Channel estimation and adaptive modulation coding method and system for hybrid communication

Abstract

The invention discloses a channel estimation and adaptive modulation coding method and system for hybrid communication. The method comprises the following steps: for a signal attenuation sequence of the RF / FSO hybrid link, dynamically optimizing key parameters of a signal decomposition algorithm by adopting an optimization algorithm to obtain optimal decomposition parameters; decomposing the attenuation sequence into a plurality of intrinsic mode function (IMF) components representing different physical mechanisms based on the parameter; inputting the IMF component and the real-time environment parameters into a deep learning network, and predicting a signal attenuation value at the next moment through feature extraction and time sequence learning; and finally, according to the predicted value, combining with a preset modulation and coding scheme library, dynamically selecting an optimal modulation and coding scheme through a classification network, and realizing adaptive switching. Through combination of parameter adaptive optimization and deep learning prediction, the problems of poor adaptability and low estimation precision of a traditional method in a complex time-varying channel are solved, and integrated improvement of channel estimation accuracy, transmission reliability and spectrum efficiency is realized.

Description

Technical Field

[0001] This invention relates to the field of wireless communication technology, specifically to a channel estimation and adaptive modulation and coding method and system for hybrid communication. Background Technology

[0002] In wireless communication, radio frequency (RF) communication has become the mainstream method due to its mature technology and widespread application. However, the increasing scarcity of RF spectrum resources has led to a shortage of spectrum resources in communication networks. Secondly, RF communication has a relatively low transmission rate. Furthermore, a single RF link may perform poorly under extreme weather conditions. To address these issues, free-space optical communication (FSO) uses laser or infrared light as the transmission signal, offering advantages such as extremely high bandwidth, low latency, and virtually unlimited spectrum resources. However, FSO communication is only suitable for line-of-sight (LOS) links. Moreover, the stability of FSO links is also limited by the platform on which they are mounted. Finally, a single FSO link is also affected by severe weather conditions such as rain, fog, and turbulence.

[0003] To address the shortcomings of both RF and FSO communication, hybrid RF / FSO communication systems have emerged as a potential solution. These systems combine the long-distance propagation and stability of RF communication with the high bandwidth and low latency advantages of FSO communication to compensate for the limitations of a single communication link. However, in practical applications, hybrid RF / FSO systems often face complex application scenarios and weather conditions. Enabling hybrid RF / FSO communication to intelligently switch mechanisms based on dynamic environmental changes to improve reliability relies heavily on the system's adaptive design—channel estimation and adaptive modulation and coding.

[0004] In recent years, deep learning has demonstrated significant advantages in channel estimation by constructing end-to-end mappings through data-driven approaches, overcoming the constraints of traditional models. For example, in large-scale MIMO systems, traditional algorithms suffer from exponentially increasing computational complexity due to the surge in channel matrix dimensions, while convolutional neural networks (CNNs) achieve efficient dimensionality reduction through local perception and weight sharing. In high-speed mobile scenarios, recurrent neural networks (RNNs) and long short-term memory networks (LSTMs) can capture the temporal correlation of time-varying channels, improving estimation robustness. In low pilot overhead scenarios, generative adversarial networks (GANs) combined with compressed sensing techniques can achieve super-resolution reconstruction of millimeter-wave channels. Adaptive modulation and coding (AMC) plays a crucial role in modern wireless communication systems, dynamically adjusting modulation and coding schemes based on channel conditions to improve system throughput and reliability. Traditional AMC methods typically rely on channel signal integration (CSI) for adjustment; however, this approach may face performance bottlenecks in complex and dynamic wireless channel environments. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a channel estimation and adaptive modulation and coding method and system for hybrid communication. The channel estimation algorithm aims to minimize the combined cost of prediction error and time delay, ensuring high efficiency and accuracy in complex time-varying environments.

[0006] This invention is achieved through the following technical solution: In a first aspect, this application provides a channel estimation and adaptive modulation and coding method for hybrid communication, comprising the following steps: Step 1: For the signal attenuation sequence of the RF and FSO hybrid communication link at the current moment, use an optimization algorithm to dynamically optimize the key decomposition parameters of the signal decomposition algorithm to obtain the optimal decomposition parameters; Step 2: Based on the optimal decomposition parameters, the signal attenuation sequence is decomposed into multiple intrinsic mode functions (IMF) components using the signal decomposition algorithm; Step 3: Input the IMF component and the real-time acquired environmental parameters into the deep learning network model. By extracting channel features and learning their temporal evolution, the signal attenuation value at the next moment can be predicted. Step 4: Based on the predicted signal attenuation value and in conjunction with the preset coding scheme library, dynamically select the modulation and coding scheme to achieve adaptive modulation and coding switching.

[0007] Preferably, the signal decomposition algorithm in step 1 is a time-domain filtered empirical mode decomposition algorithm, and the key decomposition parameters include the bandwidth threshold and the B-spline order.

[0008] Preferably, the dynamic optimization method for the key decomposition parameters in step 1 is as follows: S1. Construct an objective function, which is a weighted sum of signal decomposition delay, envelope entropy, and channel component correlation. The envelope entropy is used to evaluate the spectral separation performance of signal decomposition, and the channel component correlation is used to evaluate the correlation between the intrinsic mode function obtained from the decomposition and the actual channel components of the RF / FSO. S2. Using the snake-heron optimization algorithm, based on the objective function, the bandwidth threshold and B-spline order of the time-domain filtering empirical mode decomposition algorithm are dynamically adjusted through an iterative mechanism combining global search and local search. S3. In each iteration, calculate the objective function value based on the current parameters, and update the search direction and parameter values ​​with minimizing this value as the optimization objective; S4. When the preset number of iterations is reached or the objective function value converges to the threshold, stop the optimization and output the key decomposition parameters.

[0009] Preferably, the expression of the objective function is as follows:

[0010] in, It is the time of signal decomposition. It is the envelope entropy. This indicates the correlation of RF / FSO channel components.

[0011] Preferably, in step 2, synchronous decomposition is performed in the time and frequency domains according to the physical mechanism of the signal attenuation sequence; The RF attenuation signal is decomposed to obtain three IMF components, which correspond to the low-frequency component of rainfall attenuation, the high-frequency component of rainfall attenuation, and the multipath attenuation component, respectively. The FSO attenuation signal is decomposed to obtain five IMF components, which correspond to the pointing error attenuation component, the low-frequency precipitation attenuation component, the large-scale turbulence attenuation component, the high-frequency precipitation attenuation component, and the small-scale turbulence attenuation component, respectively.

[0012] Preferably, the deep learning network model is a CNN-BIGRU-Attention network, and its prediction method is as follows: The IMF components and environmental parameters are input into the network, and spatial channel features are extracted sequentially through convolutional layers, channel evolution laws are learned by combining time-series information through bidirectional gated recurrent unit layers, key feature weights are enhanced through attention mechanisms, and finally the features are integrated through fully connected layers to output the signal attenuation value of the RF and FSO links at the next time step.

[0013] Preferably, the preset coding scheme library in step 4 includes an RF modulation and coding scheme library and an FSO modulation and coding scheme library. Each scheme library contains multiple modulation and coding schemes, and each modulation and coding scheme is associated with a signal attenuation threshold range.

[0014] Preferably, the adaptive modulation and coding switching includes the following process: Based on the predicted signal attenuation value, determine the signal attenuation threshold range it belongs to in the preset RF modulation and coding scheme library and FSO modulation and coding scheme library; The transition probability of switching from the current modulation and coding scheme to each candidate modulation and coding scheme is calculated using a deep learning classification network. Based on the maximum transition probability, the optimal modulation and coding scheme is dynamically selected from the coding scheme library and switched to maximize the spectrum utilization while meeting the target bit error rate.

[0015] Secondly, this application provides a channel estimation and adaptive modulation and coding system for hybrid communication, comprising: The decomposition parameter optimization module is used to dynamically optimize the key decomposition parameters of the signal decomposition algorithm for the signal attenuation sequence of the hybrid communication link of RF and FSO at the current moment, so as to obtain the optimal decomposition parameters. The signal decomposition module is used to decompose the signal attenuation sequence into multiple intrinsic mode function (IMF) components based on the optimal decomposition parameters and the signal decomposition algorithm. The prediction module is used to input the IMF component and the real-time acquired environmental parameters into a deep learning network model, and predict the signal attenuation value at the next moment by extracting channel features and learning their temporal evolution law. The encoding switching module is used to dynamically select a modulation and coding scheme based on the predicted signal attenuation value and in combination with a preset encoding scheme library, so as to realize adaptive modulation and coding switching.

[0016] Thirdly, this application provides an electronic device, comprising: Memory, used to store computer programs; A processor is configured to implement the channel estimation and adaptive modulation and coding method for hybrid communication when executing the computer program.

[0017] Compared with the prior art, the present invention has the following beneficial technical effects: This application provides a channel estimation and adaptive modulation and coding scheme method for hybrid communication. First, it abandons the traditional approach of fixing the parameters of the decomposition algorithm, instead dynamically adapting the algorithm to the current channel state to ensure the accuracy of signal decomposition. Second, it fuses the decomposed IMF components, representing different physical attenuation mechanisms, with real-time environmental parameters and inputs them into a deep learning network for feature extraction and temporal pattern learning, achieving high-precision prediction of channel attenuation. Finally, based on this prediction, it drives a classification network to dynamically select the optimal modulation and coding scheme from a pre-defined scheme library. This scheme, on the one hand, significantly improves the accuracy and environmental adaptability of channel estimation through dynamic parameter optimization and deep learning prediction, overcoming the performance bottleneck of traditional models in rapidly changing channels; on the other hand, it deeply couples channel estimation with modulation and coding decision-making, using predictive information for forward-looking decision-making, achieving a dynamic optimal balance between transmission reliability (bit error rate guarantee) and spectral efficiency (throughput maximization), thereby comprehensively improving the intelligence level and overall performance of the hybrid communication system.

[0018] This application also proposes a hybrid communication channel estimation and adaptive modulation and coding system, an electronic device, and a computer storage medium, which possess all the advantages of the aforementioned hybrid communication channel estimation and adaptive modulation and coding methods. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of the channel estimation method for hybrid communication according to the present invention; Figure 2 This is a time-domain and frequency-domain IMF decomposition diagram of the TVFEMD of this invention; The left figure shows the time-domain decomposition IMF, and the right figure shows the frequency-domain decomposition IMF. Figure 3 This is a flowchart illustrating the optimization of key decomposition parameters in this invention. Figure 4 The following curves show the changes in bandwidth threshold and order with the number of SBOA iterations during the TVFEMD decomposition process of this invention; The first image on the left shows the trend of bandwidth threshold change, the second image shows the trend of order change, and the third image shows the trend of fitness change. Figure 5 This invention provides a comparison of IMF correlation coefficients before and after SBOA optimization. Figure 6 This is an architecture diagram of the CNN-BIGRU-Attention deep learning network model of this invention; Figure 7 This is a flowchart of the prediction process of the CNN-BIGRU-Attention deep learning network model of the present invention. Figure 8 This is a flowchart of the deep learning-based RF / FSO adaptive transmission process of the present invention. Figure 9 This is a flowchart of the deep learning-based RF / FSO adaptive modulation and coding process of this invention.

[0021] Figure 10 This is a diagram showing the RF link adaptive modulation and coding switching results of the present invention; The left figure shows the RF rainfall attenuation sequence, and the right figure shows the attenuation intervals corresponding to each MCS. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0023] Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0024] A channel estimation and adaptive modulation and coding method for hybrid communication includes the following steps: Step 1: Obtain the signal attenuation sequence of the hybrid communication link of radio frequency (RF) and free space optical (FSO) at the current moment; Step 2: The Snake-Heron Optimization Algorithm (SBOA) is used to dynamically optimize the key decomposition parameters of the Time-Domain Filtered Empirical Mode Decomposition Algorithm (TVFEMD) to determine the optimal key decomposition parameters, which include the bandwidth threshold and the B-spline order. The Snake-Heron Optimization Algorithm (SBOA) is used to dynamically optimize the key parameters (bandwidth threshold and B-spline order) of TVFEMD decomposition. By minimizing the comprehensive objective function of decomposition delay, envelope entropy and channel component correlation, it provides an adaptive optimal combination of decomposition parameters for signal decomposition, ensuring decomposition accuracy and efficiency.

[0025] Step 3: Using the aforementioned optimal key decomposition parameters, the RF / FSO attenuation sequence is decomposed in the time and frequency domains according to the physical mechanism of the RF / FSO attenuation sequence using the TVFEMD algorithm, thereby decomposing the RF / FSO attenuation sequence into different IMF components.

[0026] Based on the optimized decomposition parameters, the TVFEMD algorithm is used to decompose the hybrid attenuation signal into multiple intrinsic mode function components in the time and frequency domains. Each component corresponds to a physical attenuation mechanism (such as pointing error, rainfall, turbulence), thereby realizing the physical decoupling and structured extraction of composite channel characteristics.

[0027] Step 4: Input the IMF component and the real-time acquired environmental parameters into the CNN-BIGRU-Attention deep learning network model; the model extracts channel features and learns their temporal evolution rules to predict the signal attenuation value of the RF and FSO links at the next moment; The IMF components obtained from the decomposition and the real-time environmental parameters are input into the CNN-BIGRU-Attention deep learning network. The CNN extracts spatial features, the BiGRU learns the temporal evolution law, and the Attention mechanism focuses on key information to finally predict the signal attenuation value of the RF and FSO links at the next moment.

[0028] Step 5: Based on the predicted signal attenuation value, and in conjunction with the preset RF modulation and coding scheme library and FSO modulation and coding scheme library, a parallel deep learning classification network is used to dynamically select the modulation and coding scheme to achieve adaptive modulation and coding switching, so as to maximize the spectrum utilization while meeting the target bit error rate.

[0029] Based on the predicted attenuation value and combined with a predefined RF / FSO modulation and coding scheme library, the optimal modulation and coding scheme is dynamically selected through a parallel deep learning classification network. This maximizes spectrum utilization while ensuring the target bit error rate, thus completing the adaptive transmission decision.

[0030] This method first performs time-frequency decomposition on the real-time acquired signal attenuation sequence using an adaptive parameter-based snake-and-heron optimization algorithm, decoupling the mixed attenuation signal into components corresponding to different physical mechanisms. Then, through a CNN-BIGRU-Attention deep learning network, it integrates environmental parameters to extract channel features and learns their temporal evolution patterns, achieving high-precision attenuation prediction for the next time step. Finally, based on the prediction results, a parallel classification network dynamically selects the optimal transmission scheme from a pre-defined modulation and coding scheme library. This method effectively solves the problems of fixed parameters and poor adaptability in dynamic channels encountered by traditional methods, significantly improving spectrum utilization efficiency and overall system performance while ensuring communication reliability.

[0031] Example 1 A channel estimation and adaptive modulation and coding method for hybrid communication includes the following steps: Step 1: Obtain the signal attenuation sequence at the current moment.

[0032] The signal attenuation sequence is the signal acquired at the current moment by the hybrid communication system of radio frequency (RF) and free space optical (FSO), and this signal is used as the input signal. This attenuation sequence reflects the impact of complex time-varying atmospheric channels (such as rainfall, turbulence, pointing errors, etc.) on signal transmission.

[0033] Step 2: The Snake-Heron Optimization Algorithm (SBOA) is used to dynamically optimize the key decomposition parameters of the Time-Domain Filtered Empirical Mode Decomposition Algorithm (TVFEMD) to determine the optimal key decomposition parameters, including the bandwidth threshold. and B-spline order The TVFEMD algorithm controls the decomposition accuracy and aliasing degree by using optimal key decomposition parameters.

[0034] See Figure 3 The optimization methods for the key decomposition parameters are as follows: S2.1. Construct an objective function based on the signal decomposition time, error, and correlation of RF / FSO channel components, as shown in the following expression:

[0035] in, This is the signal decomposition time (delay), i.e. the time spent in the decomposition process. During optimization, we want this value to be as small as possible to ensure the real-time performance of signal processing.

[0036] It is the envelope entropy, used to measure the spectral separability of a signal. For a given signal decomposition, The smaller the value, the better the components of the decay sequence are separated, and the higher the decomposition accuracy. The expression for envelope entropy is as follows:

[0037] in, It is the probability distribution of the normalized envelope of the signal.

[0038] The correlation of the RF / FSO channel components is used to evaluate whether the Intrinsic Mode Function (IMF) obtained from the decomposition accurately reflects the RF / FSO channel correlation. The decomposed IMF can be consistent with the actual correlation of the RF and FSO signals, thus enabling the use of channel correlation to improve the accuracy of channel estimation. The correlation expression is as follows:

[0039] in, It is the covariance of the IMF and RF / FSO components. Each represents its own standard deviation.

[0040] S2.2. With minimizing the objective function as the goal, the Snake-Heron Optimization Algorithm (SBOA) is adopted. By combining global search and local search, the key decomposition parameters of the input RF / FSO attenuation sequence are dynamically and iteratively adjusted to obtain the optimal key decomposition parameters. This allows the decomposition results to more accurately separate the attenuation components caused by different channel factors in the time and frequency domains.

[0041] The SBOA algorithm employs a mechanism that combines global search and local search to dynamically adjust the bandwidth threshold. The SBOA algorithm uses B-splines to optimize the objective function, achieving an optimal balance between decomposition time, error control, and correlation. First, for the input RF / FSO decay sequence, the bandwidth threshold and B-spline order are initialized, and its fitness is calculated. The fitness consists of two parts, reflecting the SBOA algorithm's performance in local and global optimization, respectively. The fitness is mathematically expressed as:

[0042] The local optimization update strategy in the solution space is as follows:

[0043] The global optimization and update strategy is as follows:

[0044] In the above formula, These are the search step sizes for local optimization and global optimization, respectively. This represents the set of parameters to be optimized. This is the globally optimal solution. This represents the objective function value of the optimal solution.

[0045] The S2.3 and SBOA algorithms iterate repeatedly until a stopping criterion (determined by the optimization time and objective function) is reached, thus obtaining the optimal critical decomposition parameter, i.e., the bandwidth threshold. and B-spline order .

[0046] Step 3: Based on the optimal key decomposition parameters, the TVFEMD decomposition method decomposes the RF / FSO attenuation sequence in the time and frequency domains according to the physical mechanism of the RF / FSO attenuation sequence, decomposing the RF / FSO attenuation sequence into different IMF components.

[0047] TVFEMD decomposition was performed on the RF / FSO decay sequence to obtain multiple intrinsic mode functions (IMFs) of the decay sequence.

[0048] TVFEMD is decomposed simultaneously in the time and frequency domains to obtain multiple intrinsic mode functions (IMFs).

[0049] The RF attenuation signal is decomposed into three IMFs: the low-frequency component of rainfall attenuation, the high-frequency component of rainfall attenuation, and the multipath attenuation component.

[0050] The FSO attenuation signal is decomposed into five IMFs: pointing error attenuation component, rainfall attenuation low-frequency component, large-scale turbulence attenuation component, rainfall attenuation high-frequency component, and small-scale turbulence attenuation component.

[0051] Time-Variant Filtered Empirical Mode Decomposition (TVFEMD) is an adaptive signal decomposition method for processing non-stationary signals, especially for the analysis of linear and nonlinear signals. It is an extension of traditional Empirical Mode Decomposition (EMD). Compared to EMD, TVFEMD improves the frequency separation performance of signal decomposition by introducing a time-varying filter and maintains stability at lower sampling rates. EMD algorithms typically decompose a signal into a set of Intrinsic Mode Functions (IMFs), each IMF representing a specific frequency component of the signal. TVFEMD, however, introduces a time-varying filter, allowing the decomposition process to be dynamically adjusted according to the time-varying characteristics of the signal. TVFEMD can extract individual mode components more accurately for dynamically changing signals, especially FSO signals under varying environments.

[0052] In TVFEMD, the input decay sequence It is decomposed into several intrinsic mode functions (IMFs) and a residual trend component. The mathematical expression is as follows:

[0053] The key to successful TVFEMD implementation lies in selecting the appropriate filter type, cutoff frequency range, and number of decomposition layers. These parameters significantly impact the accuracy and efficiency of signal decomposition. Under varying rainfall conditions, these parameters need to be adjusted accordingly to optimize the TVFEMD decomposition effect.

[0054] The function of a time-varying filter is to extract different frequency components from a signal by applying time-varying filtering. Under rainfall conditions, signal attenuation leads to changes in different frequency bands, especially affecting the low-frequency and high-frequency components. When the rainfall intensity is high, the low-frequency component may be significantly affected; therefore, it is necessary to select an appropriate wavelet basis to extract the low-frequency and high-frequency components of the signal.

[0055] The cutoff frequency determines which frequency components of a signal are retained and which are filtered out during signal decomposition. Under rainfall conditions, the signal's frequency components are affected by raindrop scattering and path loss; therefore, the cutoff frequency should be dynamically adjusted in real-time according to the rainfall intensity. In heavy rain or torrential rain, the high-frequency components of the signal experience greater attenuation, thus requiring a lower high-frequency cutoff value to retain more low-frequency components. This approach adapts to signal changes caused by rainfall. The high-frequency cutoff value gradually decreases as rainfall intensity increases, allowing the filter to automatically adjust the frequency band it processes, thereby maintaining the accuracy of signal decomposition.

[0056] The number of decomposition levels determines the depth of signal decomposition in the TVFEMD algorithm. In a signal, mode components caused by various channel factors may exhibit aliasing, especially in complex turbulent environments where high-frequency and low-frequency components may mix together. The number of decomposition levels should be determined based on the degree of mode aliasing. When rainfall intensity is high, more decomposition levels are needed for both low-frequency and high-frequency components to accurately extract each mode. Increasing the number of decomposition levels ensures that the frequency components of each mode in the signal are accurately decomposed, avoiding aliasing.

[0057] Step 4: Based on the IMF component of the RF / FSO attenuation sequence and combined with a neural network model, predict the signal attenuation value at the next moment.

[0058] The IMF components obtained from TVFEMD decomposition, along with real-time acquired environmental parameters, are used as input features and fed into a CNN-BIGRU-Attention network. This network extracts spatial channel features from the IMF components through convolutional layers (CNNs) and utilizes bidirectional gated recurrent unit (BiGRU) layers to learn the temporal dependence and dynamic evolution of channel features by incorporating temporal information from the environmental parameters. An attention mechanism further enhances the influence weights of key features. Finally, the model synthesizes the extracted channel features and their evolution to predict the signal attenuation value at the next time step.

[0059] The CNN-BIGRU-Attention network consists of convolutional layers (CNN), bidirectional gated recurrent unit (BiGRU) layers, and an attention mechanism layer. It extracts spatial features from the imaginary function elements (IMFs), captures temporal dependencies, and performs feature fusion and weighting of key features based on correlation. Finally, it predicts the RF / FSO signal decay value at the next time step and outputs the correlation coefficients between each IMF and the original signal, providing a basis for adaptive modulation and coding decisions. The network structure includes an input layer, convolutional layers, GRU layers, an attention mechanism layer, fully connected layers, and an output layer. Through training, it achieves accurate prediction of temporal decay. The CNN-BIGRU-Attention deep learning network model receives inputs including RF / FSO decomposition components, rainfall intensity, and temperature. The data is organized in terms of time step and number of channels, meaning each time step contains data from multiple channels. The entire prediction network consists of an input layer, convolutional layers (CNN), GRU layers, an attention mechanism layer, fully connected layers, and an output layer. The network structure is as follows: Figure 6 As shown.

[0060] See Figure 7The input layer takes temporal IMF data as input and defines the dimensions of the input features. The CNN extracts channel features and consists of two convolutional layers and a max-pooling layer. The first layer has a kernel width of 3 and 32 channels; the second layer has a kernel width of 3 and 64 channels. The max-pooling layer enhances important channel features and reduces computation through window sampling, with a pooling window size of 2 and a stride of 2. The GRU layer learns the temporal dynamic evolution of channel features under rainfall conditions using a bidirectional GRU, thus quickly adapting to the characteristics of long-term rainfall in unfamiliar areas; it has 4 hidden units. The attention mechanism layer has 4 heads and assists the model in increasing the importance weights of abruptly changing features to improve prediction accuracy when rainfall intensity changes abruptly. The fully connected layer consists of two layers with ReLU activation. The first layer has 16 neurons, and the second layer has 64 neurons, gradually increasing the number of units to extract more abstract channel features. The final layer is the output layer, used for prediction, outputting the predicted signal attenuation value for the next time step.

[0061] Step 5: Based on the predicted signal attenuation value, the preset RF and FSO modulation and coding scheme library, through a parallel deep learning classification network, dynamically selects the modulation and coding scheme with the best spectrum utilization and that meets the target bit error rate requirements according to the channel attenuation threshold range where the predicted attenuation value is located, so as to realize the real-time adaptive switching of communication parameters, thereby ensuring link reliability and maximizing transmission efficiency in complex time-varying channels.

[0062] The adaptive modulation and coding scheme (EMC) functionality comprises two parts: an EMC library and an adaptive handover algorithm. The EMC library provides EMC combinations with maximum spectral efficiency under different channel conditions. Based on the attenuation prediction value estimated by the channel, the adaptive handover algorithm determines the EMC within the attenuation threshold range and performs the handover.

[0063] See Figure 8 and 9By combining a pre-defined modulation and coding scheme library (MCS library) and correlation coefficients, an adaptive dynamic switching of the modulation and coding scheme is achieved through a parallel classification network. This maximizes spectrum utilization while ensuring bit error rate performance, thereby improving the reliability and transmission efficiency of the hybrid communication system. This parallel deep learning classification network is identical to the CNN-BIGRU-Attention network model. The input layer of the classification network takes the correlation coefficients of each IMF component with the original attenuation sequence as input. After the RF / FSO attenuation sequence is optimized by SBOA and decomposed into TVFEMD, two results are generated: the IMF and the correlation coefficient. The former is input into the prediction network for channel estimation, while the latter is input into the classification network and combined with the channel estimation result of the former for MCS switching. The output layer adds a classification layer and a Softmax activation function to transform the original output into a probability distribution for the final MCS selection.

[0064] When the channel attenuation timing changes, the adaptive modulation and coding technique changes from the current... Switch to Follow the principle of transition probability. Transition probability refers to the current... Switch to The probability of that outcome is only related to the probability of that outcome. It is relevant but irrelevant to subsequent states. Assume it corresponds to the current state. The set of channel attenuation values ​​is The channel attenuation sample value at the next moment is , yes If any value is taken from the interval, then The probability can be expressed as:

[0065] from hour, arrive The transition probability can be expressed as:

[0066] in, ,therefore, The transition probability matrix of the switching can be expressed as:

[0067] According to the next moment The transition probability matrix yields the switching options for transitioning to the threshold modulation and coding scheme under the current channel state. The output layer of the classification network selects this option based on the maximum transition probability. .

[0068] The following is a library of RF link modulation and coding schemes:

[0069] FSO Link Modulation and Coding Scheme Library

[0070] See Figure 2 The image shows the TVFEMD time-domain and frequency-domain IMF decomposition diagrams of a 250-millisecond FSO attenuation sequence. IMFs 1-5 correspond to the pointing error attenuation component, the low-frequency component of rainfall attenuation, the large-scale turbulence attenuation component, the high-frequency component of rainfall attenuation, and the small-scale turbulence attenuation component, respectively. In the time domain, the five components mainly reflect the difference in fading rate from low to high; while in the frequency domain, the five components mainly reflect the difference in the distribution frequency band and the magnitude of attenuation. Through TVFEMD decomposition, we can extract the characteristics of pointing, rainfall, and turbulence factors in the time and frequency domains from an FSO attenuation sequence, thereby understanding the channel state of the entire joint channel.

[0071] See Figure 4 In the SBOA-TVFEMD decomposition of a 250ms FSO decay sequence sample, the bandwidth threshold With order The curve showing the change in the number of SBOA iterations, after several iterations of optimization, shows the bandwidth threshold. With order The optimal combination of decomposition parameters for the corresponding FSO decay sequence sample is the one that remains unchanged until it reaches a plateau. (Bandwidth threshold) The range of the IMF component envelope is controlled to fully reveal the spectral energy characteristics within each IMF component's bandwidth. For IMF components, an excessively large bandwidth threshold... Too small an order will cause the IMF to be contaminated with too many channel characteristics, while too small an order will cause the IMF to lose its channel characteristics. The complexity of the basis functions controls the fit and smoothness of the IMF components; too low an order leads to underfitting of the IMF components. After SBOA optimization, each IMF component is as smooth as possible while preserving channel characteristics to reduce high-frequency noise and achieve optimal decomposition.

[0072] For the TVFEND decomposition algorithm, the ability to completely decompose the signal attenuation is a crucial criterion for evaluating its performance. Specifically, this is reflected in the resulting IMFs (Integrated Metrics Factors). The lower the similarity between each IMF and the original attenuation sequence, and the smaller the residual ratio after decomposition, the better the performance of the current bandwidth threshold. With order The better the fitness of the sequence samples, the higher the decomposition accuracy of the TVFEMD algorithm. Minimum envelope entropy is used as a measure of fitness; envelope entropy measures the degree of change in the signal envelope to assess complexity. For envelope entropy, the smaller the entropy value, the smoother the signal change, the lower the IMF complexity, and the better the decomposition quality. Figure 4 The current bandwidth threshold is given. With order The fitness value changes with the number of SBOA optimization iterations. As the number of iterations increases, the fitness value gradually converges to the minimum value, the similarity between each IMF and the original sequence samples continuously decreases, and the decomposition accuracy is gradually optimized.

[0073] To demonstrate the above conclusions, the 250ms FSO decay sequence samples were subjected to TVFEMD decomposition and SBOA-optimized TVFEMD decomposition, respectively. The Pearson correlation coefficients of each IMF and residual with the original decay sequence samples were then calculated and compared. Figure 5 It can be seen that the correlation coefficients of each IMF obtained by the TVFEMD decomposition optimized by SBOA are lower than those of the single TVFEMD decomposition, and the residual (IMF6) is only one-third of that of the residual of the single TVFEMD decomposition. This calculation proves that the SBOA algorithm is effective and significant in optimizing the performance of TVFEMD decomposition by optimizing the bandwidth threshold. With order It can perform the highest-precision decomposition of arbitrarily varying decay sequences. This is of great significance for processing long-term, large-volume, and highly variable time-series decay sequences, and greatly helps to improve the accuracy of subsequent channel estimation results.

[0074] This paper uses adaptive modulation and coding schemes (MCS) for RF links as an example of the MCS switching principle based on a deep learning classification network. A library of MCS schemes for RF links under 20mm / h rainfall conditions is presented. To visually demonstrate the impact of different MCSs on RF link performance under rainfall conditions, a graph showing the applicable attenuation range of different MCSs in a rainfall attenuation sequence is plotted. Figure 10 As can be seen from this, when the channel attenuation value increases, the target bit error rate is satisfied. The spectral efficiency of the corresponding MCS gradually decreases. From a time perspective, the real-time changing RF link attenuation should be matched with different MCS to meet the required bit error rate performance.

[0075] This embodiment presents a hybrid communication channel estimation and adaptive modulation and coding method, mainly including attenuation signal input, SBOA parameter optimization, TVFEMD signal decomposition, CNN-BIGRU-Attention network training and prediction, and outputting predicted attenuation. This channel estimation algorithm aims to minimize the combined cost of prediction error and time delay, ensuring efficiency and accuracy in complex time-varying environments. When the RF / FSO attenuation sequences are input into the SBOA-TVFEMD-CNN-BIGRU-Attention channel estimation algorithm framework, the SBOA algorithm matches the optimal TVFEMD decomposition parameters based on the input signal type and features. According to the optimized SBOA decomposition parameters, the TVFEMD signal decomposition module decomposes the RF / FSO signal into different IMF components in the time and frequency domains according to the physical mechanism of its channel factors. All IMF components are input in parallel into the CNN-BIGRU-Attention network. The deep learning network extracts channel features from the IMFs and performs feature fusion and enhancement based on correlation, finally outputting the attenuation value of the RF / FSO at the next time step.

[0076] This method consists of two parts: a time-domain filtered empirical mode decomposition (TVFEMD) algorithm and a CNN-BIGRU-Attention network. First, the TVFEMD algorithm decomposes the input RF / FSO channel attenuation sequence according to a joint channel factor, extracting time-domain and frequency-domain features corresponding to different channel factors. Then, the decomposed intrinsic mode functions are input into the CNN-BIGRU-Attention network for deep feature enhancement and fusion, finally outputting prediction results at different adaptive stages. To improve the generalization ability and prediction accuracy of the hybrid architecture, the SBOA optimization algorithm is further introduced to dynamically optimize the hyperparameters in the TVFEMD decomposition process, enabling the system to adjust according to real-time changes in channel state, thereby meeting the high-precision requirements of channel estimation and adaptive modulation and coding switching in RF / FSO hybrid communication systems.

[0077] To reduce system complexity and redundant design, this method divides the adaptive task of RF / FSO hybrid communication into two main parts: channel estimation and adaptive modulation and coding, by adjusting the input layer structure and activation function configuration of the hybrid architecture. Channel estimation is directly achieved by the hybrid architecture through prediction of the RF / FSO attenuation sequence. The adaptive modulation and coding function includes two modules: a modulation and coding scheme library and an adaptive switching algorithm. The modulation and coding scheme library pre-stores modulation and coding combinations with optimal spectral efficiency under different channel conditions, while the adaptive switching algorithm determines the attenuation threshold range to which the predicted value belongs based on the attenuation output of the channel estimation and dynamically switches to the appropriate modulation and coding scheme accordingly.

[0078] Example 2 Correspondingly, this application provides a channel estimation and adaptive modulation and coding system for hybrid communication, including: The decomposition parameter optimization module is used to dynamically optimize the key decomposition parameters of the signal decomposition algorithm for the signal attenuation sequence of the hybrid communication link of RF and FSO at the current moment, so as to obtain the optimal decomposition parameters. The signal decomposition module is used to decompose the signal attenuation sequence into multiple intrinsic mode function (IMF) components based on the optimal decomposition parameters and the signal decomposition algorithm. The prediction module is used to input the IMF component and the real-time acquired environmental parameters into a deep learning network model, and predict the signal attenuation value at the next moment by extracting channel features and learning their temporal evolution law. The encoding switching module is used to dynamically select a modulation and coding scheme based on the predicted signal attenuation value and in combination with a preset encoding scheme library, so as to realize adaptive modulation and coding switching.

[0079] It should be noted that, in the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another device, or some features may be ignored or not executed. The modules described as separate components may or may not be physically separated. The components shown as modules may be one or more physical units, that is, they may be located in one place or distributed in multiple different places. Some or all of the modules can be selected to achieve the purpose of the solution in this embodiment according to actual needs.

[0080] Furthermore, in the various embodiments of the present invention, the modules can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The integrated unit described above can be implemented in hardware or as a software functional unit.

[0081] An electronic device provided in this application includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps of the channel estimation and adaptive modulation and coding method for hybrid communication as described in any of the above embodiments.

[0082] Another electronic device provided in this application embodiment may further include: an input port connected to a processor for transmitting multimodal data collected by an external acquisition device to the processor; a display unit connected to the processor for displaying the processor's processing results to the outside world; and a communication module connected to the processor for enabling communication between the electronic device and the outside world. The display unit may be a display panel, a laser scanning display, etc.; the communication method adopted by the communication module includes, but is not limited to, Mobile High Definition Link (HML), Universal Serial Bus (USB), High Definition Multimedia Interface (HDMI), and wireless connection (including Wi-Fi, Bluetooth, Bluetooth Low Energy, and IEEE 802.11s-based communication technology).

[0083] This application provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the steps of the channel estimation and adaptive modulation and coding method for hybrid communication as described in any of the above embodiments.

[0084] For descriptions of relevant parts in the hybrid communication channel estimation and adaptive modulation and coding system, electronic device, and computer-readable storage medium provided in the embodiments of this application, please refer to the detailed description of the corresponding parts in the hybrid communication channel estimation and adaptive modulation and coding method provided in the embodiments of this application, which will not be repeated here. Furthermore, parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of corresponding technical solutions in the prior art have not been described in detail to avoid excessive elaboration.

[0085] The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A channel estimation and adaptive modulation and coding method for hybrid communication, characterized in that, Includes the following steps: Step 1: For the signal attenuation sequence of the RF and FSO hybrid communication link at the current moment, use an optimization algorithm to dynamically optimize the key decomposition parameters of the signal decomposition algorithm to obtain the optimal decomposition parameters; Step 2: Based on the optimal decomposition parameters, the signal attenuation sequence is decomposed into multiple intrinsic mode functions (IMF) components using the signal decomposition algorithm; Step 3: Input the IMF component and the real-time acquired environmental parameters into the deep learning network model. By extracting channel features and learning their temporal evolution, the signal attenuation value at the next moment can be predicted. Step 4: Based on the predicted signal attenuation value and in conjunction with the preset coding scheme library, dynamically select the modulation and coding scheme to achieve adaptive modulation and coding switching.

2. The channel estimation and adaptive modulation and coding method for hybrid communication according to claim 1, characterized in that, The signal decomposition algorithm described in step 1 is a time-domain filtered empirical mode decomposition algorithm, and the key decomposition parameters include the bandwidth threshold and the B-spline order.

3. The channel estimation and adaptive modulation and coding method for hybrid communication according to claim 2, characterized in that, The dynamic optimization method for the key decomposition parameters mentioned in step 1 is as follows: S1. Construct an objective function, which is a weighted sum of signal decomposition delay, envelope entropy, and channel component correlation. The envelope entropy is used to evaluate the spectral separation performance of signal decomposition, and the channel component correlation is used to evaluate the correlation between the intrinsic mode function obtained from the decomposition and the actual channel components of the RF / FSO. S2. Using the snake-heron optimization algorithm, based on the objective function, the bandwidth threshold and B-spline order of the time-domain filtering empirical mode decomposition algorithm are dynamically adjusted through an iterative mechanism combining global search and local search. S3. In each iteration, calculate the objective function value based on the current parameters, and update the search direction and parameter values ​​with minimizing this value as the optimization objective; S4. When the preset number of iterations is reached or the objective function value converges to the threshold, stop the optimization and output the key decomposition parameters.

4. The channel estimation and adaptive modulation and coding method for hybrid communication according to claim 3, characterized in that, The expression for the objective function is as follows: in, It is the time of signal decomposition. It is the envelope entropy. This indicates the correlation of RF / FSO channel components.

5. The channel estimation and adaptive modulation and coding method for hybrid communication according to claim 1, characterized in that, In step 2, synchronous decomposition is performed in the time and frequency domains according to the physical mechanism of the signal attenuation sequence; The RF attenuation signal is decomposed to obtain three IMF components, which correspond to the low-frequency component of rainfall attenuation, the high-frequency component of rainfall attenuation, and the multipath attenuation component, respectively. The FSO attenuation signal is decomposed to obtain five IMF components, which correspond to the pointing error attenuation component, the low-frequency precipitation attenuation component, the large-scale turbulence attenuation component, the high-frequency precipitation attenuation component, and the small-scale turbulence attenuation component, respectively.

6. The channel estimation and adaptive modulation and coding method for hybrid communication according to claim 1, characterized in that, The deep learning network model is a CNN-BIGRU-Attention network, and its prediction method is as follows: The IMF components and environmental parameters are input into the network, and spatial channel features are extracted sequentially through convolutional layers, channel evolution laws are learned by combining time-series information through bidirectional gated recurrent unit layers, key feature weights are enhanced through attention mechanisms, and finally the features are integrated through fully connected layers to output the signal attenuation value of the RF and FSO links at the next time step.

7. The channel estimation and adaptive modulation and coding method for hybrid communication according to claim 1, characterized in that, The preset coding scheme library mentioned in step 4 includes an RF modulation and coding scheme library and an FSO modulation and coding scheme library. Each scheme library contains multiple modulation and coding schemes, and each modulation and coding scheme is associated with a signal attenuation threshold range.

8. The channel estimation and adaptive modulation and coding method for hybrid communication according to claim 1, characterized in that, The adaptive modulation and coding switching includes the following process: Based on the predicted signal attenuation value, determine the signal attenuation threshold range it belongs to in the preset RF modulation and coding scheme library and FSO modulation and coding scheme library; The transition probability of switching from the current modulation and coding scheme to each candidate modulation and coding scheme is calculated using a deep learning classification network. Based on the maximum transition probability, the optimal modulation and coding scheme is dynamically selected from the coding scheme library and switched to maximize the spectrum utilization while meeting the target bit error rate.

9. A channel estimation and adaptive modulation and coding system for hybrid communication, characterized in that, include: The decomposition parameter optimization module is used to dynamically optimize the key decomposition parameters of the signal decomposition algorithm for the signal attenuation sequence of the hybrid communication link of RF and FSO at the current moment, so as to obtain the optimal decomposition parameters. The signal decomposition module is used to decompose the signal attenuation sequence into multiple intrinsic mode function (IMF) components based on the optimal decomposition parameters and the signal decomposition algorithm. The prediction module is used to input the IMF component and the real-time acquired environmental parameters into a deep learning network model, and predict the signal attenuation value at the next moment by extracting channel features and learning their temporal evolution law. The encoding switching module is used to dynamically select a modulation and coding scheme based on the predicted signal attenuation value and in combination with a preset encoding scheme library, so as to realize adaptive modulation and coding switching.

10. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the channel estimation and adaptive modulation and coding method for hybrid communication as described in any one of claims 1-8 when executing the computer program.

Images