Beam steering optimization method and system for millimeter wave networks

Abstract

The present invention discloses a beam steering optimization method and system for millimeter wave networks, relating to the technical field of beam steering optimization, and comprising the following steps: using a DTW algorithm and a dynamic time warping algorithm to match the multi-hop path delay characteristics between a base station and a user device, and screening the optimal reflection path sequence; obtaining the path topology, constructing a quadratic unconstrained binary optimization model and solving it through quantum annealing to obtain a discrete phase distribution; calculating the quantization error of the discrete phase and using an interpolation algorithm to compensate in the frequency domain to obtain a phase compensation vector; loading the compensation vector into programmable metasurface hardware to form a reflection link, and constructing a phase error covariance matrix; incrementally correcting the weight parameters of the optimization model based on the matrix; finally, combining the corrected weights and the compensation vector to construct an objective function to achieve channel capacity optimization, thereby solving the problems of difficult multi-hop path selection, insufficient phase error compensation, and high complexity of beam optimization in millimeter wave communications, and improving channel capacity and communication quality.

Description

Technical Field

[0001] The present invention relates to the technical field of beam steering optimization, and more particularly, to a beam steering optimization method and system for millimeter wave networks. Background Art

[0002] In the development of current fifth-generation (5G) and future sixth-generation (6G) mobile communication systems, millimeter-wave communication technology has become a research hotspot due to its extremely high bandwidth and transmission speed. However, millimeter-wave signals face severe path loss, weak penetration, and susceptibility to obstruction during transmission. In complex urban environments, direct path obstruction is extremely common, significantly limiting communication stability and service coverage.

[0003] To overcome these physical limitations, beam steering technologies based on reflection paths and programmable metasurfaces have emerged in recent years. Programmable metasurfaces are a new type of artificial structure that can dynamically adjust the phase, amplitude, and direction of incident electromagnetic waves, thereby compensating for the shortcomings of direct channels by constructing multi-hop reflection links. However, key challenges remain: how to select the optimal path among multiple reflection paths, how to construct the optimal beamforming strategy, and how to address the discrete errors and control biases introduced by actual hardware.

[0004] Existing technologies often use heuristic search or traditional optimization algorithms to configure metasurface states, but these methods suffer from low solution efficiency, suboptimal performance, and difficulty adapting to dynamic channel changes. Furthermore, most approaches ignore the impact of discrete hardware characteristics on overall beam steering accuracy, resulting in performance degradation in practical deployments and failing to fully realize the potential of metasurfaces in millimeter-wave communications.

[0005] In view of the above problems, the present invention proposes a solution. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a beam control optimization method and system for millimeter wave networks. By combining dynamic time warping path matching, quantum annealing optimization model solution and frequency domain error compensation, high-precision beam control and channel capacity optimization based on programmable metasurfaces in millimeter wave networks are realized to solve the problems of difficulty in multi-hop reflection path selection, insufficient phase discrete error compensation and high complexity of beam optimization in millimeter wave communications, which make it difficult to improve channel capacity and communication quality.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The beam control optimization method for millimeter wave networks includes the following steps: using the DTW algorithm to match the multi-hop path delay characteristics between the base station and the user equipment to screen the optimal reflection path sequence; obtaining the topological structure of the optimal reflection path sequence, constructing a quadratic unconstrained binary optimization model and performing quantum annealing to solve it to obtain a discrete phase distribution; calculating the quantization error eigenvalue of the discrete phase distribution, and using an interpolation algorithm to compensate the quantization error eigenvalue in the frequency domain to obtain a phase compensation vector; loading the phase compensation vector into a programmable metasurface hardware unit to obtain a first reflection link, and constructing a phase error covariance matrix; incrementally correcting the weight parameters of the quadratic unconstrained binary optimization model based on the phase error covariance matrix; constructing an objective function based on the corrected weight parameters and the phase compensation vector, and reconstructing the channel capacity through a gradient projection method to obtain the optimal channel capacity.

[0009] In a preferred embodiment, the DTW algorithm is used to match the multi-hop path delay characteristics between the base station and the user equipment to screen the optimal reflection path sequence. Specifically, the delay sequence data of several candidate reflection paths between the base station and the user equipment are collected, and a path delay feature library is constructed; based on a preset reference delay template, the DTW algorithm is used to align the delay sequence of each candidate reflection path and calculate the minimum cumulative path distance; and several candidate reflection paths are screened according to the minimum cumulative path distance to obtain the optimal reflection path sequence.

[0010] In a preferred embodiment, the topological structure of the optimal reflection path sequence is obtained, a quadratic unconstrained binary optimization model is constructed and quantum annealing is performed to obtain a discrete phase distribution. Specifically, the activation state variables and phase adjustment variables of the programmable metasurface hardware unit are determined according to the topological structure of the optimal reflection path sequence; the activation state variables and phase adjustment variables are used as input, and a quadratic unconstrained binary optimization model is constructed based on preset activation state constraints and phase discretization constraints, with maximum signal strength as the optimization goal; the quadratic unconstrained binary optimization model is mapped to the Ising model of the quantum annealing machine, and the discrete phase distribution is obtained by using the quantum annealing algorithm to solve.

[0011] In a preferred embodiment, the quantization error characteristic value of the discrete phase distribution is calculated, and the quantization error characteristic value is compensated in the frequency domain by an interpolation algorithm to obtain a phase compensation vector, specifically: the difference between the discrete phase distribution and the preset continuous phase distribution is calculated, and the quantization error value of each programmable metasurface hardware unit is calculated based on the difference; the quantization error value is Fourier transformed, and the frequency domain characteristic component is extracted to obtain a quantization error characteristic vector; the error characteristic vector in the frequency domain is spectrally analyzed, and an adaptive fluctuation detection method is used to identify missing frequency bands; the frequency bands of the quantization error characteristic vector are dynamically divided by a fuzzy clustering algorithm to obtain several first frequency bands; the ratio of the mean and standard deviation of the quantization error characteristic vector of each first frequency band is calculated, and the node spacing density of the cubic spline interpolation is dynamically adjusted according to the ratio; the missing frequency bands are adaptively interpolated and compensated according to the adjusted node spacing density, and the frequency domain compensation result after interpolation is inversely transformed to the time domain to generate a phase compensation vector.

[0012] In a preferred embodiment, the phase compensation vector is loaded into the programmable metasurface hardware unit to obtain a first reflection link, and a phase error covariance matrix is ​​constructed. Specifically, the phase response parameters of the programmable metasurface hardware unit are adjusted according to the phase compensation vector to obtain the first reflection link; the deviation value between the actual phase response and the preset first phase response is measured through the first reflection link to construct a phase error sample set; based on the phase error sample set, the phase error covariance value between each programmable metasurface hardware unit is calculated to obtain the phase error covariance matrix.

[0013] In a preferred embodiment, the weight parameters of the quadratic unconstrained binary optimization model are incrementally corrected based on the phase error covariance matrix, specifically: the phase error covariance matrix is ​​input into the quadratic unconstrained binary optimization model as a regularization term; the weight parameters of the quadratic unconstrained binary optimization model are iteratively updated using a gradient descent method with an adaptive step size until the error converges to obtain the corrected weight parameters.

[0014] In a preferred embodiment, the objective function is constructed based on the corrected weight parameters and phase compensation vector, and the channel capacity is reconstructed by the gradient projection method to obtain the optimal channel capacity, specifically: taking the corrected weight parameters and phase compensation vector as input, a multi-objective optimization function is constructed; the multi-objective optimization function is iteratively optimized by the gradient projection method to calculate the maximum channel capacity that meets the preset power constraints; the first channel capacity value is initialized, and the gradient direction of the multi-objective optimization function under the current parameters is calculated; the gradient direction is projected to the feasible solution space, and the first channel capacity value is dynamically updated; the iteration is repeated until the increment of the first channel capacity value is less than the preset first threshold value to obtain the optimal channel capacity.

[0015] The technical effects and advantages of the beam steering optimization method and system for millimeter wave networks of the present invention are as follows:

[0016] 1. This invention achieves precise modeling and beamforming of complex channel states in high-frequency communication environments through a multi-level, multi-stage joint design involving a dynamic time warping algorithm, multi-hop reflection path modeling, quantum annealing, frequency domain compensation, and gradient projection optimization. Compared to traditional methods, this invention not only effectively matches the optimal reflection path between the base station and the user, significantly improving signal coverage and stability, but also solves the large-scale combinatorial optimization problem inherent in phase control by constructing a quadratic unconstrained binary optimization model and combining it with quantum annealing. Furthermore, a frequency domain interpolation algorithm is used to compensate for discrete phase quantization errors, significantly improving phase control accuracy. The system further constructs a phase error covariance matrix based on actual measurements and incrementally modifies the optimization model weight parameters based on this matrix, enhancing the model's robustness to hardware uncertainties. Finally, a gradient projection method is used to achieve channel capacity reconstruction under multi-objective constraints, thereby improving the overall communication performance of the system. While ensuring high computational efficiency, this invention effectively balances accuracy, adaptability, and engineering feasibility. It is particularly suitable for next-generation millimeter-wave communication systems in ultra-dense deployments and dynamically changing scenarios, possessing significant practical value and industrial application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 Schematic diagram of the flow of the beam steering optimization method for millimeter wave networks according to the present invention.

[0018] Figure 2 Schematic diagram of the structure of the beam steering optimization system for millimeter wave networks according to the present invention. DETAILED DESCRIPTION

[0019] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0020] Example 1, Figure 1 The present invention provides a beam steering optimization method for a millimeter wave network, comprising the following steps:

[0021] S1, uses the DTW algorithm to match the multi-hop path delay characteristics between the base station and the user equipment to select the optimal reflection path sequence;

[0022] In this example, the DTW algorithm is used to match the multi-hop path delay characteristics between the base station and the user equipment and select the optimal reflection path sequence. Specifically,

[0023] Collect delay sequence data of several candidate reflection paths between the base station and the user equipment, and build a path delay feature library;

[0024] Based on the preset reference delay template, the DTW algorithm is used to align the delay sequence of each candidate reflection path and calculate the minimum cumulative path distance;

[0025] Several candidate reflection paths are screened according to the minimum cumulative path distance to obtain the optimal reflection path sequence.

[0026] For example, in a millimeter wave communication environment, there are multiple reflection paths between the base station (BS) and the user equipment (UE). The system uses a ranging module or time of arrival (ToA) estimation technology to collect the delay sequence on each candidate reflection path. For example, there are three reflection paths as follows:

[0027] The delay sequences for the three candidate reflection paths of Path A are 2.1ns, 2.3ns, and 2.5ns, respectively; the delay sequences for the four candidate reflection paths of Path B are 1.9ns, 2.0ns, 2.4ns, and 2.6ns, respectively; and the delay sequences for the three candidate reflection paths of Path C are 2.0ns, 2.4ns, and 2.7ns, respectively. These delay sequence data are used to construct a path delay feature library. Furthermore, the system sets a set of reference delay templates T representing ideal path characteristics, specifically 2.0ns, 2.4ns, and 2.7ns, representing the propagation delay characteristics under an ideal reflection topology. These templates can be obtained based on historical environmental data, manually set, or derived from a model. Finally, the DTW algorithm is used to dynamically align the delay sequences of Paths A, B, and C with the reference templates to calculate their minimum cumulative path distance. The results are: DTW(A, T) = 0.1, DTW(B, T) = 0.2, and DTW(C, T) = 0.4. Path A has the smallest distance according to the minimum DTW distance sorting method and is therefore determined to be the optimal reflection path. Optionally, a threshold can be set or the top N paths can be selected as the optimal path sequence. The final output optimal reflection path sequence is: Path A.

[0028] S2, obtain the topological structure of the optimal reflection path sequence, build a quadratic unconstrained binary optimization model and perform quantum annealing to obtain the discrete phase distribution;

[0029] In this example, the topological structure of the optimal reflection path sequence is obtained, a quadratic unconstrained binary optimization model is constructed and quantum annealing is performed to obtain the discrete phase distribution, specifically:

[0030] Determining activation state variables and phase adjustment variables of the programmable metasurface hardware unit according to the topological structure of the optimal reflection path sequence;

[0031] Taking the activation state variable and phase adjustment variable as input, a quadratic unconstrained binary optimization model is constructed based on the preset activation state constraint and phase discretization constraint, with the maximum signal strength as the optimization goal.

[0032] The model is mapped to the Ising model of the quantum annealing machine, and the discrete phase distribution is obtained by using the quantum annealing algorithm.

[0033] It should be noted that the topology can be represented as a directed graph, where each reflection point corresponds to a programmable metasurface hardware unit. Each unit has two variables: an activation state variable indicating whether the unit is activated, and a phase adjustment variable representing the adjusted discrete phase. A quadratic unconstrained binary optimization model is constructed to achieve signal enhancement in millimeter-wave communications. First, based on a selected optimal reflection path sequence, the state of each programmable metasurface hardware unit—namely, whether it participates in reflection and its specific phase adjustment value—is determined. To accommodate the discrete phase settings available in the hardware, each phase adjustment variable is discretized into a fixed set of angles and encoded using a set of mutually exclusive binary variables. Based on this, the model optimizes the signal strength at the receiver by transforming the coupling relationships and phase differences between all reflection units into a quadratic objective function involving binary variables. This objective function does not contain any constraints and, through proper construction, can be transformed into a standard quadratic unconstrained binary optimization model, making it suitable for efficient solution using discrete optimization methods such as quantum annealing. This model simultaneously considers the interactions between multiple units and their discrete control behaviors, thereby providing an optimal strategy within the feasible solution space for subsequent optimization.

[0034] Furthermore, the specific steps of quantum annealing solution are as follows:

[0035] The initial state is set to quantum superposition state;

[0036] The system gradually evolves in the simulated annealing path, and the Hamiltonian evolves from the initial Hamiltonian to the target Hamiltonian;

[0037] After the final measurement, the lowest energy state (i.e. the bit string of the optimal solution) is obtained;

[0038] The process minimizes the Hamiltonian function and finds the bit configuration that maximizes the objective function.

[0039] S3, calculating the quantization error characteristic value of the discrete phase distribution, and using an interpolation algorithm to perform frequency domain compensation on the quantization error characteristic value to obtain a phase compensation vector;

[0040] In this example, the quantization error eigenvalue of the discrete phase distribution is calculated, and the interpolation algorithm is used to compensate the quantization error eigenvalue in the frequency domain to obtain the phase compensation vector, which is specifically:

[0041] Calculating the difference between the discrete phase distribution and the preset continuous phase distribution, and calculating the quantization error value of each programmable metasurface hardware unit based on the difference;

[0042] Perform Fourier transform on the quantization error value and extract the frequency domain feature component to obtain the quantization error feature vector;

[0043] Perform spectral analysis on the error eigenvectors in the frequency domain and use adaptive fluctuation detection method to identify missing frequency bands;

[0044] A fuzzy clustering algorithm is used to dynamically divide the frequency bands of the quantization error characteristic vector to obtain a plurality of first frequency bands;

[0045] Calculating the ratio of the mean and standard deviation of the quantization error eigenvector of each first frequency band, and dynamically adjusting the node spacing density of the cubic spline interpolation according to the ratio;

[0046] Adaptive interpolation compensation is performed on the missing frequency bands according to the adjusted node interval density, and the interpolated frequency domain compensation results are inversely transformed into the time domain to generate a phase compensation vector.

[0047] It should be noted that in millimeter wave networks, in order to reduce the error caused by the discrete phase distribution and the ideal continuous phase distribution, this phase quantization error needs to be compensated in the frequency domain. Specifically, the difference between the actual discrete phase value and the expected continuous value of each metasurface unit is first calculated, and these errors are transformed into a vector to obtain a frequency domain representation reflecting the frequency characteristics of the error, namely the error characteristic vector. Next, the missing frequency bands in the frequency domain signal are identified through an adaptive fluctuation detection method, and the spectrum is divided using fuzzy clustering to obtain multiple error frequency bands. Subsequently, the ratio of the mean to the standard deviation of the error vector in each frequency band is calculated to reflect the stationary or fluctuating error distribution of the frequency band. Based on this, the node spacing density of the cubic spline interpolation is dynamically adjusted to compensate for the missing frequency bands. Finally, the frequency domain compensation results are restored to the time domain through an inverse Fourier transform to generate a phase compensation vector that can be used for phase control, thereby improving beam accuracy and overall system performance.

[0048] S4, loading the phase compensation vector into the programmable metasurface hardware unit to obtain the first reflection link and construct the phase error covariance matrix;

[0049] In this example, the phase compensation vector is loaded into the programmable metasurface hardware unit to obtain the first reflection link, and the phase error covariance matrix is ​​constructed, specifically:

[0050] Adjusting a phase response parameter of the programmable metasurface hardware unit according to a phase compensation vector to obtain a first reflection link;

[0051] Measuring a deviation between an actual phase response and a preset first phase response through a first reflection link to construct a phase error sample set;

[0052] Based on the phase error sample set, the phase error covariance value between each programmable metasurface hardware unit is calculated to obtain the phase error covariance matrix.

[0053] It should be noted that after obtaining the phase compensation vector, the vector will be loaded into the programmable metasurface hardware unit to adjust the phase response of each unit, thereby forming an actual first reflection link. In order to evaluate the phase control accuracy of this link, the system will compare the phase response of the actual reflection link with the theoretical preset response, calculate the phase deviation of each metasurface unit, and form a sample set of these deviation data. Based on this sample set, the phase error covariance between any two units is calculated by statistical methods to construct a complete phase error covariance matrix. This matrix not only reflects the spatial correlation distribution of the error, but also provides a data basis for the precise correction of the weight parameters in the subsequent optimization model, which helps to further improve the accuracy and robustness of beam control.

[0054] S5, incrementally modifying the weight parameters of the quadratic unconstrained binary optimization model based on the phase error covariance matrix;

[0055] In this example, the weight parameters of the quadratic unconstrained binary optimization model are incrementally modified based on the phase error covariance matrix, specifically:

[0056] The phase error covariance matrix is ​​input as a regularization term into the quadratic unconstrained binary optimization model;

[0057] The gradient descent method with adaptive step size is used to iteratively update the weight parameters of the quadratic unconstrained binary optimization model until the error converges and the corrected weight parameters are obtained.

[0058] It should be noted that after phase compensation is completed and loaded onto the programmable metasurface, the system obtains the actual phase response deviation of each unit through measurements on the first reflection link, constructing a phase error sample set covering all units. After further statistical analysis, a phase error covariance matrix reflecting the correlation of errors between units is calculated and used as external feedback information for the current model.

[0059] In the original quadratic unconstrained binary optimization model, the coupling relationships between individual metasurface units are represented as a weight matrix to describe their joint influence. However, due to the discreteness and response offset of actual hardware, the original weight setting fails to accurately reflect the error propagation effect. Therefore, a new model structure with a regularization term is adopted. The phase error covariance matrix is ​​embedded as a regularization factor in the optimization model, which modifies the original objective function. This allows the model to consider actual error correlations while optimizing the objective, thereby enhancing robustness to noise and interference.

[0060] In order to update the weight parameters efficiently, a gradient descent method based on adaptive step size is used for iterative update. The specific process is as follows:

[0061] Initialize the weight matrix to the parameters obtained after the last quantum annealing optimization;

[0062] The covariance matrix is ​​added to the objective function as a regularization penalty term to form a new loss function with constraints;

[0063] Calculate the gradient direction of the current loss function for each weight;

[0064] Adjust the weight value according to the gradient direction, and introduce dynamic step size control to automatically adjust the amplitude of each iteration according to the rate of decrease of the loss function;

[0065] Continue iterating until the loss function decreases by less than the preset convergence threshold, indicating that the model has fully absorbed the error statistical characteristics and outputs the final corrected weight parameters.

[0066] S6, constructs the objective function based on the modified weight parameters and phase compensation vector, and reconstructs the channel capacity through the gradient projection method to obtain the optimal channel capacity.

[0067] In this example, the objective function is constructed based on the modified weight parameters and phase compensation vector, and the channel capacity is reconstructed using the gradient projection method to obtain the optimal channel capacity, which is specifically:

[0068] Taking the modified weight parameters and phase compensation vector as input, a multi-objective optimization function is constructed;

[0069] The multi-objective optimization function is iteratively optimized by the gradient projection method to calculate the maximum channel capacity that meets the preset power constraints;

[0070] Initialize the first channel capacity value and calculate the gradient direction of the multi-objective optimization function under the current parameters;

[0071] Project the gradient direction to the feasible solution space and dynamically update the first channel capacity value;

[0072] Repeat the iteration until the increment of the first channel capacity value is less than a preset first threshold value, and obtain the optimal channel capacity.

[0073] It should be noted that the reconstruction process is based on two types of input: one is the optimization model weight parameters obtained through incremental correction in the previous stage, which already includes the adaptation of hardware error characteristics; the other is the phase compensation vector generated based on frequency domain compensation, which is used to reflect the actual control capabilities of the current programmable metasurface hardware unit. With the goal of improving the system channel capacity, a multi-objective optimization function is constructed, comprehensively considering the following factors:

[0074] Maximize beam gain; minimize signal interference; ensure smooth phase transitions between adjacent units; and ensure that the transmit power does not exceed a preset constraint. The objective function takes into account both transmission performance and engineering constraints. While the overall optimization problem is non-convex, its structure is gradient-differentiable.

[0075] Furthermore, in order to efficiently solve the above objective function, the gradient projection method is used for iterative optimization. The main steps are as follows:

[0076] Initialize the channel capacity value to the theoretical capacity of the current system under the original configuration;

[0077] Calculate the gradient direction of the multi-objective optimization function under the current phase compensation and weight combination;

[0078] Project the gradient direction into the feasible solution space that satisfies the power and physical adjustable range;

[0079] According to the projection results, update the channel capacity estimation value and record the optimization step size;

[0080] If the increment between the current capacity value and the previous round is less than the set threshold, or the optimization function changes tend to be stable, the iteration is stopped;

[0081] The final output is the phase configuration and model weight corresponding to the maximum channel capacity, which is used for actual beam pointing control. Through the gradient projection optimization method, the system can continuously approach the optimal value of the channel capacity under multiple constraints and effectively avoid falling into the local optimal solution. At the same time, this method fully utilizes the complementary advantages of the previous optimization model and the frequency domain error compensation results, realizing the integrated processing from "error correction" to "capacity enhancement", which greatly improves the transmission efficiency and robustness of the millimeter wave communication link.

[0082] Example 2, Figure 2 The present invention provides a beam steering optimization system for millimeter wave networks, including a path matching module, an optimization solution module, a phase compensation module, a hardware control module, a parameter correction module, and a capacity reconstruction module:

[0083] The path matching module is used to match the multi-hop path delay characteristics between the base station and the user equipment using the DTW algorithm to select the optimal reflection path sequence;

[0084] The optimization solution module is used to obtain the topological structure of the optimal reflection path sequence, build a quadratic unconstrained binary optimization model and perform quantum annealing to obtain the discrete phase distribution;

[0085] Determining activation state variables and phase adjustment variables of the programmable metasurface hardware unit according to the topological structure of the optimal reflection path sequence;

[0086] Taking the activation state variable and phase adjustment variable as input, a quadratic unconstrained binary optimization model is constructed based on the preset activation state constraint and phase discretization constraint, with the maximum signal strength as the optimization goal.

[0087] The quadratic unconstrained binary optimization model is mapped to the Ising model of the quantum annealing machine, and the discrete phase distribution is obtained by using the quantum annealing algorithm.

[0088] The phase compensation module is used to calculate the quantization error characteristic value of the discrete phase distribution and use the interpolation algorithm to perform frequency domain compensation on the quantization error characteristic value to obtain the phase compensation vector;

[0089] Calculating the difference between the discrete phase distribution and the preset continuous phase distribution, and calculating the quantization error value of each programmable metasurface hardware unit based on the difference;

[0090] Perform Fourier transform on the quantization error value and extract the frequency domain feature component to obtain the quantization error feature vector;

[0091] Perform spectral analysis on the error eigenvectors in the frequency domain and use adaptive fluctuation detection method to identify missing frequency bands;

[0092] A fuzzy clustering algorithm is used to dynamically divide the frequency bands of the quantization error characteristic vector to obtain a plurality of first frequency bands;

[0093] Calculating the ratio of the mean and standard deviation of the quantization error eigenvector of each first frequency band, and dynamically adjusting the node spacing density of the cubic spline interpolation according to the ratio;

[0094] Adaptively interpolate and compensate the missing frequency bands according to the adjusted node spacing density, and inversely transform the interpolated frequency domain compensation results into the time domain to generate a phase compensation vector;

[0095] a hardware control module, configured to load the phase compensation vector into the programmable metasurface hardware unit, obtain a first reflection link, and construct a phase error covariance matrix;

[0096] A parameter correction module is used to incrementally correct the weight parameters of the quadratic unconstrained binary optimization model based on the phase error covariance matrix;

[0097] The capacity reconstruction module is used to construct the objective function based on the corrected weight parameters and phase compensation vector, and reconstruct the channel capacity through the gradient projection method to obtain the optimal channel capacity.

[0098] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0099] The above embodiments may be implemented in whole or in part through software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0100] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0101] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0102] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0103] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A beam steering optimization method for millimeter wave networks, characterized in that: The following steps are involved: The DTW algorithm is used to match the multi-hop path delay characteristics between the base station and the user equipment to select the optimal reflection path sequence; The topological structure of the optimal reflection path sequence is obtained, a quadratic unconstrained binary optimization model is constructed and quantum annealing is performed to obtain the discrete phase distribution, specifically: Based on the topological structure of the optimal reflection path sequence, the activation state variables and phase adjustment variables of the programmable metasurface hardware unit are determined. A quadratic unconstrained binary optimization model is constructed based on the activation state and phase discretization constraints, taking maximum signal strength as the optimization objective, using the activation state and phase adjustment variables as input. The quadratic unconstrained binary optimization model is mapped to the Ising model of a quantum annealing machine, and the discrete phase distribution is obtained using a quantum annealing algorithm. Calculate the quantization error eigenvalue of the discrete phase distribution, and use the interpolation algorithm to perform frequency domain compensation on the quantization error eigenvalue to obtain the phase compensation vector, which is specifically: The difference between the discrete phase distribution and the preset continuous phase distribution is calculated, and the quantization error value of each programmable metasurface hardware unit is calculated based on the difference; the quantization error value is Fourier transformed, and the frequency domain characteristic components are extracted to obtain a quantization error characteristic vector; the error characteristic vector in the frequency domain is spectrally analyzed, and an adaptive fluctuation detection method is used to identify missing frequency bands; the frequency bands of the quantization error characteristic vector are dynamically divided using a fuzzy clustering algorithm to obtain a number of first frequency bands; the ratio of the mean and standard deviation of the quantization error characteristic vector of each first frequency band is calculated, and the node spacing density of the cubic spline interpolation is dynamically adjusted according to the ratio; the missing frequency bands are adaptively interpolated and compensated according to the adjusted node spacing density, and the interpolated frequency domain compensation result is inversely transformed to the time domain to generate a phase compensation vector; Loading the phase compensation vector into the programmable metasurface hardware unit to obtain the first reflection link and construct the phase error covariance matrix; Incremental modification of the weight parameters of the quadratic unconstrained binary optimization model based on the phase error covariance matrix; The objective function is constructed based on the corrected weight parameters and phase compensation vector, and the channel capacity is reconstructed by the gradient projection method to obtain the optimal channel capacity.

2. The beam steering optimization method for millimeter wave networks according to claim 1, characterized in that: The DTW algorithm is used to match the multi-hop path delay characteristics between the base station and the user equipment to select the optimal reflection path sequence, specifically: Collect delay sequence data of several candidate reflection paths between the base station and the user equipment, and build a path delay feature library; Based on the preset reference delay template, the DTW algorithm is used to align the delay sequence of each candidate reflection path and calculate the minimum cumulative path distance; Several candidate reflection paths are screened according to the minimum cumulative path distance to obtain the optimal reflection path sequence.

3. The beam steering optimization method for millimeter wave networks according to claim 2, characterized in that: The phase compensation vector is loaded into the programmable metasurface hardware unit to obtain the first reflection link and construct the phase error covariance matrix, specifically: Adjusting a phase response parameter of the programmable metasurface hardware unit according to a phase compensation vector to obtain a first reflection link; Measuring a deviation between an actual phase response and a preset first phase response through a first reflection link to construct a phase error sample set; Based on the phase error sample set, the phase error covariance value between each programmable metasurface hardware unit is calculated to obtain the phase error covariance matrix.

4. The beam steering optimization method for millimeter wave networks according to claim 3, characterized in that: The weight parameters of the quadratic unconstrained binary optimization model are incrementally modified based on the phase error covariance matrix, specifically: The phase error covariance matrix is ​​input as a regularization term into the quadratic unconstrained binary optimization model; The gradient descent method with adaptive step size is used to iteratively update the weight parameters of the quadratic unconstrained binary optimization model until the error converges and the corrected weight parameters are obtained.

5. The beam steering optimization method for millimeter wave networks according to claim 4, characterized in that: The objective function is constructed based on the modified weight parameters and phase compensation vector, and the channel capacity is reconstructed by the gradient projection method to obtain the optimal channel capacity, which is specifically: Taking the modified weight parameters and phase compensation vector as input, a multi-objective optimization function is constructed; The multi-objective optimization function is iteratively optimized by the gradient projection method to calculate the maximum channel capacity that meets the preset power constraints; Initialize the first channel capacity value and calculate the gradient direction of the multi-objective optimization function under the current parameters; Project the gradient direction to the feasible solution space and dynamically update the first channel capacity value; Repeat the iteration until the increment of the first channel capacity value is less than a preset first threshold value, and obtain the optimal channel capacity.

6. A system using the beam steering optimization method for a millimeter wave network according to any one of claims 1 to 5, characterized in that: Including path matching module, optimization solution module, phase compensation module, hardware control module, parameter correction module, and capacity reconstruction module: The path matching module is used to match the multi-hop path delay characteristics between the base station and the user equipment using the DTW algorithm to select the optimal reflection path sequence; The optimization solution module is used to obtain the topological structure of the optimal reflection path sequence, build a quadratic unconstrained binary optimization model and perform quantum annealing to obtain the discrete phase distribution, specifically: Based on the topological structure of the optimal reflection path sequence, the activation state variables and phase adjustment variables of the programmable metasurface hardware unit are determined. A quadratic unconstrained binary optimization model is constructed based on the activation state and phase discretization constraints, taking maximum signal strength as the optimization objective, using the activation state and phase adjustment variables as input. The quadratic unconstrained binary optimization model is mapped to the Ising model of a quantum annealing machine, and the discrete phase distribution is obtained using a quantum annealing algorithm. The phase compensation module is used to calculate the quantization error eigenvalue of the discrete phase distribution and use the interpolation algorithm to perform frequency domain compensation on the quantization error eigenvalue to obtain the phase compensation vector, which is specifically: The difference between the discrete phase distribution and the preset continuous phase distribution is calculated, and the quantization error value of each programmable metasurface hardware unit is calculated based on the difference; the quantization error value is Fourier transformed, and the frequency domain characteristic components are extracted to obtain a quantization error characteristic vector; the error characteristic vector in the frequency domain is spectrally analyzed, and an adaptive fluctuation detection method is used to identify missing frequency bands; the frequency bands of the quantization error characteristic vector are dynamically divided using a fuzzy clustering algorithm to obtain a number of first frequency bands; the ratio of the mean and standard deviation of the quantization error characteristic vector of each first frequency band is calculated, and the node spacing density of the cubic spline interpolation is dynamically adjusted according to the ratio; the missing frequency bands are adaptively interpolated and compensated according to the adjusted node spacing density, and the interpolated frequency domain compensation result is inversely transformed to the time domain to generate a phase compensation vector; a hardware control module, configured to load the phase compensation vector into the programmable metasurface hardware unit, obtain a first reflection link, and construct a phase error covariance matrix; A parameter correction module is used to incrementally correct the weight parameters of the quadratic unconstrained binary optimization model based on the phase error covariance matrix; The capacity reconstruction module is used to construct the objective function based on the corrected weight parameters and phase compensation vector, and reconstruct the channel capacity through the gradient projection method to obtain the optimal channel capacity.

Images