positions, such as... Figure 6 As shown in (a), placing the transmitting antenna outside the laboratory wall demonstrates that the system functions even in NLoS scenarios. In this experiment, as the target moved through different regions of space while the receiving antenna remained stationary, changes in the communication quality at the receiving end could be observed. Two typical locations were chosen for illustration, as shown in [examples omitted]. Figure 6 (b) and Figure 6 As shown in (c), when the target within the green box moves to the receiving antenna area, the signal quality in that area significantly improves, video transmission is smooth, and the symbols in the constellation diagram cluster together. However, when the target moves out of the receiving antenna area, the received signal-to-noise ratio at the antenna location becomes too low due to the distributed intelligent metasurface system directing the main electromagnetic energy to the target location, resulting in a high bit error rate and degraded video transmission quality. Furthermore, experimental results demonstrate that the proposed system has the capability to track a specified target, which can include multiple targets. This capability holds great potential and scalability for future multi-target, multi-intelligent metasurface scenarios.

Claims

1. A low-power dynamic occlusion optimization method based on distributed intelligent metasurfaces, characterized in that: Includes the following steps: Step 1: Design a low-power smart metasurface unit with 2-bit phase response and beamforming effect in the 5GHz band, determine the composition architecture of its array, and rationally design the FPGA driver control board and its control strategy for the smart metasurface. Step 2: Use several of the aforementioned smart metasurfaces to build a distributed smart metasurface system, and use the conditional sample mean codebook optimization algorithm to calculate the optimal code for each region in this scenario; Step 3: Build a visual perception system to recognize human targets in conjunction with the host computer. At the same time, convert the target location information into the corresponding code of the area in real time and transmit it to the FPGA driver control board to regulate the intelligent metasurface. The intelligent metasurface consists of 8×12 metasurface units with 2-bit quantization phase. The basic pattern of each unit consists of an irregular metal patch and two feed rectangular patches. The geometric parameters of the unit are: L1=20mm, L2=0.89mm, L3=11.32mm, W1=20mm, W2=3.8mm, W3=2.9mm, W4=2.47mm, W5=6.2mm, P1=6.62mm, P2=6.25mm, H=5mm, R=0.6mm. To improve the oblique incidence stability of the unit, a row of metal vias is placed on each side of the unit along the polarization orthogonal direction. The stacked structure of the unit includes a top metal pattern layer, a bottom all-metal backplate layer, and an F4B dielectric substrate layer between them. An active tunable electronic component is embedded in the basic structure of the metasurface unit to achieve dynamic control of the phase of the reflected electromagnetic wave; the tunable electronic component is a single-pole single-throw switch chip based on CMOS technology; the bias voltage of the switch chip in the on state is +2.5V, and the bias voltage in the off state is -2.5V. The metasurface unit uses two switching chips with different state combinations to realize four reflection phase states with a 90-degree phase difference, defined as "00", "01", "10" and "11"; at 5 GHz, the reflection phases of the four coded states are successively 90 degrees apart, realizing four discrete phase states within a 360-degree range. In step 2, the optimal code for each region in the scenario is calculated using the conditional sample mean codebook optimization algorithm. The specific steps include: dividing the experimental space into multiple regions of equal area, numbering each region, and deploying smart metasurfaces at multiple locations in the environment. Using a horizontally polarized omnidirectional antenna with a center operating frequency of 5 GHz as the receiver, the received power at the antenna location is measured. The antenna is placed at a designated location to receive modulated electromagnetic waves from various directions in space, and the signal power value is fed back to the host computer in real time using a spectrum analyzer, thereby obtaining a set of sample data for the designated location. The above steps are repeated a predetermined number of times to obtain a dataset, and then the blind beam coding suitable for the location is calculated using the conditional sample mean algorithm on this dataset.

2. The low-power dynamic coverage optimization method based on distributed intelligent metasurfaces according to claim 1, characterized in that: The visual perception system includes a binocular camera and a target tracking host computer capable of providing real-time feedback on target position information. In step 3, the host computer is used to identify human targets and simultaneously convert the target position information into the corresponding code for that region in real time, which is then transmitted to the FPGA driver control board to regulate the intelligent metasurface. Specifically, this includes: using a deep learning target detection algorithm, the binocular camera captures image information and perceives three-dimensional structural information, continuously identifying and locating the three-dimensional coordinates of the target in real time; using a pre-trained blind beam codebook, the host computer indexes the code at the corresponding position and distributes it synchronously to the control boards of each intelligent metasurface through a switch. The control board maps the received code to the voltage driving each metasurface unit, jointly regulating the complex electromagnetic waves in space to achieve dynamic signal coverage optimization for the target position.

3. The low-power dynamic coverage optimization method based on distributed intelligent metasurfaces according to claim 2, characterized in that: The deep learning-based object detection algorithm is a visual perception algorithm based on the YOLO-v5 framework.

4. The low-power dynamic coverage optimization method based on distributed intelligent metasurfaces according to claim 1, characterized in that: The number of intelligent metasurfaces is no more than 4.

5. The low-power dynamic coverage optimization method based on distributed intelligent metasurfaces according to claim 1, characterized in that: It also includes step 4: building a video wireless transmission platform through software radio, and intuitively demonstrating the system's real-time signal coverage enhancement for the target user in non-line-of-sight scenarios through video transmission.

6. A low-power dynamic occlusion optimization system based on a distributed intelligent metasurface, used to implement the method described in any one of claims 1-5, characterized in that, The system comprises several low-power 2-bit smart metasurfaces and their control boards, a binocular camera, and a supporting host computer control system. The system uses the binocular camera to capture image data in real time and employs a deep learning-based target detection algorithm to identify human targets in the images and calculate their relative coordinates in three-dimensional space. Based on the coordinates of the grid cell containing the target, the system indexes the corresponding smart metasurface encoding sequence optimized by the conditional sample mean algorithm and converts it into a voltage control signal for the smart metasurface via a control board with an FPGA as the main controller chip. This enables real-time dynamic enhancement of wireless signal coverage in the target area by the distributed smart metasurface.

Images