Intelligent broadband transmission method based on Mimo communication

Abstract

The invention discloses an intelligent broadband transmission method based on Mimo communication, and relates to the technical field of broadband Mimo wireless communication, and the method comprises the steps: determining the number and arrangement mode of transmitting and receiving antennas, generating a dynamic channel matrix through employing a geometric statistical modeling method in combination with a ray tracing method, and introducing a deep neural network to predict a change trend; optimizing antenna resources, transmitting power and spectrum allocation by adopting a reinforcement learning algorithm, and constructing a multi-constraint optimization objective function; an improved QR decomposition algorithm and a post-sorting algorithm are adopted, and an interference alignment technology and a minimum mean square error algorithm are combined to suppress interference and noise; signal preprocessing, compression and path optimization are carried out through edge calculation, and transmission delay is reduced; based on channel capacity, bit error rate, throughput and delay key performance indexes, in combination with deep learning and reinforcement learning models, resource allocation and channel modeling are optimized and adjusted. The method has the characteristics of high efficiency and self-adaption, and can be widely applied to the field of wireless communication.

Description

Technical Field

[0001] The present invention relates to the technical field of broadband Mimo wireless communication, and particularly to an intelligent broadband transmission method based on Mimo communication. Background Art

[0002] With the rapid development of 5G and future 6G communication technologies, broadband wireless communication systems are facing growing demands for high data rate, low latency, and high reliability. Multiple-input multiple-output (Mimo) communication technology, as a core means to improve spectrum utilization and communication efficiency, has been widely applied in the field of wireless communication. However, the performance of Mimo communication in complex channel environments is limited by the accuracy of channel modeling, the dynamic scheduling ability of multi-antenna resources, and the robustness of signal detection and anti-interference technologies.

[0003] Currently, mainstream Mimo communication methods mostly rely on traditional channel modeling techniques, such as statistical modeling or deterministic modeling. These methods lack high-precision prediction ability for channel dynamic changes in complex scenarios. At the same time, existing resource allocation strategies are mostly based on static optimization algorithms and are difficult to adapt to the real-time requirements of channel changes over time and environment. In addition, traditional signal detection algorithms are limited in performance under low signal-to-noise ratio and multipath propagation conditions and cannot effectively suppress interference and ensure high-precision signal decoding. With the diversification of wireless communication scenarios, such as intelligent transportation, industrial Internet of Things, and edge computing application scenarios, the limitations of the existing technology are becoming increasingly obvious.

[0004] To address the above problems, the present invention proposes an intelligent broadband transmission method based on Mimo communication. By introducing the combination of geometric statistical channel modeling method and deep neural network; using reinforcement learning algorithm to optimize antenna resources, transmit power, and spectrum allocation; and combining improved QR decomposition and post-sorting algorithm with interference alignment and minimum mean square error algorithm, the signal detection accuracy and anti-interference ability are greatly improved. Summary of the Invention

[0005] The present invention addresses the above problems and provides an intelligent broadband transmission method based on Mimo communication to solve the problems of insufficient channel modeling accuracy, low intelligence level of resource allocation, weak signal detection and anti-interference ability, and heavy terminal computing burden in the existing technology in complex dynamic environments.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: An intelligent broadband transmission method based on Mimo communication, comprising the following steps:

[0007] Step S1, by configuring a Mimo communication system, determine the number of transmit antennas and receive antennas, the signal frequency range, and the array type, and actively collect channel characteristic data; based on a geometric statistical modeling method, combine the ray tracing method to generate a dynamic channel matrix, and introduce a deep neural network to predict the dynamic changes of the channel state;

[0008] Among them, in step S1, the following sub-steps are also included:

[0009] S1-1, determine the number of transmit antennas M t and receive antennas M r Set the transmit frequency range, signal bandwidth, and antenna array type, actively detect the channel characteristics in the environment using a preset training signal sequence, capture the transmitted signal through the receive antenna array, and obtain the initial channel state information, including channel gain, path loss, multipath delay, and interference intensity parameters. The core objective of the acquisition is to generate a channel matrix H, and its definition is specifically as shown in Equation (1):

[0010] H = [h ij M r ×M t Equation (1)

[0011] Among them, H is the channel matrix, [h ij represents the channel coefficient between the i-th receive antenna and the j-th transmit antenna, M t is the number of transmit antennas, and M r is the number of receive antennas;

[0012] S1-2, based on the collected channel data, select a geometric statistical modeling method. During the modeling process, analyze channel parameters, including path loss, multipath delay, and spatial correlation. Describe the distribution characteristics of scatterers through geometric modeling. Assume that the scatterer distribution satisfies a certain probability density function, randomly generate the positions of scatterers, and combine the ray tracing method to calculate the propagation path of the signal and its corresponding gain. The mathematical expression of the channel model is specifically as shown in Equation (2):

[0013]

[0014] Among them, [h ij represents the channel coefficient between the i-th receive antenna and the j-th transmit antenna, L is the number of multipaths, α l is the fading coefficient of the l-th path, τ l is the time delay of this path, and f is the signal frequency; combined with this model, generate a dynamic channel matrix H(t,f) that includes time, frequency, and spatial characteristics;

[0015] Introduce a machine learning optimization module, and use historical channel data and real-time updated channel status to train a deep neural network (DNN) to predict the dynamic change trend of the channel matrix. In the specific implementation process, input the historical collected data and the current channel matrix H(t,f) into the DNN model, and this model outputs the predicted channel matrix. Its objective function is defined as minimizing the error between the predicted matrix and the real matrix, specifically as shown in Equation (3):

[0016]

[0017] Where, ζ is the error objective function, N is the number of samples, ||·|| 2 is the matrix norm, H n (t+Δt,f) is the actual channel matrix, is the predicted channel matrix;

[0018] S1-3, after completing the channel modeling, verify the generated channel model H. The channel verification process includes performance testing, model comparison, and environmental adaptability analysis. Design standard performance indicators to evaluate the accuracy and stability of the channel model. These indicators include the mean square error and the correlation coefficient;

[0019] The mean square error (MSE) is used to evaluate the difference between the predicted matrix of the model and the actual measured matrix H, specifically as shown in Equation (4):

[0020]

[0021] Where, MSE is the mean square error, M t and M r are the number of transmit antennas and receive antennas respectively, H ij is the element of the actual channel matrix, is the element of the predicted channel matrix;

[0022] The correlation coefficient (CR) is used to measure the degree of retention of the channel correlation by the model, specifically as shown in Equation (5):

[0023]

[0024] Where, CR is the correlation coefficient, tr(·) represents the trace of the matrix, ||·|| F is the Frobenius norm, H H is the conjugate transpose of the channel matrix;

[0025] Compare the generated channel model with the standard channel model to evaluate the performance improvement of the modeling method. By running different models in the same simulation environment, compare their description accuracies of channel gain, signal-to-noise ratio, and multipath characteristics, and verify the advantages of the geometric statistical model;

[0026] In the actual communication environment, the adaptability of the model is further verified through measured data. The modeling results are applied to scenarios in a dynamic environment to verify whether the model can effectively cope with changes in channel characteristics. The tests include the dynamic update ability of the channel matrix and the robustness performance in low signal-to-noise ratio and high interference environments.

[0027] Step S2: After the channel modeling is completed, the channel state information is obtained in real time, key parameters such as channel capacity, interference intensity, and multipath characteristics are calculated, and the correlation matrix of antenna configuration is generated; based on the reinforcement learning algorithm, the antenna resources, transmit power, and spectrum allocation are dynamically optimized, and a comprehensive objective function that meets various constraint conditions is constructed.

[0028] Among them, in step S2, the following sub-steps are also included:

[0029] S2-1: Obtain the channel state information in real time, collect and calculate the key parameters in the channel in real time, and provide data support for dynamic resource allocation. The specific steps are as follows:

[0030] Utilize the dynamic channel matrix H(t,f) provided by the channel modeling module to capture the real-time channel change information through the receiving antenna array; combine the multi-slot data to calculate the time derivative characteristic ΔH = H(t + Δt,f) - H(t,f) of the channel matrix, which is used to describe the change trend of the channel over time.

[0031] Based on the collected channel matrix, calculate the instantaneous channel capacity C, specifically as shown in Equation (6):

[0032]

[0033] where C is the instantaneous channel capacity, I is the identity matrix, P is the transmit power, σ 2 is the noise power density, H is the channel matrix, and H H is the conjugate transpose of the channel matrix; through the signal processing unit at the receiving end, extract the signal-to-interference ratio SINR of the signal, and combine the interference analysis to identify the main interference sources in the current channel and their impact on the transmission performance.

[0034] Separate the multipath components of the signal through the time-frequency analysis method, and obtain the delay τ l of each path, the gain α l and the phase shift, and form the multipath information matrix Θ = {(α l , τ l )} l = 1 L for resource allocation optimization; calculate the spatial correlation matrix R between antennas, evaluate the correlation between antennas, specifically as shown in Equation (7):

[0035] R = HH Formula (7) of H

[0036] where R is the spatial correlation matrix, H is the channel matrix, and H H is the conjugate transpose of the channel matrix;

[0037] Evaluate the spatial multiplexing ability of the antenna configuration through eigenvalue analysis to help select the optimal antenna subset; combine historical channel data to perform short-term prediction on future channel states, and use the trained deep learning model f DNN (H(t,f)) to output the channel matrix at the next moment Improve the foresight of resource allocation;

[0038] S2-2, after obtaining the channel state information, use the optimization algorithm to design a dynamic resource allocation strategy, construct an optimization objective function for resource allocation according to the performance requirements of communication, and the goal of optimization is to maximize the channel capacity C and the signal-to-noise ratio SINR, and minimize the power consumption P and the delay D. The comprehensive objective function is defined as follows in Formula (8):

[0039]

[0040] where is the objective function, C is the channel capacity, SINR is the signal-to-noise ratio, P is the power consumption, D is the delay, ω 1 , ω 2 , ω 3 , ω 4 is the adjustable weight used to balance the importance of different performance indicators;

[0041] During the optimization process, the following constraint conditions need to be satisfied: power constraint, spectrum constraint, and antenna constraint; the power constraint means that the total transmission power P total shall not exceed the power upper limit, as shown in Formula (9):

[0042]

[0043] where P i is the transmission power of the i-th antenna, and P max is the maximum power upper limit;

[0044] The spectrum constraint means that the allocated frequency band f must be within the permitted frequency band range, as shown in Formula (10):

[0045] f min ≤ f i ≤ f max Formula (10)

[0046] where f i is the spectrum allocation frequency band of the i-th antenna, f min is the minimum frequency limit, fmax is the maximum frequency limit;

[0047] The antenna constraint means that the number of selected transmit antennas k should be less than or equal to the total number of antennas M t ;

[0048] Use the reinforcement learning (RL) algorithm to dynamically optimize the resource allocation strategy. In communication, define the key elements of reinforcement learning: State, Action, Reward; The reward value calculated based on the objective function is specifically as shown in Equation (11):

[0049]

[0050] where R is the reward value, is the objective function, Penalty represents the penalty for violating the constraint conditions, and λ is the penalty factor;

[0051] Use the deep Q-learning algorithm to optimize the strategy, estimate the expected reward value of the current state s and action a through the neural network Q(s,a;θ), and improve the learning efficiency through experience replay;

[0052] S2-3, after the resource allocation strategy optimization is completed, enter the actual execution stage of resource allocation, and apply the optimization result to Mimo communication. The specific execution steps are as follows:

[0053] According to the optimal transmit power allocation result output by the optimization algorithm Dynamically adjust the power of each transmit antenna to meet the channel requirements and power constraints. The power control module monitors the transmit power of each antenna in real time to ensure that the allocated power does not exceed the limit P max , for the antennas with low signal-to-noise ratio, according to the weak signal paths in the channel matrix H, preferentially increase the power to improve their transmission performance;

[0054] Execute the spectrum allocation result, and allocate the frequency band f i and frequency bandwidth Δf obtained by the optimization calculation to the corresponding antenna subsets: f = {f 1 , f 2 ,..., f k}, Δf = {f 1,bw , f 2,bw ,..., f k,bw}, and through the dynamic spectrum allocation algorithm, efficiently allocate the broadband resources to the channels with higher priorities to avoid spectrum resource conflicts;

[0055] According to the optimized antenna subsets Activate the selected transmit antennas and turn off the unnecessary antennas, dynamically adjust the antenna configuration through the switching matrix, and concentrate the resources on the signal paths with higher signal-to-noise ratio to maximize the channel capacity;

[0056] During the execution of the resource allocation strategy, the change of the signal interference strength SINR is monitored in real time. The antenna transmission direction is adjusted by combining the multi-antenna interference alignment technology to minimize the interference. The interference alignment matrix G is used to optimize the transmission matrix G to minimize the influence of the interference signal on the target receiving antenna, as shown in Equation (12):

[0057] G = argmin ∥H·G∥, subject to: ∥G∥ ≤ 1 Equation (12)

[0058] where G is the interference alignment matrix, H is the channel matrix, and ∥G∥ is the norm of matrix G;

[0059] During the execution process, the channel state feedback is collected in real time, including the signal-to-noise ratio and the channel capacity index, and the actual effect of the resource allocation is monitored. If it is found that the allocation result does not match the expected performance index, for example, the bit error rate is higher than the threshold, the optimization strategy is recalculated and adjusted through fast iteration.

[0060] Step S3: Through the improved QR decomposition algorithm and the post-sorting algorithm, signal detection is performed. The interference signal is aligned to the interference space through the interference alignment technology, and the minimum mean square error algorithm is used to suppress the noise; based on the real-time monitoring of the bit error rate, the detection algorithm is dynamically adjusted;

[0061] In step S3, the following sub-steps are further included:

[0062] S3-1: Signal detection is one of the core tasks in Mimo communication. Especially in an environment with multipath and large interference, the effective decoding of the signal is particularly crucial. Based on the optimized resource configuration, the signal detection process includes the following steps:

[0063] In Mimo, the received signal y is composed of the superposition of multiple signal sources and noise, as shown in Equation (13):

[0064] y = Hx + n Equation (13)

[0065] where H is the channel matrix, x is the transmitted signal, and n is the noise vector; according to the received signal y and the channel matrix H, the original signal x is recovered through different detection algorithms;

[0066] Before signal detection, the received signal needs to be preprocessed, including denoising and signal amplification. The denoising algorithm suppresses the interference signal by maximizing the signal-to-noise ratio (SNR) to improve the quality of the received signal;

[0067] S3-2, an improved signal detection algorithm is adopted, including improved QR decomposition and post sorting algorithm (PSA), to enhance the signal detection performance in high interference and low signal-to-noise ratio situations; the optimized QR decomposition process includes the following steps:

[0068] Calculate the QR decomposition of the channel matrix to obtain the orthogonal matrix Q and the upper triangular matrix R. Rearrange the columns of the channel matrix according to the signal-to-noise ratio to ensure that the signal with the best signal-to-noise ratio is detected first. Simplify the detection process through QR decomposition, reduce the influence of interference signals on signal decoding, and improve the overall performance;

[0069] The PSA algorithm performs eigenvalue sorting on the upper triangular matrix obtained by QR decomposition and selects the optimal path for signal demodulation according to the eigenvalues. In actual operation, the PSA algorithm gradually eliminates the interference of low signal-to-noise ratio paths and adjusts the estimated value of the signal layer by layer, so that the error does not accumulate during the propagation process;

[0070] The key steps of the PSA algorithm include:

[0071] During the signal detection process, first perform QR decomposition on the channel matrix; adjust the signal reception order according to the eigenvalue size of each path, thereby reducing the influence of low signal-to-noise ratio paths; improve the signal decoding accuracy by sorting and adjusting the paths;

[0072] Combine the improved QR decomposition and PSA algorithms and dynamically adjust the decoding strategy to cope with changes in different channel environments; this combined algorithm can still maintain a high signal detection accuracy in the case of poor signal quality and is particularly suitable for complex multipath and dynamic channel environments;

[0073] S3-3, anti-interference and bit error rate control. Anti-interference and bit error rate control are key steps in a high interference environment. The goal of this stage is to minimize the bit error rate (BER) and ensure reliable transmission of signals even after multiple interferences;

[0074] In multi-antenna Mimo, through interference alignment technology, align the interference of multiple signal sources to the interference space, so that the influence of these interference signals on the target receiving antenna is minimized. Separate the interference signals from the useful signals by adjusting the signal direction of the transmitting antenna, reducing multi-user interference and co-channel interference;

[0075] The MMSE algorithm calculates the optimal filter matrix by introducing the noise covariance matrix and comprehensively considering the influence of signals and noise, as shown in Equation (14):

[0076]

[0077] where, G MMSE is the optimal filter matrix, H is the channel matrix, HH is the conjugate transpose of the channel matrix, σ 2 is the noise power, P is the signal power, and I is the identity matrix; The MMSE algorithm can effectively suppress noise in the case of low signal-to-noise ratio and ensure correct decoding of signals in the presence of noise interference;

[0078] During the detection process, the bit error rate (BER) is monitored in real time, and the demodulation algorithm is adjusted according to the detection results; If the bit error rate exceeds the set threshold, it will trigger the re-optimization of the signal detection algorithm, including reselecting the antenna subset, adjusting the transmit power, or reallocating the spectrum; In addition, based on the bit error rate feedback, the resource configuration can be corrected in real time, the signal detection process can be optimized, and the transmission stability can be ensured.

[0079] Step S4, use the edge node to process part of the computing tasks, perform channel modeling optimization, signal preprocessing, and data compression, reduce the computing burden of the terminal and reduce the latency; Optimize the transmission path through multipath transmission, load balancing, and path switching, and dynamically adjust the transmission strategy in combination with the feedback mechanism;

[0080] Among them, in step S4, the following sub-steps are also included:

[0081] S4-1, use edge computing to improve performance, reduce the computing burden of the terminal and reduce the latency by transferring part of the computing tasks from the terminal device to the edge node; The data processing module of the edge node mainly includes the following aspects:

[0082] The edge node receives the channel state information (CSI) from multiple terminal devices, and through a fast channel estimation algorithm, generates the channel matrix H to accurately model the channel; At the same time, the edge node uses a method that combines historical data and real-time data to optimize channel prediction and reduce errors; The edge node can also further improve the accuracy of channel estimation through a machine learning model;

[0083] In the signal preprocessing stage, the edge node is responsible for filtering, denoising, and signal enhancement of the raw data from the terminal device; Through advanced digital signal processing technology, the edge node can effectively improve the quality of the received signal; This process can reduce the computing burden of the terminal device and provide a cleaner signal input for signal detection and decoding;

[0084] The edge node compresses the data from the terminal to reduce the amount of data transmission; The compression algorithm can be based on the compressed sensing (CS) theory or other efficient compression technologies to ensure that the bandwidth required for data transmission is greatly reduced without losing key data; In addition, the edge node is also responsible for optimizing the data transmission path, selecting the optimal transmission channel, reducing latency, improving bandwidth utilization, and ensuring real-time response capabilities in a dynamic environment;

[0085] S4-2. Through intelligent path selection and optimization algorithms, edge nodes can provide the best communication paths for different terminal devices, reducing relay, interference, and transmission latency. The specific steps are as follows:

[0086] Based on real-time channel information and interference analysis, edge nodes can dynamically select the optimal transmission path; switch paths according to the real-time channel state and user requirements in a complex network environment; edge nodes can judge whether to switch to a better transmission path by real-time monitoring of channel quality;

[0087] In the network, edge nodes can optimize data transmission through multipath transmission, using multiple independent paths to transmit data, improving the transmission speed and increasing the network's fault tolerance; the load balancing algorithm reasonably distributes different communication loads to multiple paths to prevent a single path from being overloaded, resulting in transmission latency or packet loss;

[0088] During the path optimization process, edge nodes will evaluate the performance of each transmission path in real-time, mainly by monitoring indicators such as latency, packet loss rate, bandwidth, and signal-to-noise ratio; after the transmission is completed, edge nodes will feedback the transmission performance and further optimize the path selection strategy based on these feedbacks to adapt to the changing network environment;

[0089] S4-3. During the transmission process, it is crucial to ensure the accuracy, stability, and efficiency of the transmission; the transmission result verification module can monitor and feedback the performance of data transmission in real-time, so as to adjust the optimization strategy in a timely manner when abnormalities occur. The key steps are as follows:

[0090] During the transmission process, edge nodes monitor the key performance indicators (KPIs) of the transmission in real-time, such as throughput, transmission rate, bit error rate (BER), latency, and packet loss rate; these monitoring data can help edge nodes judge whether the current transmission meets the requirements and evaluate the effectiveness of the current transmission path and resource allocation;

[0091] To ensure the reliability of data transmission, edge nodes need to control the bit error rate; once a high bit error rate is detected, edge nodes will initiate an error correction mechanism and use techniques such as retransmission, modulation method adjustment, and automatic gain control (AGC) for processing; at the same time, combined with signal processing modules, such as using advanced coding methods like LDPC codes or Turbo codes, to improve the anti-interference ability under poor channel conditions;

[0092] Transmission result verification is not only a detection of data transmission quality but also provides a feedback mechanism; edge nodes optimize the current resource allocation and path selection based on performance monitoring and bit error feedback;

[0093] Step S5. After completing the transmission task, evaluate the system performance through key metrics such as channel capacity, bit error rate, throughput, and latency to verify the effectiveness of the resource allocation and signal detection strategies. Based on the performance evaluation results, optimize the channel modeling and resource allocation strategies using deep learning and reinforcement learning models.

[0094] Among them, in step S5, the following sub-steps are also included:

[0095] S5-1. After completing the transmission task, it is necessary to evaluate the overall performance to verify the effectiveness of the resource allocation and signal transmission strategies. The performance evaluation mainly starts from the following aspects: Calculate the actual channel capacity to verify the improvement effect of the resource allocation strategy on the channel utilization rate, specifically as shown in Equation (15):

[0096]

[0097] Among them, C is the channel capacity, I is the identity matrix, P is the transmit power, σ 2 is the noise power density, H is the channel matrix, and H H is the conjugate transpose of the channel matrix; in different scenarios, including high interference environment, dynamic user distribution, and multipath channel conditions, test the key performance indicators to evaluate the stability and robustness;

[0098] S5-2. According to the performance evaluation results, combine the feedback mechanism to perform intelligent optimization and continuously improve the communication performance. The intelligent optimization is achieved through the following steps: Collect the feedback data obtained from the performance evaluation, including bit error rate, signal-to-noise ratio, and throughput metrics, and input these data into the optimization module to extract abnormal patterns, such as high bit error rate or low signal-to-noise ratio regions;

[0099] Use deep learning (DNN) or reinforcement learning (RL) models to retrain according to the feedback data, specifically as follows:

[0100] Channel modeling optimization, adjust the weights of the model or introduce more historical data to improve the accuracy of channel state prediction;

[0101] Resource allocation optimization, re-optimize the transmit power, antenna selection, and spectrum allocation strategies based on reinforcement learning;

[0102] Introduce a dynamic learning mechanism to enable the model to quickly adapt to the new channel environment;

[0103] Dynamically adjust the key parameters according to the optimization results, including increasing the transmit power in the interference environment, reallocating the spectrum resources to avoid interference, and adjusting the configuration of the antenna array to optimize the spatial multiplexing performance.

[0104] Compared with the prior art, the beneficial effects of the present invention are:

[0105] The present invention realizes high-precision prediction and dynamic modeling of the channel state through a geometric statistical modeling method combined with a deep neural network, can adapt to the changes in complex channel environments, improves the description accuracy of the channel matrix and the real-time performance of modeling, and overcomes the limitations of traditional channel modeling methods in multipath and dynamic interference environments.

[0106] The present invention designs a dynamic resource allocation strategy using a reinforcement learning algorithm, optimizes antenna resources, transmit power, and spectrum allocation according to real-time channel state information, and achieves a dynamic balance among channel capacity, signal-to-noise ratio, and power consumption by constructing a multi-objective optimization function, significantly improving the resource utilization rate and solving the low-efficiency problem caused by the static nature of resource allocation in the prior art.

[0107] The present invention combines an improved QR decomposition algorithm and a post-sorting algorithm, improving the accuracy and efficiency of signal detection, especially performing outstandingly in low signal-to-noise ratio and multi-interference scenarios. Through interference alignment technology and the least mean square error algorithm, it can effectively suppress multi-user interference and the influence of noise, greatly reducing the bit error rate and ensuring the reliable transmission of signals.

[0108] The present invention introduces edge computing technology, distributes signal preprocessing, data compression, and path optimization tasks to edge nodes, significantly reducing the computational burden of terminal devices and the system transmission delay. By optimizing the transmission strategy through multi-path transmission and load balancing, it improves the transmission speed and fault tolerance of the communication system, meeting the requirements of dynamic and complex scenarios. Brief Description of the Drawings

[0109] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention, and thus should not be regarded as limiting the scope. For those of ordinary skill in the art, other relevant drawings can also be obtained based on these drawings without creative efforts.

[0110] Figure 1 It is the method flowchart of the present invention. Detailed Embodiments

[0111] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed present invention, but is merely for the selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0112] Please refer to Figure 1 which is a schematic diagram of an intelligent broadband transmission method based on Mimo communication provided by an embodiment of the present invention, and includes the following steps:

[0113] Step S1: By configuring the Mimo communication system, determine the number of transmitting antennas and receiving antennas, the signal frequency range, and the array type, and actively collect channel characteristic data; based on a geometric statistical modeling method, combine the ray tracing method to generate a dynamic channel matrix, and introduce a deep neural network to predict the dynamic changes of the channel state.

[0114] Among them, in step S1, the following sub-steps are further included:

[0115] S1-1: Configure Mimo communication, determine the number of transmitting antennas M t and receiving antennas M r according to the target scenario, and set the transmission frequency range, signal bandwidth, and antenna array type, such as linear, rectangular, or three-dimensional array; in the channel acquisition stage, use a preset training signal sequence to actively detect the channel characteristics in the environment, capture the transmitted signal through the receiving antenna array, and obtain the initial channel state information, including channel gain, path loss, multipath delay, and interference intensity parameters. The core objective of the acquisition is to generate the channel matrix H, and its definition is specifically as shown in Equation (1):

[0116] H = [h ij M r ×M t Equation (1)

[0117] Among them, H is the channel matrix, [h ij represents the channel coefficient between the i-th receiving antenna and the j-th transmitting antenna, M t is the number of transmitting antennas, and M r is the number of receiving antennas;

[0118] During the acquisition process, synchronous signal sampling technology is adopted to incorporate time-domain and frequency-domain data into the calculation simultaneously, extract the channel impulse response and its dynamic changes, and estimate the noise power density N introduced by channel acquisition. At the same time, the distribution of scatterers and the multipath propagation paths are monitored in real time. Distributed data acquisition is used, and preliminary processing is performed at each receiving point through edge computing devices to reduce data redundancy and generate a high-confidence basic channel dataset. 0

[0119] S1-2. Based on the acquired channel data, a channel model H is constructed using a modeling method suitable for the target communication scenario. Considering the diversity and variability of the channel in complex scenarios, a geometric-based statistical modeling method is selected. During the modeling process, first, channel parameters are analyzed, including path loss, multipath delay, and spatial correlation. The distribution characteristics of scatterers are described through geometric modeling. Assuming that the scatterer distribution satisfies a certain probability density function, the positions of scatterers are randomly generated, and the propagation paths of signals and their corresponding gains are calculated by combining the ray-tracing method. The mathematical expression of the channel model is specifically as shown in Equation (2):

[0120]

[0121] Among them, [h ij represents the channel coefficient between the i-th receiving antenna and the j-th transmitting antenna, L is the number of multipaths, α l is the fading coefficient of the l-th path, τ l is the time delay of this path, and f is the signal frequency. Combining this model, a dynamic channel matrix H(t, f) containing time, frequency, and spatial characteristics is generated.

[0122] A machine learning optimization module is introduced. Using historical channel data and real-time updated channel states, a deep neural network (DNN) is trained to predict the dynamic change trend of the channel matrix. In the specific implementation process, the historical acquisition data and the current channel matrix H(t, f) are input into the DNN model, and the model outputs the predicted channel matrix Its objective function is defined as minimizing the error between the predicted matrix and the real matrix, specifically as shown in Equation (3):

[0123]

[0124] Among them, ζ is the error objective function, N is the number of samples, ||·|| 2 is the matrix norm, H n (t + Δt, f) is the actual channel matrix, is the predicted channel matrix. By iteratively optimizing this loss function, the dynamic prediction ability of channel modeling is continuously improved.

[0125] After model optimization, the channel matrix H is transformed into a feature representation form that is convenient for subsequent steps, including the channel gain matrix G = |H| 2 and the correlation matrix R = HH H H, where H H is the conjugate transpose, and the path information matrix Θ; these features can be directly used for dynamic resource allocation and signal detection;

[0126] S1-3, after completing the channel modeling, it is necessary to verify the generated channel model H. The channel verification process includes performance testing, model comparison, and environmental adaptability analysis. Design standard performance indicators to evaluate the accuracy and stability of the channel model. These indicators include the mean square error and the correlation coefficient;

[0127] The mean square error (MSE) is used to evaluate the difference between the model prediction matrix and the actual measurement matrix H, specifically as shown in Equation (4):

[0128]

[0129] where MSE is the mean square error, M t and M r are the numbers of transmit antennas and receive antennas respectively, H ij is the element of the actual channel matrix, is the element of the predicted channel matrix;

[0130] The correlation coefficient (CR) is used to measure the degree to which the model retains the channel correlation, specifically as shown in Equation (5):

[0131]

[0132] where CR is the correlation coefficient, tr(·) represents the trace of the matrix, ||·|| F is the Frobenius norm, and H H is the conjugate transpose of the channel matrix;

[0133] Compare the generated channel model with the standard channel model to evaluate the performance improvement of the modeling method. By running different models in the same simulation environment, compare their description accuracies of channel gain, signal-to-noise ratio, and multipath characteristics to verify the advantages of the geometric statistical model;

[0134] In the actual communication environment, further test the adaptability of the model through measured data. Apply the modeling results to scenarios in a dynamic environment to verify whether the model can effectively cope with changes in channel characteristics. The tests include the dynamic update ability of the channel matrix and the robustness performance in low signal-to-noise ratio and high interference environments;

[0135] Adjust and optimize the channel model according to the verification results. If deviations are found in the verification, including increased errors at low signal-to-noise ratios or inaccurate descriptions in multipath environments, the model performance can be further improved by improving the quality of the input data, optimizing the machine learning model structure, or enhancing the scatterer distribution modeling method.

[0136] Step S2: After the channel modeling is completed, obtain the channel state information in real time, calculate key parameters such as channel capacity, interference intensity, and multipath characteristics, and generate the correlation matrix of antenna configuration; based on the reinforcement learning algorithm, dynamically optimize the antenna resources, transmit power, and spectrum allocation, and construct a comprehensive objective function that satisfies multiple constraint conditions.

[0137] Among them, in step S2, the following sub-steps are also included:

[0138] S2-1: After the channel modeling is completed and its accuracy is verified, enter the dynamic resource allocation stage. First, it is necessary to obtain the channel state information (CSI) in real time to support subsequent resource optimization decisions. The goal of state information acquisition is to collect and calculate the key parameters in the channel in real time to provide data support for dynamic resource allocation. The specific steps are as follows:

[0139] Utilize the dynamic channel matrix H(t,f) provided by the channel modeling module to capture real-time channel change information through the receiving antenna array; combine multi-slot data to calculate the time derivative characteristic ΔH = H(t + Δt,f) - H(t,f) of the channel matrix, which is used to describe the change trend of the channel over time;

[0140] Based on the collected channel matrix, calculate the instantaneous channel capacity C, specifically as shown in Equation (6):

[0141]

[0142] Among them, C is the instantaneous channel capacity, I is the identity matrix, P is the transmit power, σ 2 is the noise power density, H is the channel matrix, and H H is the conjugate transpose of the channel matrix;

[0143] Through the receiving end signal processing unit, extract the signal-to-interference ratio SINR of the signal, and combine interference analysis to identify the main interference sources in the current channel and their impact on the transmission performance;

[0144] Separate the multipath components of the signal through time-frequency analysis methods, and obtain the delay τ l , gain α l and phase shift of each path, and form the multipath information matrix Θ = {(α l , τ l )} l = 1 L for resource allocation optimization;

[0145] Calculate the spatial correlation matrix R between antennas and evaluate the correlation between antennas, specifically as shown in Equation (7):

[0146] R = H H H Equation (7)

[0147] where R is the spatial correlation matrix, H is the channel matrix, and H H is the conjugate transpose of the channel matrix;

[0148] Evaluate the spatial multiplexing ability of the antenna configuration through eigenvalue analysis to help select the optimal antenna subset;

[0149] Combined with historical channel data, perform short-term prediction on future channel states, and use the trained deep learning model f DNN (H(t,f)) to output the channel matrix at the next moment Improve the forward-looking nature of resource allocation;

[0150] S2-2, after obtaining the channel state information, use the optimization algorithm to design a dynamic resource allocation strategy. The core goal is to maximize the channel capacity and minimize the interference by adjusting the antenna resources, transmit power, and spectrum allocation. The specific process is as follows:

[0151] According to the performance requirements of communication, construct an optimization objective function for resource allocation. The optimization goal is to maximize the channel capacity C and the signal-to-noise ratio SINR, and minimize the power consumption P and the delay D. The comprehensive objective function is defined as shown in Equation (8):

[0152]

[0153] where is the objective function, C is the channel capacity, SINR is the signal-to-noise ratio, P is the power consumption, D is the delay, ω 1 , ω 2 , ω 3 , ω 4 is the adjustable weight used to balance the importance of different performance metrics;

[0154] During the optimization process, the following constraint conditions need to be satisfied: power constraint, spectrum constraint, and antenna constraint; the power constraint means that the total transmit power P total shall not exceed the power upper limit, specifically as shown in Equation (9):

[0155]

[0156] where P i is the transmit power of the i-th antenna, and P max is the maximum power upper limit;

[0157] The spectrum constraint means that the allocated frequency band f must be within the licensed frequency band range, specifically as shown in Equation (10):

[0158] f min ≤f i ≤f max Equation (10)

[0159] where f i is the spectrum allocation frequency band of the i-th antenna, f min is the minimum frequency limit, and f max is the maximum frequency limit;

[0160] The antenna constraint means that the number of selected transmit antennas k should be less than or equal to the total number of antennas M t ;

[0161] Using the reinforcement learning (RL) algorithm to dynamically optimize the resource allocation strategy, in communication, the key elements of reinforcement learning are defined:

[0162] State: The current channel state information, including the channel matrix H, the interference strength SINR, and the power distribution P;

[0163] Action: The resource adjustment strategy, including the power reallocation P i 、antenna selection k and spectrum division f i ;

[0164]

[0165] where R is the reward value, is the objective function, Penalty represents the penalty for violating the constraint conditions, and λ is the penalty factor;

[0166] where Penalty represents the penalty for violating the constraint conditions, and λ is the penalty factor;

[0167] Using the deep Q-learning algorithm to optimize the strategy, estimating the expected reward value of the current state s and action a through the neural network Q(s,a;θ), and improving the learning efficiency through experience replay;

[0168] During the optimization process, the reinforcement learning algorithm continuously iterates and updates the strategy until the following conditions are met:

[0169] The value of the optimization objective function reaches the preset threshold; the reward value R converges to a stable state;

[0170] Through iterative calculation, the optimal resource allocation scheme is output, including: the transmit power P of each antenna i ; the selected spectrum interval [f i , f j; the optimal transmit antenna subset κ;

[0171] By dynamically optimizing the resource allocation strategy, it is possible to achieve efficient utilization of resources in a complex and changing channel environment, and at the same time provide an optimal configuration for signal detection and transmission optimization;

[0172] S2-3, after the resource allocation strategy optimization is completed, enter the actual execution stage of resource allocation, and apply the optimization results to Mimo communication to ensure the real-time and stability of dynamic resource adjustment;

[0173] The specific execution steps are as follows:

[0174] According to the optimal transmit power allocation result output by the optimization algorithm Dynamically adjust the power of each transmit antenna to meet the channel requirements and power constraints. The power control module monitors the transmit power of each antenna in real time to ensure that the allocated power does not exceed the limit P max , for the antennas with low signal-to-noise ratio, according to the weak signal paths in the channel matrix H, preferentially increase the power to improve their transmission performance;

[0175] Execute the spectrum allocation result, and allocate the frequency band f i and the frequency bandwidth Δf obtained by optimization calculation to the corresponding antenna subsets: f = {f 1 , f 2 ,..., f k}, Δf = {f 1,bw , f 2,bw ,..., f k,bw}, through the dynamic spectrum allocation algorithm, efficiently allocate the broadband resources to the channels with higher priorities to avoid spectrum resource conflicts;

[0176] According to the optimized antenna subset Activate the selected transmit antennas and turn off the unnecessary antennas to reduce power consumption. Dynamically adjust the antenna configuration through the switching matrix, and concentrate the resources on the signal paths with higher signal-to-noise ratio, so as to maximize the channel capacity;

[0177] During the execution of the resource allocation strategy, monitor the change of the signal interference intensity SINR in real time, and adjust the transmit direction of the antennas in combination with the multi-antenna interference alignment technology to minimize the interference to the greatest extent. Use the interference alignment matrix G to optimize the transmit matrix G to minimize the influence of the interference signal on the target receive antenna, specifically as shown in Equation (12):

[0178] G = argmin ∥H·G∥, subject to: ∥G∥ ≤ 1 Equation (12)

[0179] Where G is the interference alignment matrix, H is the channel matrix, and ∥G∥ is the norm of the matrix G;

[0180] During the execution process, the channel state feedback is collected in real time, including the signal-to-noise ratio and channel capacity metrics, and the actual effect of resource allocation is monitored. If it is found that the allocation result does not match the expected performance metrics, for example, the bit error rate is higher than the threshold, the optimization strategy is recalculated and adjusted through rapid iteration to ensure the dynamic adaptability of resource allocation;

[0181] After the allocation execution is completed, the communication performance metrics at the receiving end, such as the bit error rate BER, transmission rate, and delay, are used to verify the strategy effect. If the performance metrics meet the preset requirements, it is confirmed that the current resource allocation strategy can be stably applied; if not, the resource allocation module is triggered to re-optimize.

[0182] Step S3: Signal detection is performed through an improved QR decomposition algorithm and a post-sorting algorithm. The interference signal is aligned to the interference space through interference alignment technology, and the minimum mean square error algorithm is used to suppress the noise; based on the real-time monitoring of the bit error rate, the detection algorithm is dynamically adjusted.

[0183] Among them, in step S3, the following sub-steps are also included:

[0184] S3-1: Signal detection is one of the core tasks in Mimo communication. Especially in an environment with multipath and high interference, the effective decoding of signals is particularly crucial. Based on the optimized resource configuration, the signal detection process includes the following steps:

[0185] In Mimo, the received signal y is composed of the superposition of multiple signal sources and noise, specifically as shown in Equation (13):

[0186] y = Hx + n Equation (13)

[0187] Among them, H is the channel matrix, x is the transmitted signal, and n is the noise vector; according to the received signal y and the channel matrix H, the original signal x is recovered through different detection algorithms;

[0188] Before signal detection, preprocessing of the received signal is required, including denoising and signal amplification. The denoising algorithm suppresses the interference signal by maximizing the signal-to-noise ratio (SNR) to improve the quality of the received signal;

[0189] An improved QR decomposition algorithm is used for signal detection. The received signal matrix is transformed into an upper triangular matrix through QR decomposition, simplifying the calculation process and effectively eliminating the interference between signals;

[0190] For each detected signal, hard decision or soft decision is performed:

[0191] Hard decision: Decode the signal into the most likely symbol, which is suitable for application scenarios with high real-time requirements;

[0192] Soft decision: Provides probability information of the signal, which helps to further optimize the performance, especially in the high bit error rate environment;

[0193] S3-2. Adopt an improved signal detection algorithm, including improved QR decomposition and post sorting algorithm (PSA), to enhance the performance of signal detection in high interference and low signal-to-noise ratio situations;

[0194] Based on the traditional QR decomposition algorithm, adopt an improved Gram-Schmidt orthogonalization process. By selecting the minimum norm of the column vectors of the channel matrix for sorting, the signal-to-noise ratio (SNR) is maximized. The optimized QR decomposition process includes the following steps:

[0195] Calculate the QR decomposition of the channel matrix to obtain the orthogonal matrix Q and the upper triangular matrix R. Rearrange the columns of the channel matrix according to the signal-to-noise ratio to ensure that the signals with the best signal-to-noise ratio are detected first. Simplify the detection process through QR decomposition, reduce the influence of interference signals on signal decoding, and improve the overall performance;

[0196] The PSA algorithm performs eigenvalue sorting on the upper triangular matrix obtained by QR decomposition and selects the optimal path for signal demodulation according to the eigenvalues. In actual operation, the PSA algorithm gradually eliminates the interference of low signal-to-noise ratio paths and adjusts the estimated value of the signal layer by layer, so that the error does not accumulate during the propagation process;

[0197] The key steps of the PSA algorithm include:

[0198] During the signal detection process, first perform QR decomposition on the channel matrix; adjust the receiving order of the signals according to the eigenvalue size of each path, so as to reduce the influence of low signal-to-noise ratio paths; improve the decoding accuracy of the signal by sorting and adjusting the paths;

[0199] Combine the improved QR decomposition and PSA algorithm and dynamically adjust the decoding strategy to cope with the changes in different channel environments; this combined algorithm can still maintain a high signal detection accuracy in the case of poor signal quality, and is especially suitable for complex multipath and dynamic channel environments;

[0200] S3-3. Anti-interference and bit error rate control. Anti-interference and bit error rate control are key steps in the high interference environment. The goal of this stage is to minimize the bit error rate (BER) and ensure the reliable transmission of signals after experiencing multiple interferences;

[0201] In multi-antenna Mimo, through interference alignment technology, align the interferences of multiple signal sources to the interference space, so that the influence of these interference signals on the target receiving antenna is minimized. Separate the interference signals from the useful signals by adjusting the signal direction of the transmitting antenna, and reduce multi-user interference and co-channel interference;

[0202] The MMSE algorithm calculates the optimal filter matrix by introducing the noise covariance matrix and comprehensively considering the effects of signals and noise, as shown in Equation (14):

[0203]

[0204] where G MMSE is the optimal filter matrix, H is the channel matrix, and H H is the conjugate transpose of the channel matrix. σ 2 is the noise power, P is the signal power, and I is the identity matrix. The MMSE algorithm can effectively suppress noise in the case of low signal-to-noise ratio and ensure the correct decoding of signals in the presence of noise interference;

[0205] During the detection process, the bit error rate (BER) is monitored in real time, and the demodulation algorithm is adjusted according to the detection results. If the bit error rate exceeds the set threshold, it will trigger the re-optimization of the signal detection algorithm, including re-selecting the antenna subset, adjusting the transmission power, or reallocating the spectrum. In addition, based on the bit error rate feedback, the resource configuration can be corrected in real time, the signal detection process can be optimized, and the transmission stability can be ensured.

[0206] Step S4: Use the edge node to process part of the computing tasks, perform channel modeling optimization, signal preprocessing, and data compression to reduce the computing burden of the terminal and reduce the latency; optimize the transmission path through multi-path transmission, load balancing, and path switching, and dynamically adjust the transmission strategy in combination with the feedback mechanism.

[0207] Among them, in step S4, the following sub-steps are also included:

[0208] S4-1: In communication, edge computing can significantly improve performance, especially in the case of limited resources. By transferring part of the computing tasks from the terminal device to the edge node, the computing burden of the terminal is reduced and the latency is reduced. The data processing module of the edge node mainly includes the following aspects:

[0209] The edge node receives the channel state information (CSI) from multiple terminal devices and generates the channel matrix H through a fast channel estimation algorithm, such as the minimum mean square error (MMSE) algorithm, to accurately model the channel. At the same time, the edge node uses a method that combines historical data and real-time data to optimize channel prediction and reduce errors. The edge node can also further improve the accuracy of channel estimation through a machine learning model, such as a deep neural network, and provide more accurate channel state information for subsequent resource allocation;

[0210] In the signal preprocessing stage, the edge node is responsible for filtering, denoising, and signal enhancement of the raw data from the terminal device; through advanced digital signal processing technologies, such as adaptive filtering and noise suppression algorithms, the edge node can effectively improve the quality of the received signal; this process can reduce the computational burden of the terminal device and provide a cleaner signal input for signal detection and decoding;

[0211] The edge node compresses the data from the terminal to reduce the amount of data transmission; the compression algorithm can be based on the compressed sensing (CS) theory or other efficient compression technologies to ensure a significant reduction in the bandwidth required for data transmission without losing key data; in addition, the edge node is also responsible for optimizing the data transmission path, selecting the optimal transmission channel, reducing latency, and improving bandwidth utilization to ensure real-time response capabilities in a dynamic environment;

[0212] S4-2, through intelligent path selection and optimization algorithms, the edge node can provide the best communication path for different terminal devices, reducing relay, interference, and transmission latency. The specific steps are as follows:

[0213] Based on real-time channel information and interference analysis, the edge node can dynamically select the optimal transmission path; in a complex network environment, path switching may be required according to the real-time channel state and user requirements; the edge node can judge whether to switch to a better transmission path through real-time monitoring of the channel quality;

[0214] For example, in a mobile communication scenario, the edge node selects the best transmission path according to the user's movement trajectory and the signal strength of adjacent base stations to avoid a decline in transmission quality caused by signal attenuation or interference;

[0215] In the network, the edge node can optimize data transmission through multi-path transmission, using multiple independent paths to transmit data, improving the transmission speed and increasing the network's fault tolerance; the load balancing algorithm reasonably distributes different communication loads to multiple paths to prevent a single path from being overloaded, resulting in transmission latency or packet loss;

[0216] During the path optimization process, the edge node will evaluate the performance of each transmission path in real time, mainly by monitoring indicators such as latency, packet loss rate, bandwidth, and signal-to-noise ratio; after the transmission is completed, the edge node will feedback the transmission performance and further optimize the path selection strategy according to these feedbacks to adapt to the changing network environment;

[0217] S4-3, during the transmission process, it is crucial to ensure the accuracy, stability, and efficiency of the transmission; the transmission result verification module can monitor and feedback the performance of data transmission in real time, so as to adjust and optimize the strategy in a timely manner when anomalies occur. The key steps of this module are as follows:

[0218] During the transmission process, the edge node monitors the key performance indicators (KPIs) of the transmission in real time, such as throughput, transmission rate, bit error rate (BER), latency, and packet loss rate; this monitoring data can help the edge node determine whether the current transmission meets the requirements and evaluate the effectiveness of the current transmission path and resource allocation;

[0219] To ensure the reliability of data transmission, the edge node needs to control the bit error rate; once a high bit error rate is detected, the edge node will initiate an error correction mechanism and adopt processing methods such as retransmission, modulation method adjustment, and automatic gain control (AGC) technology; at the same time, combined with a signal processing module, such as using advanced coding methods like LDPC codes or Turbo codes, to improve the anti-interference ability under poor channel conditions;

[0220] The verification of the transmission result is not only a detection of the data transmission quality but also provides a feedback mechanism; based on the performance monitoring and bit error feedback, the edge node optimizes the current resource allocation and path selection;

[0221] For example, when the detected bit error rate rises, it will automatically trigger a re-routing or transmit power adjustment strategy to reduce interference and improve the accuracy of data transmission; in addition, the edge computing node can also send the feedback information back to the core network to achieve end-to-end performance optimization.

[0222] Step S5, after completing the transmission task, evaluate the system performance through key indicators such as channel capacity, bit error rate, throughput, and latency, and verify the effectiveness of the resource allocation and signal detection strategies; based on the performance evaluation results, optimize the channel modeling and resource allocation strategies using deep learning and reinforcement learning models.

[0223] Among them, in step S5, the following sub-steps are also included:

[0224] S5-1, after completing the transmission task, it is necessary to evaluate the overall performance to verify the effectiveness of the resource allocation and signal transmission strategies and provide data support for subsequent optimization. The performance evaluation mainly starts from the following aspects:

[0225] Calculate the actual channel capacity and verify the improvement effect of the resource allocation strategy on the channel utilization rate, specifically as shown in Equation (15):

[0226]

[0227] Among them, C is the channel capacity, I is the identity matrix, P is the transmit power, σ 2 is the noise power density, H is the channel matrix, H H is the conjugate transpose of the channel matrix;

[0228] Bit Error Rate (BER): Monitor the bit error rate during data transmission to verify the effectiveness of signal detection and anti-interference strategies;

[0229] Throughput: Calculate the amount of effective data transmitted per unit time to evaluate data transmission efficiency;

[0230] Latency: Measure the end-to-end communication latency to ensure meeting real-time requirements;

[0231] Power Consumption: Evaluate the overall energy efficiency to verify the energy-saving effect of power allocation strategies;

[0232] Under different scenarios, such as high-interference environments, dynamic user distributions, and multipath channel conditions, test the key performance indicators to evaluate stability and robustness;

[0233] S5-2, according to the performance evaluation results, combined with the feedback mechanism, perform intelligent optimization on it to continuously improve communication performance. The intelligent optimization is achieved through the following steps:

[0234] Collect the feedback data obtained from performance evaluation, especially the bit error rate, signal-to-noise ratio, and throughput metrics. Input these data into the optimization module to extract abnormal patterns, such as high bit error rate or low signal-to-noise ratio regions;

[0235] Utilize deep learning (DNN) or reinforcement learning (RL) models to retrain according to the feedback data:

[0236] Channel modeling optimization: Adjust the weights of the model or introduce more historical data to improve the accuracy of channel state prediction;

[0237] Resource allocation optimization: Re-optimize the transmission power, antenna selection, and spectrum allocation strategies based on reinforcement learning;

[0238] Introduce a dynamic learning mechanism to enable the model to quickly adapt to new channel environments;

[0239] Dynamically adjust key parameters according to the optimization results, such as increasing the transmission power in high-interference environments, reallocating spectrum resources to avoid interference, and adjusting the configuration of the antenna array to optimize spatial multiplexing performance.

[0240] The above is only the preferred embodiment of the present invention and is not used to limit the present invention. For those skilled in the art, various changes and modifications can be made to the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An intelligent broadband transmission method based on Mimo communication, characterized in that: The following steps are involved: Step S1, by configuring the Mimo communication system, determining the number of transmitting antennas and receiving antennas, the signal frequency range and the array type, and actively collecting channel characteristic data; based on the geometric statistical modeling method, combined with the ray tracing method, a dynamic channel matrix is ​​generated, and a deep neural network is introduced to predict the dynamic changes of the channel state; Step S2, after the channel modeling is completed, the channel state information is obtained in real time, key parameters such as channel capacity, interference intensity and multipath characteristics are calculated, and a correlation matrix of antenna configuration is generated; Based on the reinforcement learning algorithm, the antenna resources, transmission power and spectrum allocation are dynamically optimized to construct a comprehensive objective function that meets multiple constraints. Step S3, performing signal detection by using an improved QR decomposition algorithm and a post-sorting algorithm, aligning the interference signal to the interference space by using an interference alignment technique, and suppressing the noise by using a minimum mean square error algorithm; Based on real-time monitoring of bit error rate, the detection algorithm is dynamically adjusted; Step S4: Use edge nodes to process some computing tasks, perform channel modeling optimization, signal preprocessing, and data compression to reduce the terminal computing burden and delay; optimize the transmission path through multi-path transmission, load balancing, and path switching, and dynamically adjust the transmission strategy in combination with the feedback mechanism; Step S5, after completing the transmission task, evaluate the system performance through key indicators such as channel capacity, bit error rate, throughput and delay to verify the effectiveness of resource allocation and signal detection strategies; Based on the performance evaluation results, deep learning and reinforcement learning models are used to optimize channel modeling and resource allocation strategies.

2. According to claim 1, an intelligent broadband transmission method based on Mimo communication is characterized in that: Wherein step S1 also includes the following sub-steps: S1-1, determine the transmitting antenna M t and receiving antenna M r The number of antennas is set, the transmission frequency range, signal bandwidth and antenna array type are set, and the channel characteristics in the environment are actively detected using the preset training signal sequence. The transmission signal is captured by the receiving antenna array to obtain the initial channel state information, including channel gain, path loss, multipath delay and interference intensity parameters. The core goal of the acquisition is to generate the channel matrix H, which is defined as follows: H= h ij ]M r ×M t Formula (1) Where H is the channel matrix, [h ij ] represents the channel coefficient between the i-th receiving antenna and the j-th transmitting antenna, M t is the number of transmitting antennas, M r is the number of receiving antennas; S1-2, based on the collected channel data, a geometry-based statistical modeling method is selected. In the modeling process, channel parameters are analyzed, including path loss, multipath delay and spatial correlation. The distribution characteristics of scatterers are described by geometric modeling. It is assumed that the scatterer distribution satisfies a certain probability density function, and the scatterer positions are randomly generated. The signal propagation path and its corresponding gain are calculated by combining the ray tracing method. The mathematical expression of the channel model is as follows: Among them, [h ij ] represents the channel coefficient between the i-th receiving antenna and the j-th transmitting antenna, L is the multipath number, α l is the fading coefficient of the lth path, τ1 is the delay of the path, and f is the signal frequency; combining this model, a dynamic channel matrix H(t,f) containing time, frequency and space characteristics is generated; The machine learning optimization module is introduced to use historical channel data and real-time updated channel status to train a deep neural network (DNN) to predict the dynamic change trend of the channel matrix. In the specific implementation process, the historical collection data and the current channel matrix H(t,f) are input into the DNN model, and the model outputs the predicted channel matrix Its objective function is defined as minimizing the error between the predicted matrix and the true matrix, as shown in formula (3): Among them, ζ is the error objective function, N is the number of samples, ||·|| 2 is the matrix norm, H n (t+Δt,f) is the actual channel matrix, is the predicted channel matrix; S1-3, after completing the channel modeling, the generated channel model H is verified. The channel verification process includes performance testing, model comparison and environmental adaptability analysis. Standard performance indicators are designed to evaluate the accuracy and stability of the channel model. These indicators include mean square error and correlation coefficient. The mean square error (MSE) is the evaluation model prediction matrix The difference with the actual measurement matrix H is as follows: Among them, MSE is the mean square error, M t and M r are the number of transmitting antennas and receiving antennas, respectively, ij is the actual channel matrix element, is the predicted channel matrix element; The correlation coefficient (CR) is a measure of the degree to which the model retains channel correlation, as shown in formula (5): Where CR is the correlation coefficient, tr(·) represents the trace of the matrix, ||·|| F is the Frobenius norm, H H is the conjugate transpose of the channel matrix; The generated channel model is compared with the standard channel model to evaluate the performance improvement of the modeling method. The advantages of the geometric statistical model are verified by running different models in the same simulation environment to compare their description accuracy of channel gain, signal-to-noise ratio and multipath characteristics. In the actual communication environment, the adaptability of the model is further verified through measured data, and the modeling results are applied to scenarios in dynamic environments to verify whether the model can effectively cope with changes in channel characteristics. The test includes the dynamic update capability of the channel matrix and its robustness in low signal-to-noise ratio and high interference environments.

3. The intelligent broadband transmission method based on Mimo communication according to claim 1, characterized in that: Wherein, in step S2, the following sub-steps are also included: S2-1, real-time acquisition of channel status information, real-time collection and calculation of key parameters in the channel, and providing data support for dynamic resource allocation. The specific steps are as follows: Using the dynamic channel matrix H(t,f) provided by the channel modeling module, the real-time channel change information is captured through the receiving antenna array; combined with multi-time slot data, the time-derived characteristic of the channel matrix ΔH=H(t+Δt,f)-H(t,f) is calculated to describe the change trend of the channel over time; Based on the collected channel matrix, the instantaneous channel capacity C is calculated as follows: Where C is the instantaneous channel capacity, I is the unit matrix, P is the transmit power, σ 2 is the noise power density, H is the channel matrix, H H is the conjugate transpose of the channel matrix; extract the signal interference ratio (SINR) through the receiving end signal processing unit, and combine it with interference analysis to identify the main interference sources in the current channel and their impact on transmission performance; The multipath components of the signal are separated by time-frequency analysis to obtain the time delay τ of each path. l , gain α l and phase offset, forming a multipath information matrix Θ = {(α l ,τ l )} l =1 L , used for resource allocation optimization; calculate the spatial correlation matrix R between antennas and evaluate the correlation between antennas, as shown in formula (7): R=H H H formula (7) Among them, R is the spatial correlation matrix, H is the channel matrix, and H H is the conjugate transpose of the channel matrix; Eigenvalue analysis is used to evaluate the spatial multiplexing capability of the antenna configuration and help select the optimal antenna subset. The historical channel data is combined to make short-term predictions of future channel states and use the trained deep learning model f DNN (H(t,f)) outputs the channel matrix at the next moment Improve the foresight of resource allocation; S2-2, after obtaining the channel state information, the dynamic resource allocation strategy is designed using the optimization algorithm. According to the performance requirements of the communication, the optimization objective function of resource allocation is constructed. The optimization goal is to maximize the channel capacity C and the signal-to-noise ratio SINR, and minimize the power consumption P and the delay D. The comprehensive objective function is defined as follows: in, is the objective function, C is the channel capacity, SINR is the signal-to-noise ratio, P is the power consumption, D is the delay, ω1, ω2, ω3, ω4 are adjustable weights used to balance the importance of different performance indicators; During the optimization process, the following constraints must be met: power constraint, spectrum constraint and antenna constraint; power constraint is the total transmit power P total The power limit must not be exceeded, as shown in formula (9): Among them, P i is the transmission power of the ith antenna, P max is the maximum power upper limit; The spectrum constraint means that the allocated frequency band f must be within the licensed frequency band, as shown in formula (10): f min ≤f i ≤f max Formula (10) Among them, f i is the frequency band allocated to the spectrum of the ith antenna, f min is the minimum frequency limit, f max is the maximum frequency limit; Antenna constraint means that the number of transmitting antennas k selected should be less than or equal to the total number of antennas M t ; The resource allocation strategy is dynamically optimized using the reinforcement learning (RL) algorithm. In communication, the key elements of reinforcement learning are defined: state, action, and reward. The reward value calculated based on the objective function is as follows: Among them, R is the reward value, is the objective function, Penalty represents the penalty for violating the constraint conditions, and λ is the penalty factor; Use the deep Q learning algorithm to optimize the strategy, estimate the expected reward value of the current state s and action a through the neural network Q(s,a;θ), and improve learning efficiency through experience replay; S2-3, after the resource allocation strategy optimization is completed, the actual execution phase of resource allocation is entered, and the optimization results are applied to Mimo communication. The specific execution steps are as follows: The optimal transmission power allocation result output by the optimization algorithm Dynamically adjust the power of each transmitting antenna to meet channel requirements and power constraints. The power control module monitors the transmit power of each antenna in real time to ensure that the allocated power does not exceed the limit P max ,For low signal-to-noise ratio antennas, according to the weak channel paths in the channel matrix H, the power is increased first to improve its transmission performance; Execute the spectrum allocation result and optimize the calculated frequency band f i The sum bandwidth Δf is allocated to the corresponding antenna subset: f = {f1,f2,...,f k },Δf={f 1, ,f 2, ,...,f k, }, through the dynamic spectrum allocation algorithm, broadband resources are efficiently allocated to channels with higher priority to avoid spectrum resource conflicts; According to the optimized antenna subset Activate the selected transmit antennas and shut down unnecessary antennas, dynamically adjust antenna configurations through the switching matrix, and allocate resources to channel paths with higher signal-to-noise ratios to maximize channel capacity; During the execution of the resource allocation strategy, the change of signal interference intensity SINR is monitored in real time, and the antenna transmission direction is adjusted in combination with the multi-antenna interference alignment technology to minimize the interference. The interference alignment matrix G is used to optimize the transmission matrix G to minimize the impact of the interference signal on the target receiving antenna. Specifically, as shown in formula (12): G=argmin∥H·G∥, subject to: ∥G∥≤1 Formula (12) Where G is the interference alignment matrix, H is the channel matrix, and ∥G∥ is the norm of the matrix G; During the execution process, channel status feedback is collected in real time, including signal-to-noise ratio and channel capacity indicators, to monitor the actual effect of resource allocation. If it is found that the allocation result does not meet the expected performance indicators, such as the bit error rate is higher than the threshold, the optimization strategy is recalculated and adjusted through rapid iteration.

4. The intelligent broadband transmission method based on Mimo communication according to claim 1, characterized in that: Wherein step S3 also includes the following sub-steps: S3-1, signal detection is one of the core tasks in Mimo communication, especially in multipath and interference-intensive environments, effective signal decoding is particularly critical. Based on the optimized resource configuration, the signal detection process includes the following steps: In Mimo, the received signal y is composed of multiple signal sources and noise superimposed, as shown in formula (13): y=Hx+n Formula (13) Where H is the channel matrix, x is the transmitted signal, and n is the noise vector. According to the received signal y and the channel matrix H, the original signal x is restored through different detection algorithms. Before signal detection, the received signal needs to be preprocessed, including denoising and signal amplification. The denoising algorithm suppresses interference signals and improves the quality of the received signal by maximizing the signal-to-noise ratio (SNR); S3-2, adopts an improved signal detection algorithm, including improved QR decomposition and post-sorting (PSA) algorithm, to improve the performance of signal detection under high interference and low signal-to-noise ratio conditions; the optimized QR decomposition process includes the following steps: Calculate the QR decomposition of the channel matrix to obtain the orthogonal matrix Q and the upper triangular matrix R. Rearrange the columns of the channel matrix according to the signal-to-noise ratio to ensure that the signal with the best signal-to-noise ratio is detected first. Simplify the detection process through QR decomposition, reduce the impact of interference signals on signal decoding, and improve the overall performance; The PSA algorithm sorts the eigenvalues ​​of the upper triangular matrix obtained by QR decomposition and selects the optimal path for signal demodulation based on the eigenvalues. In actual operation, the PSA algorithm gradually eliminates the interference of low signal-to-noise ratio paths and adjusts the estimated value of the signal layer by layer, so that the error does not accumulate during the propagation process. The key steps of the PSA algorithm include: In the signal detection process, the channel matrix is ​​firstly decomposed by QR; the order of signal reception is adjusted according to the eigenvalue of each path, thereby reducing the impact of low signal-to-noise ratio path pairs; by sorting and adjusting the paths, the decoding accuracy of the signal is improved; The improved QR decomposition and PSA algorithms are combined to dynamically adjust the decoding strategy to cope with changes in different channel environments. This combined algorithm can maintain high signal detection accuracy even when the signal quality is poor, and is particularly suitable for complex multipath and dynamic channel environments. S3-3, anti-interference and bit error rate control, anti-interference and bit error rate control are key steps in a high-interference environment. The goal of this stage is to minimize the bit error rate (BER) and ensure that the signal can still be transmitted reliably when it passes through multiple interferences; In multi-antenna Mimo, interference from multiple signal sources is aligned to the interference space through interference alignment technology, so that the impact of these interference signals on the target receiving antenna is minimized. By adjusting the signal direction of the transmitting antenna, the interference signal is separated from the useful signal, reducing multi-user interference and co-channel interference; The MMSE algorithm introduces the noise covariance matrix, comprehensively considers the influence of signal and noise, and calculates the optimal filter matrix, as shown in formula (14): Among them, G MMSE is the optimal filter matrix, H is the channel matrix, H H is the conjugate transpose of the channel matrix, σ 2 is the noise power, P is the signal power, and I is the unit matrix; the MMSE algorithm can effectively suppress the noise under low signal-to-noise ratio conditions and ensure that the signal is correctly decoded under noise interference; During the detection process, the bit error rate (BER) is monitored in real time and the demodulation algorithm is adjusted according to the detection results. If the bit error rate exceeds the set threshold, it will trigger the re-optimization of the signal detection algorithm, including reselecting the antenna subset, adjusting the transmission power or reallocating the spectrum. In addition, based on the bit error rate feedback, the resource configuration can be corrected in real time to optimize the signal detection process and ensure transmission stability.

5. The intelligent broadband transmission method based on Mimo communication according to claim 1, characterized in that: Wherein step S4 also includes the following sub-steps: S4-1, using edge computing to improve performance, by transferring some computing tasks from terminal devices to edge nodes, reducing the computing burden of the terminal and reducing latency; the data processing module of the edge node mainly includes the following aspects: The edge node receives channel state information (CSI) from multiple terminal devices and generates a channel matrix H through a fast channel estimation algorithm to accurately model the channel. At the same time, the edge node uses a combination of historical data and real-time data to optimize channel prediction and reduce errors. The edge node can also further improve the accuracy of channel estimation through machine learning models. In the signal preprocessing stage, the edge node is responsible for filtering, denoising and signal enhancement of the raw data from the terminal device. Through advanced digital signal processing technology, the edge node can effectively improve the quality of the received signal. This process can reduce the computing burden of the terminal device and provide cleaner signal input for signal detection and decoding. The edge node compresses the data from the terminal to reduce the amount of data transmission; The compression algorithm can be based on the compressed sensing (CS) theory or other efficient compression technologies to ensure that the bandwidth required for data transmission is greatly reduced without losing key data. In addition, the edge node is also responsible for optimizing the data transmission path, selecting the optimal transmission channel, reducing latency, improving bandwidth utilization, and ensuring real-time response capabilities in dynamic environments. S4-2, through intelligent path selection and optimization algorithms, edge nodes can provide the best communication paths for different terminal devices, reducing relay, interference and transmission delay. The specific steps are as follows: Based on real-time channel information and interference analysis, edge nodes can dynamically select the optimal transmission path; in complex network environments, they can switch paths based on real-time channel status and user needs; edge nodes can determine whether they need to switch to a better transmission path by real-time monitoring of channel quality; In the network, edge nodes can optimize data transmission through multi-path transmission, using multiple independent paths to transmit data, improving transmission speed and increasing network fault tolerance; load balancing algorithms reasonably distribute different communication loads to multiple paths to prevent a certain path from being overloaded and causing transmission delays or packet loss; During the path optimization process, the edge nodes will evaluate the performance of each transmission path in real time, mainly by monitoring the delay, packet loss rate, bandwidth and signal-to-noise ratio indicators; after the transmission is completed, the edge nodes will feedback the transmission performance to the network, and further optimize the path selection strategy based on these feedbacks to adapt to the ever-changing network environment; S4-3, during the transmission process, it is crucial to ensure the accuracy, stability and efficiency of the transmission; the transmission result verification module can monitor and feedback the performance of data transmission in real time, so as to adjust the optimization strategy in time when anomalies occur. The key steps are as follows: During the transmission process, the edge node monitors the key performance indicators (KPIs) of the transmission in real time, such as throughput, transmission rate, bit error rate (BER), latency, and packet loss rate; These monitoring data can help edge nodes determine whether the current transmission meets the requirements and evaluate the effectiveness of the current transmission path and resource allocation; To ensure the reliability of data transmission, edge nodes need to control the bit error rate. Once a high bit error rate is detected, the edge node will start the error correction mechanism, using technologies such as retransmission, modulation adjustment, and automatic gain control (AGC) for processing. At the same time, combined with the signal processing module, such as the use of advanced coding methods such as LDPC code or Turbo code, the anti-interference ability under poor channel conditions is improved. Transmission result verification is not only a test of data transmission quality, but also provides a feedback mechanism; edge nodes optimize current resource allocation and path selection based on performance monitoring and error feedback.

6. The intelligent broadband transmission method based on Mimo communication according to claim 1, characterized in that: Wherein, in step S5, the following sub-steps are also included: S5-1, after completing the transmission task, the overall performance needs to be evaluated to verify the effect of resource allocation and signal transmission strategy. The performance evaluation is mainly carried out from the following aspects: calculating the actual channel capacity and verifying the effect of resource allocation strategy on channel utilization, as shown in formula (15): Where C is the channel capacity, I is the unit matrix, P is the transmit power, σ 2 is the noise power density, H is the channel matrix, H H is the conjugate transpose of the channel matrix; under different scenarios, including high interference environment, dynamic user distribution, multipath channel conditions, the key performance indicators of the test are evaluated for stability and robustness; S5-2, based on the performance evaluation results, combined with the feedback mechanism, intelligent optimization is performed to continuously improve communication performance. Intelligent optimization is achieved through the following steps: collecting feedback data obtained from performance evaluation, including bit error rate, signal-to-noise ratio, and throughput indicators, and inputting these data into the optimization module to extract abnormal patterns, such as high bit error rate or low signal-to-noise ratio areas; Use a deep learning (DNN) or reinforcement learning (RL) model and retrain based on feedback data as follows: Channel modeling optimization: adjust the model weight or introduce more historical data to improve the accuracy of channel state prediction; Resource allocation optimization: re-optimizing transmit power, antenna selection, and spectrum allocation strategies based on reinforcement learning; Introducing a dynamic learning mechanism to enable the model to quickly adapt to new channel environments; Dynamically adjust key parameters based on the optimization results, including increasing the transmit power in an interference environment, reallocating spectrum resources to avoid interference, and adjusting the configuration of the antenna array to optimize spatial multiplexing performance.

Images