Method and system for communication transmission control based on channel state feedback

Abstract

The application discloses a communication transmission control method and system based on channel state feedback, and the method comprises the following steps: deploying a lightweight link sensing module on the terminal side, collecting link environment sensing data, classifying the channel characteristics of the link environment sensing data, and obtaining the channel characteristic type; constructing a differentiated compression feedback strategy double channel, matching and switching the channel characteristic type, determining the target channel feedback strategy channel; collecting channel state feedback data, selecting pre-coding for the channel state feedback data, and obtaining the base station side pre-coding matrix; based on the base station side pre-coding matrix and the link environment sensing data, the communication transmission scheduling is analyzed, the target communication transmission parameter is determined, and the communication closed-loop control is carried out through the target communication transmission parameter. The application achieves the optimal balance of feedback overhead, calculation complexity and system transmission performance, and can fully adapt to the communication transmission requirements of different mobile speeds and service types.

Description

Technical Field

[0001] This invention belongs to the field of communication transmission and relates to communication transmission control technology, specifically to a communication transmission control method and system based on channel state feedback. Background Technology

[0002] With the large-scale deployment of MIMO technology in 5G and mobile communication systems, real-time channel state information feedback has become a core support for achieving efficient precoding, beamforming, and wireless resource scheduling. However, traditional full-channel state feedback grows exponentially with the increase in the number of base station antennas and users, severely crowding out uplink control channel resources. Furthermore, a single compression feedback strategy cannot adapt to differentiated channel characteristics. In high-speed, rapidly changing channel scenarios such as high-speed rail and vehicle-mounted applications, the channel tracking accuracy of fixed codebook feedback is insufficient, easily leading to precoding mismatch and a surge in transmission error rate. In static or slow-changing channel scenarios such as indoor offices and smart homes, differential quantization feedback has high redundancy, making it difficult to effectively reduce terminal power consumption and feedback overhead. At the same time, existing terminal-side link sensing modules are mostly independent hardware units with high computational and memory overhead, making them difficult to deploy in low-power IoT terminals. Moreover, transmission parameter scheduling often only optimizes throughput as a single objective, failing to take into account the latency and reliability requirements of real-time services such as voice and video, as well as the transmission efficiency requirements of non-real-time services such as file downloads. Summary of the Invention

[0003] Purpose of the invention: To address the technical problem that existing technologies cannot simultaneously adapt to the differentiated requirements of fast-changing and slow-changing channels, resulting in a difficulty in balancing uplink and downlink feedback overhead, channel tracking accuracy, and overall system transmission performance, this invention provides a communication transmission control method and system based on channel state feedback. This achieves an optimal balance between feedback overhead, computational complexity, and system transmission performance, and can fully adapt to the communication transmission requirements of different mobile speeds and service types.

[0004] Technical solution: To achieve the above objectives, the present invention provides a communication transmission control method based on channel state feedback, comprising the following steps:

[0005] S1: Deploy a lightweight link awareness module on the terminal side, collect link environment awareness data through the lightweight link awareness module, and classify the link environment awareness data by channel characteristics to obtain the channel characteristic type;

[0006] S2: Construct a dual-channel differential compression feedback strategy, and use the dual-channel differential compression feedback strategy to match and switch the channel feature type to determine the target channel feedback strategy channel;

[0007] S3: Collect channel state feedback data through the target channel feedback strategy channel, perform precoding selection on the channel state feedback data, and obtain the base station side precoding matrix;

[0008] S4: Based on the precoding matrix on the base station side and the link environment awareness data, perform communication transmission scheduling analysis, determine the target communication transmission parameters, and perform closed-loop control of communication through the target communication transmission parameters.

[0009] Furthermore, the method for classifying the channel feature types of the link environment-aware data in step S1 includes:

[0010] A1: Construct a set of link performance metrics, which includes received signal strength, signal-to-interference-plus-noise ratio, bit error rate and delay jitter, channel multipath spread and Doppler shift estimation;

[0011] A2: Perform channel feature extraction and calculation on the link environment perception data according to the link performance index set to obtain the channel performance index feature vector;

[0012] A3: Set feature judgment thresholds for the link performance index set according to the link transmission performance requirements, and determine multi-level judgment thresholds for channel types;

[0013] A4: Based on the multi-level judgment threshold of channel type, the channel performance index feature vector is classified to obtain the channel feature type.

[0014] Furthermore, the method for determining the target channel feedback strategy channel by using a differentiated compression feedback strategy to match and switch channel feature types in dual channels in step S2 includes:

[0015] B1: The specific channel architecture of the dual-channel differential compression feedback strategy includes differential quantization feedback strategy channel and base index feedback strategy channel;

[0016] The channel switcher is logically connected to each branch channel in the dual-channel differential compression feedback strategy, wherein the channel switcher adopts a MOSFET power switching circuit.

[0017] B2: Construct channel switching logic, which includes fast-changing channel activating differential quantization feedback strategy channel, slow-changing channel activating base index feedback strategy channel, and hybrid channel activating both differential quantization feedback strategy channel and base index feedback strategy channel simultaneously.

[0018] B3: The channel switcher performs channel matching and switching based on channel characteristic types according to the channel switching logic to determine the target channel feedback strategy channel.

[0019] Furthermore, the method for precoding the channel state feedback data in step S3 to obtain the base station-side precoding matrix includes:

[0020] C1: The channel state feedback data is received and decoded by the base station to obtain the channel state decoded data;

[0021] C2: Analyze the reconstruction algorithm for each logical scenario channel in the dual-channel differential compression feedback strategy to determine the scenario channel association reconstruction algorithm;

[0022] C3: The scenario channel association reconstruction algorithm is used to reconstruct the channel features of the channel state decoding data to obtain the key channel feature vector;

[0023] C4: The base station pre-stores a hierarchical precoding matrix codebook, and performs precoding selection in the hierarchical precoding matrix codebook based on the channel key feature vectors to obtain the base station-side precoding matrix.

[0024] Further, step C3 includes:

[0025] The scene channel association reconstruction algorithm specifically includes the back projection reconstruction algorithm for the differential quantization feedback strategy channel, the base index feedback strategy channel for the codebook lookup table reconstruction algorithm, and the hybrid strategy channel for the joint superposition reconstruction algorithm;

[0026] A scenario-based channel association reconstruction algorithm is used to match the channel state decoding data with the scenario algorithm to determine the target channel association reconstruction algorithm.

[0027] The current channel rank is obtained by performing channel rank determination on the channel state decoding data through the scene channel association reconstruction algorithm;

[0028] The scenario channel association reconstruction algorithm is used to reconstruct the channel features of the channel state decoding data according to the current channel rank, and obtain the key channel feature vector.

[0029] Furthermore, the method for determining the target communication transmission parameters in step S4 based on the base station-side precoding matrix and link environment-aware data through communication transmission scheduling analysis includes:

[0030] D1: Obtain the communication transmission parameter dimension, and perform communication transmission scheduling analysis based on the base station-side precoding matrix and link environment awareness data according to the communication transmission parameter dimension to obtain the communication transmission parameter selection space;

[0031] D2: Construct the link communication transmission reward function based on the link communication transmission target;

[0032] D3: Use the link communication transmission reward function to optimize the communication transmission parameter selection space and determine the target communication transmission parameters.

[0033] Further, step D3 includes:

[0034] E1: Randomly select N communication transmission parameters within the communication transmission parameter selection space;

[0035] E2: The link communication transmission reward function is used to evaluate and optimize N communication transmission parameters to obtain the frontier communication transmission parameter set;

[0036] E3: Based on the frontier communication transmission parameter set, the link communication transmission reward function is solved and optimized to determine the target communication transmission parameters.

[0037] Further, step E3 includes:

[0038] F1: The frontier communication transmission parameter set is used as the initial parameter solution and cross-mutation is performed to expand it, so as to obtain the initial communication transmission parameter population.

[0039] F2: Use the link communication transmission reward function to iteratively evaluate and update the initial communication transmission parameter population until the preset termination condition is met, and obtain the communication transmission parameter population space.

[0040] F3: Based on the population space of communication transmission parameters, the link communication transmission reward function is solved and optimized to determine the target communication transmission parameters.

[0041] Further, step F3 includes:

[0042] The first communication transmission parameter is obtained by globally maximizing the link communication transmission reward function based on the population space of the communication transmission parameter.

[0043] The target communication transmission parameters are determined by growing and optimizing neighboring individuals based on the first communication transmission parameter.

[0044] The present invention also provides a communication transmission control system based on channel state feedback, comprising:

[0045] The classification module is used to deploy a lightweight link awareness module on the terminal side. The lightweight link awareness module collects link environment awareness data, classifies the link environment awareness data by channel characteristics, and obtains the channel feature type.

[0046] The matching module is used to construct a dual-channel differential compression feedback strategy. It uses the dual-channel differential compression feedback strategy to match and switch the channel feature type to determine the target channel feedback strategy channel.

[0047] The encoding module is used to collect channel state feedback data through the target channel feedback strategy channel, perform precoding selection on the channel state feedback data, and obtain the base station side precoding matrix.

[0048] The parsing module is used to perform communication transmission scheduling parsing based on the base station-side precoding matrix and link environment awareness data, determine the target communication transmission parameters, and perform communication closed-loop control based on the target communication transmission parameters.

[0049] Beneficial Effects: Compared with existing technologies, this invention deploys a lightweight link sensing module on the terminal side that reuses the idle computing resources of existing RF front-end and baseband chip. It constructs a six-dimensional link performance index set including received signal strength, signal-to-interference-plus-noise ratio, bit error rate, latency jitter, channel multipath spread, and Doppler shift estimation to achieve accurate composite classification of channel characteristics. Furthermore, it designs a differentiated compression feedback strategy combining differential quantization and base indexing for dual-channel operation. Based on the channel change rate attribute, it dynamically switches or jointly activates the feedback channel, and uses corresponding back-projection, codebook lookup, and joint superposition reconstruction algorithms to achieve efficient and low-complexity reconstruction of channel characteristics. Simultaneously, it constructs a multi-objective weighted link communication transmission reward function, combining global optimization using a genetic algorithm with fine-grained neighborhood search to achieve optimal scheduling of transmission parameters. Ultimately, it achieves an optimal balance between feedback overhead, computational complexity, and system transmission performance, and can fully adapt to the communication transmission needs of different mobile speeds and service types. Attached Figure Description

[0050] Figure 1 This is a schematic flowchart of the method of the present invention;

[0051] Figure 2 This is a schematic diagram of the system of the present invention.

[0052] Explanation of the attached diagram labels: Classification module 11, Matching module 12, Encoding module 13, Parsing module 14. Detailed Implementation

[0053] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading this invention, any modifications of the invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.

[0054] Example 1:

[0055] like Figure 1 As shown, this embodiment provides a communication transmission control method based on channel state feedback, including the following steps:

[0056] S100: Deploy a lightweight link awareness module on the terminal side, collect link environment awareness data through the lightweight link awareness module, classify the link environment awareness data by channel characteristics, and obtain the channel characteristic type.

[0057] Specifically, a lightweight link sensing module is first deployed on the terminal side. This module is not an independent hardware unit, but a dedicated software sensing unit formed by reusing the idle computing resources of the terminal's existing RF front-end and baseband chip, and by modifying the traditional complex channel sensing algorithm through operator pruning, computation graph optimization, and fixed-point transformation. Its peak computing overhead is strictly controlled within 5% of the terminal's total baseband computing power, and its memory usage does not exceed 128KB. Subsequently, the lightweight link sensing module collects link environment sensing data at an adaptive sampling frequency. The link environment sensing data is pre-processed physical layer and link layer structured data, specifically including channel impulse response segments extracted from the terminal's baseband channel estimator, the received signal power moving average sequence output by the RF receiver, and data packet transmission error records and delay difference records statistically collected by the MAC layer. The sampling frequency is dynamically adjusted according to the speed of the initially detected channel changes, increasing to once per millisecond in fast-changing channels to ensure real-time sensing, and decreasing to once per hundred milliseconds in slow-changing channels to further reduce terminal power consumption.

[0058] Next, we construct a set of link performance metrics. This set of metrics is a collection of quantitative indicators that has been selected through theoretical derivation and simulation verification, and can comprehensively and without redundancy characterize the channel transmission characteristics. Specifically, it includes six core indicators: received signal strength, which characterizes the overall received power of the signal and reflects the basic coverage quality of the link; signal-to-interference-plus-noise ratio (SINR), which is the ratio of useful signal power to interference signal plus noise power, directly determining the Shannon limit that the link can achieve, i.e., the theoretical maximum transmission rate; bit error rate, which is the ratio of the number of erroneous bits to the total number of bits during transmission, reflecting the actual transmission reliability of the link; and delay jitter, which is the amount of variation in the arrival time of consecutive data packets, and is crucial to the user experience of real-time services such as voice and video. Multipath propagation, the maximum time difference between signals arriving at the receiver via different reflection paths in the channel, determines the frequency-selective fading characteristics and Doppler shift estimation of the channel. It is generated by the relative motion between the terminal and the base station and directly reflects the time-selective fading characteristics of the channel, i.e., the rate of channel change. According to this link performance index set, feature extraction and quantization calculation are performed on the collected link environment perception data for each index. The unstructured data that was originally scattered in different protocol layers and different dimensions is converted into a channel performance index feature vector in a unified format. This vector is a one-dimensional numerical vector formed by arranging the above six quantization indices in a fixed order, and each element corresponds to the real-time calculated value of a link performance index.

[0059] Subsequently, feature judgment thresholds are set for each indicator in the link performance indicator set according to the link transmission performance requirements of different services, forming a multi-level judgment threshold for channel type. The multi-level judgment threshold is a graded threshold matrix constructed for each indicator. In this embodiment, for the Doppler frequency shift indicator, 0-10Hz is set as the slow-changing channel threshold range, 10-100Hz as the mixed channel threshold range, and greater than 100Hz as the fast-changing channel threshold range. For the signal-to-interference-plus-noise ratio indicator, greater than 20dB is set as the high-quality channel threshold, 10-20dB as the medium-quality channel threshold, and less than 10dB as the low-quality channel threshold. The multi-level thresholds of all indicators together constitute a six-dimensional channel judgment space, and the threshold parameters can be dynamically adjusted according to the differentiated requirements of different services such as voice, video, and data.

[0060] Finally, based on the multi-level judgment threshold of the channel type, the channel performance index feature vector is classified in multiple dimensions to obtain the final channel feature type. This type is a composite label that simultaneously includes the channel change rate attribute, slow / mixed / fast change and channel quality attribute, high / medium / low. For example, slow-changing high-quality channel, fast-changing medium-quality channel, etc., comprehensively and accurately describe the actual state of the current channel.

[0061] S200: Construct a dual-channel differential compression feedback strategy, and use the dual-channel differential compression feedback strategy to match and switch the channel feature type to determine the target channel feedback strategy channel.

[0062] Specifically, a dual-channel differential compression feedback strategy is first constructed. This dual-channel approach shares existing uplink control channel resources, operates logically in parallel, and employs two completely different compression principles: a differential quantization feedback strategy channel and a base-indexed feedback strategy channel. The differential quantization feedback strategy channel utilizes the time correlation of the channel, quantizing and encoding only the difference between the current and previous channel states. Differential quantization refers to the technique of discretizing continuous channel difference signals into a finite number of quantization levels and encoding them. Its quantization step size adaptively adjusts according to the channel change rate; the faster the channel changes, the smaller the quantization step size and the higher the accuracy. The system can accurately capture the rapid fluctuation characteristics of the channel with low feedback overhead, making it particularly suitable for fast-changing channels in high-speed mobile scenarios. The base index feedback strategy channel is based on pre-shared codebook technology. The base station and terminal pre-store identical hierarchical precoding matrix codebooks. This codebook is composed of a large number of orthogonal channel basis vectors organized hierarchically according to channel rank and accuracy level. The terminal only needs to feed back the index number of the basis vector that best matches the current channel in the codebook. The base index is the unique numerical identifier corresponding to each precoding vector in the codebook, which usually only requires a few bits. This reduces the feedback overhead by more than an order of magnitude, making it particularly suitable for slow-changing channels in stationary or low-speed mobile scenarios. Next, the channel switcher is logically connected to each branch channel in the differentiated compression feedback strategy dual-channel system. This channel switcher is implemented using a MOSFET power switching circuit. MOSFETs, or metal-oxide-semiconductor field-effect transistors, have advantages such as nanosecond-level switching speed, milliohm-level on-resistance, and microwatt-level static power consumption. They can achieve seamless millisecond-level channel switching without introducing additional signal delay and power loss. Furthermore, the terminal power consumption can be further reduced by shutting down the power supply of unused channels.

[0063] Subsequently, a channel switching logic is constructed, which strictly maps one-to-one with the channel change rate attribute in the output channel feature type. Specifically: when the channel feature type is a fast-changing channel, i.e., the Doppler frequency shift is greater than 100Hz, the differential quantization feedback strategy channel is activated and the base index feedback strategy channel is turned off to prioritize the real-time performance and tracking accuracy of the feedback; when the channel feature type is a slow-changing channel, i.e., the Doppler frequency shift is less than 10Hz, the base index feedback strategy channel is activated and the differential quantization feedback strategy channel is turned off to minimize uplink feedback overhead; when the channel feature type is a mixed channel, i.e., the Doppler frequency shift is between 10-100Hz, both channels are activated simultaneously and share uplink control channel resources using time-division multiplexing. In this case, the differential quantization feedback strategy channel is responsible for feeding back the fast fluctuation component of the channel, and the base index feedback strategy channel is responsible for feeding back the stable base component of the channel. The two are combined at the base station side through a joint superposition reconstruction algorithm to obtain complete channel state information.

[0064] Finally, the channel switcher matches and determines the input channel feature type according to the above switching logic, while introducing a 10% threshold hysteresis margin to avoid ping-pong switching problems caused by frequent fluctuations of the channel near the determination threshold, and finally determines the unique target channel feedback strategy channel or dual-channel joint working mode.

[0065] S300: Collect channel state feedback data through the target channel feedback strategy channel, perform precoding selection on the channel state feedback data, and obtain the base station side precoding matrix.

[0066] Specifically, the base station first receives channel state feedback data from the target channel feedback strategy channel via the uplink control channel. The format and content of this data are entirely determined by the target channel type: if the target channel is a differential quantization feedback strategy channel, the feedback data is a sequence of differences between the current and previous channel states after adaptive quantization coding; if the target channel is a base-indexed feedback strategy channel, the feedback data is a fixed-length index frame containing channel rank information and the optimal base vector number; if the target channel is a dual-channel joint working mode, the feedback data is a mixed frame of time-division multiplexed index frames and difference sequences. The base station first performs CRC check and channel decoding on the received data to correct bit errors introduced during transmission and obtain error-free channel state decoded data.

[0067] Next, we perform reconstruction algorithm mapping analysis on each logical scenario channel in the dual-channel differential compression feedback strategy to determine the scenario channel associated reconstruction algorithm. This algorithm is a set of inverse transformation algorithms that correspond one-to-one with the compression principle of each channel, specifically including three categories: the back projection reconstruction algorithm for the differential quantization feedback strategy channel, which utilizes the temporal correlation of the channel to project the dequantized channel difference onto the channel state space reconstructed at the previous moment, and obtains the complete channel matrix at the current moment through iterative updates, which can effectively track the instantaneous fluctuations of fast-changing channels; the codebook lookup table reconstruction algorithm for the base index feedback strategy channel, which utilizes the hierarchical precoding matrix codebook pre-shared by the base station and the terminal, and directly retrieves the corresponding channel basis vector from the codebook according to the base index obtained by decoding, without complex matrix operations, and has extremely low computational complexity; and the joint superposition reconstruction algorithm for the hybrid strategy channel, which linearly superimposes the channel stable basis component reconstructed from the base index channel and the channel fast fluctuation component reconstructed from the differential quantization channel, and corrects the superposition result through the minimum mean square error algorithm, balancing reconstruction accuracy and computational efficiency.

[0068] Subsequently, the matched scene channel association reconstruction algorithm is used to determine the channel rank of the channel state decoding data. The channel rank refers to the maximum number of independent parallel data streams that a MIMO channel can support. It is determined by the spatial correlation and signal-to-noise ratio of the channel and directly determines the spatial multiplexing gain of the system. The determination process is to perform eigenvalue decomposition on the channel covariance matrix obtained by reconstruction and count the number of eigenvalues ​​greater than a preset threshold to obtain the current channel rank. Then, the channel state decoding data is reconstructed by dimensionality reduction according to the channel rank to extract the key channel feature vector that can characterize the main transmission characteristics of the channel. This vector eliminates noise components and secondary features in the channel and retains only the core information that has the greatest impact on transmission performance, which greatly reduces the computational complexity of subsequent precoding selection.

[0069] Finally, a hierarchical precoding matrix codebook identical to that of the terminal is pre-stored on the base station side. This codebook is a set of precoding matrices pre-designed according to the 3GPP standard and actual channel statistical characteristics, organized hierarchically according to channel rank and precision level. Each precoding matrix corresponds to a set of optimal transmit beamforming weights, which can concentrate the transmit energy in the direction of the terminal while suppressing interference to other users. Based on the channel key feature vector obtained by reconstruction, the inner product value of it and all precoding matrices of the same rank in the codebook is calculated. The larger the inner product value, the higher the matching degree of the precoding matrix with the current channel. Finally, the precoding matrix with the largest inner product value is selected as the precoding matrix on the base station side, and the index of the matrix is ​​stored for differential feedback reference in the next time step.

[0070] S400: Based on the precoding matrix and link environment awareness data on the base station side, communication transmission scheduling is analyzed to determine the target communication transmission parameters, and communication closed-loop control is performed through the target communication transmission parameters.

[0071] Specifically, the communication transmission parameter dimension is first obtained. This dimension refers to the set of physical layer and link layer parameters that can directly affect the transmission performance of the communication system and can be dynamically adjusted. In this embodiment, it specifically includes modulation and coding scheme, transport block size, downlink transmit power, number of physical resource blocks allocated, number of hybrid automatic repeat request processes, and number of spatial multiplexing layers. The number of spatial multiplexing layers is directly determined by the rank of the base station-side precoding matrix obtained in the previous step. The other parameters are strongly correlated with indicators such as signal-to-interference-plus-noise ratio, bit error rate, and latency jitter in the link environment perception data. Based on the rank constraint of the base station-side precoding matrix and the performance threshold constraint of the link environment perception data, the full set of parameter combinations is pruned and filtered to remove all invalid parameter combinations that do not meet the current channel conditions and service requirements. This constructs a communication transmission parameter selection space, which is a set of low-dimensional feasible parameters after constraints. Compared with the full parameter space, its size can be reduced by more than 90%, which greatly reduces the computational complexity of subsequent optimization. Next, based on the differentiated transmission objectives of the current carried services, a link communication transmission reward function is constructed. This function is not a single throughput indicator function, but a multi-objective weighted comprehensive performance evaluation function, and its general expression is:

[0072] Reward = α × Normalized throughput - β × Normalized delay - γ × Normalized bit error rate;

[0073] α, β, and γ are dynamic weighting coefficients that can be adjusted in real time according to the service type: For real-time services such as voice and video conferencing, the latency weight β and the bit error rate weight γ are set to 0.4 and 0.3 respectively, and the throughput weight α is set to 0.3, prioritizing the real-time performance and reliability of transmission; For non-real-time services such as file download and video caching, the throughput weight α is set to 0.7, the latency weight β is set to 0.1, and the bit error rate weight γ is set to 0.2, prioritizing maximizing system transmission efficiency.

[0074] Subsequently, the reward function is used to perform hierarchical optimization of the communication transmission parameter selection space: First, N initial communication transmission parameter combinations (N = 50) are randomly selected from the parameter selection space, and their reward values ​​are calculated one by one. The top 10 parameter combinations with the highest reward values ​​are selected to form the frontier communication transmission parameter set, which represents the locally optimal solution in the current parameter space. The frontier communication transmission parameter set is used as the initial parameter solution for crossover and mutation expansion. The crossover operation swaps the corresponding parameter values ​​of two frontier parameter combinations with a probability of 0.7 to generate a new parameter combination that inherits the superior characteristics of its parent. The mutation operation randomly changes the value of a parameter with a probability of 0.1 to introduce a new solution space to avoid getting trapped in local optima. The final generated solution has a scale of 20. An initial population of communication transmission parameters with a value of 0 is established. The initial population is iteratively evaluated and updated using the link communication transmission reward function. In each generation, the 50 individuals with the highest reward are retained as parents, and crossover and mutation are continued to generate the next generation until the preset maximum number of iterations is reached, denoted as 20 iterations or the reward value converges, i.e., the reward value increases by less than 0.1% for three consecutive generations, resulting in a converged communication transmission parameter population space. Based on this population space, the reward function is globally maximized to obtain the globally optimal first communication transmission parameter. Then, neighborhood individuals are grown around this parameter, i.e., small parameter combinations with a step size of ±1 are generated in each parameter dimension for fine optimization, and finally, the target communication transmission parameter with the best overall performance is determined.

[0075] Finally, the communication closed-loop control is executed. The base station configures the transmitter according to the target communication transmission parameters, and transmits data in combination with the determined base station-side precoding matrix. At the same time, it monitors the performance indicators such as throughput, bit error rate, and latency in real time during the actual transmission process. These measured performance indicators are fed back to the lightweight link awareness module on the terminal side through the downlink control channel. As an important part of the next round of link environment awareness data, it is used to update the channel feature classification results and switch the differentiated feedback strategy, forming a control loop of perception-feedback-precoding-scheduling-re-perception.

[0076] In this embodiment, channel feature classification is performed on the link environment-aware data to obtain channel feature types. This includes: constructing a link performance index set, which includes received signal strength, signal-to-interference-plus-noise ratio, bit error rate and delay jitter, channel multipath spread, and Doppler shift estimation; extracting and calculating channel features from the link environment-aware data according to the link performance index set to obtain channel performance index feature vectors; setting feature judgment thresholds for the link performance index set according to link transmission performance requirements to determine multi-level channel type judgment thresholds; and classifying the channel performance index feature vectors based on the multi-level channel type judgment thresholds to obtain channel feature types.

[0077] Specifically, a set of link performance metrics is first constructed. This set consists of six quantitative metrics selected after channel theory derivation and extensive simulation verification in real-world scenarios. These metrics comprehensively and without redundancy characterize the channel transmission characteristics. Each metric corresponds to a key dimension of the channel: Received Signal Strength (RSS), which characterizes the average power of the base station's transmitted signal received by the terminal, reflecting the basic coverage quality of the link and serving as a prerequisite for normal communication; Signal-to-Interference-Ratio (SINR), the ratio of useful signal power to the power of co-channel interference signal plus background noise, directly determining the Shannon limit that the link can achieve, i.e., the theoretical maximum transmission rate; and Bit Error Rate (BER), the ratio of the number of incorrectly received bits to the total number of transmitted bits, reflecting the actual transmission rate of the link. Transmission reliability is the most intuitive indicator of communication quality. Delay jitter, the standard deviation of the time difference between two consecutive data packets arriving at the terminal, has a decisive impact on the user experience of real-time services such as voice and video conferencing. Channel multipath spread, the maximum time difference between signals arriving at the receiver through different reflection and scattering paths in the channel, determines the frequency-selective fading characteristics of the channel, that is, signals of different frequencies will experience different attenuation during transmission. Doppler shift estimation, the frequency offset of the received signal caused by the relative motion between the terminal and the base station, directly reflects the time-selective fading characteristics of the channel, that is, how fast the channel state changes over time. These six indicators complement each other and together constitute a complete description of the channel characteristics.

[0078] Next, feature extraction and quantification calculations are performed on the collected link environment perception data according to the link performance index set to obtain the channel performance index feature vector: For received signal strength, it is calculated by averaging the instantaneous power sampled value output by the RF receiver using a sliding window, with the window size adaptively adjusted according to the channel change rate; for signal-to-interference-plus-noise ratio (SINR), it is obtained by dividing the useful signal power output by the baseband channel estimator by the interference noise power; for bit error rate (BER), it is obtained by statistically analyzing the ratio of the number of CRC check error bits to the total number of transmitted bits in a statistical period at the MAC layer; for delay jitter, it is obtained by calculating the standard deviation of the arrival time interval of 100 consecutive data packets; for channel multipath spread, it is obtained by performing energy detection on the channel impulse response and statistically analyzing the maximum time difference of all paths with energy exceeding 10% of the peak energy; for Doppler shift estimation, it is obtained by calculating the phase change rate of the channel estimation results of two adjacent time slots. The above six quantification indicators are arranged in a fixed order to form a six-dimensional numerical vector. This vector realizes the mathematical and structured expression of the unstructured perception data that was originally scattered in different protocol layers and different dimensions, providing a computational basis for subsequent classification and judgment. Then, based on the link transmission performance requirements of the current carrying services, feature judgment thresholds are set for each indicator in the link performance indicator set to determine multi-level judgment thresholds for channel types. These multi-level judgment thresholds are hierarchical threshold matrices constructed for each indicator, enabling fine-grained classification of channel characteristics. For example, for the Doppler frequency shift indicator, 0-10Hz is set as the slow-changing channel threshold range, 10-100Hz as the mixed channel threshold range, and greater than 100Hz as the fast-changing channel threshold range. For the signal-to-interference-plus-noise ratio (SINR) indicator, greater than 20dB is set as the high-quality channel threshold, and 10-20dB as the medium-quality channel threshold. A threshold of less than 10dB is defined as a low-quality channel threshold. A 10% hysteresis margin is set for each threshold; that is, the channel only switches to the next level when the index value rises from below the threshold to 110% of the threshold, and only switches to the next level when it falls from above the threshold to 90% of the threshold. This effectively avoids the ping-pong switching problem caused by frequent channel fluctuations near the decision threshold. Furthermore, the threshold parameters are not fixed but can be dynamically adjusted according to the differentiated needs of different services such as voice, video, and data. For example, for latency-sensitive voice services, the Doppler shift threshold can be appropriately increased to prioritize real-time feedback.

[0079] Finally, based on the multi-level judgment threshold of the channel type, the channel performance index feature vector is classified in multiple dimensions to obtain the final channel feature type. This classification process adopts a weighted voting method for multi-dimensional comprehensive judgment: First, the channel change rate attribute is determined based on the judgment result of the Doppler frequency shift index, which is slow / mixed / fast, and the weight of this attribute is set to 0.5 because it directly determines the selection of subsequent feedback strategies; then, the channel quality attribute is determined based on the judgment results of the three indicators of signal-to-interference-plus-noise ratio, bit error rate, and received signal strength, which is high / medium / low, and the weights of these three indicators are set to 0.2, 0.15, and 0.15, respectively; finally, a composite label that simultaneously contains the channel change rate attribute and the channel quality attribute is obtained, such as slow-changing high-quality channel, fast-changing medium-quality channel, mixed low-quality channel, etc. The composite label can more comprehensively and accurately describe the actual state of the current channel.

[0080] In this embodiment, a differentiated compression feedback strategy dual-channel is used to match and switch channel feature types to determine the target channel feedback strategy channel. This includes: the specific channel architecture of the differentiated compression feedback strategy dual-channel includes a differential quantization feedback strategy channel and a base-index feedback strategy channel; logically connecting a channel switcher to each branch channel in the differentiated compression feedback strategy dual-channel, wherein the channel switcher uses a MOSFET power switching circuit; constructing channel switching logic, specifically, a fast-changing channel activates the differential quantization feedback strategy channel, a slow-changing channel activates the base-index feedback strategy channel, and a mixed channel simultaneously activates both the differential quantization feedback strategy channel and the base-index feedback strategy channel; and the channel switcher performs channel matching and switching according to the channel switching logic to determine the target channel feedback strategy channel.

[0081] Specifically, the architecture of the dual-channel differential compression feedback strategy is first clarified. This dual-channel architecture consists of two feedback processing paths that share the existing uplink control channel spectrum resources of the terminal, operate logically in parallel, and employ completely different compression principles: a differential quantization feedback strategy channel and a base-indexed feedback strategy channel. The differential quantization feedback strategy channel utilizes the temporal correlation of the channel to design a compression algorithm. Differential quantization refers to a technique that does not transmit the complete channel state matrix but only quantizes and encodes the difference between the current and previous channel states. Because the channel states at adjacent times have a strong correlation, the dynamic range of the difference is much smaller than the original channel state. Therefore, high-precision quantization can be achieved with fewer bits, and its quantization step size will adjust according to the channel change rate. Adaptive adjustment: the faster the channel changes, the smaller the quantization step size and the higher the tracking accuracy. It can accurately capture the instantaneous fluctuation characteristics of the channel in high-speed moving scenarios, and is particularly suitable for fast-changing channels with Doppler shift greater than 100Hz. The basis index feedback strategy channel is implemented based on the hierarchical precoding matrix codebook pre-shared by the base station and the terminal. The basis index refers to the unique digital identifier corresponding to each orthogonal channel basis vector in the codebook. It usually only requires 4-8 bits to represent. The terminal does not need to transmit the complete channel vector. It only needs to feed back the index number of the basis vector that best matches the current channel. The base station can directly restore the channel state by looking up the table, which can reduce the feedback overhead by more than an order of magnitude. It is particularly suitable for stationary or slow-changing channels with Doppler shift less than 10Hz.

[0082] Next, the channel switcher is logically and physically connected to each branch channel in the differential compression feedback strategy dual channel. The channel switcher is implemented using a MOSFET power switching circuit integrated in the power management module of the terminal baseband chip. MOSFET, or metal-oxide-semiconductor field-effect transistor, has the core advantages of nanosecond-level switching speed, milliohm-level on-resistance, and microwatt-level static power consumption. It can not only achieve seamless millisecond-level switching of channels without introducing additional signal delay and power loss, but also further reduce the standby and operating power consumption of the terminal by completely shutting down the power supply of unused channels. The connection method is to connect two MOSFET switches in series on the power input and data output paths of the differential quantization feedback channel and the base index feedback channel, respectively, to achieve independent on / off control of the two channels.

[0083] Subsequently, a channel switching logic is constructed that strictly maps to the channel feature type output in the previous step. This logic uses the channel change rate attribute in the channel feature type as the core judgment criterion, determined by the Doppler frequency shift index, and is fine-tuned with reference to channel quality attributes. Specifically: when the channel feature type is a fast-changing channel, i.e., the Doppler frequency shift is greater than 100Hz, the differential quantization feedback strategy channel is activated and the base index feedback strategy channel is completely turned off to prioritize ensuring the real-time tracking accuracy of the feedback; when the channel feature type is a slow-changing channel, i.e., the Doppler frequency shift is less than 10Hz, the base index feedback strategy channel is activated and differential quantization is completely turned off. The feedback strategy channel minimizes uplink feedback overhead. When the channel characteristic type is a hybrid channel, i.e., the Doppler frequency shift is between 10-100Hz, two channels are activated simultaneously and share uplink control channel resources using dynamic time-division multiplexing. In this case, the differential quantization feedback strategy channel is responsible for the fast fluctuation component of the feedback channel, and the base index feedback strategy channel is responsible for the stable base component of the feedback channel. The resource ratio of the two channels is dynamically adjusted according to the specific value of the Doppler frequency shift. The closer the Doppler frequency shift is to 100Hz, the higher the resource ratio of the differential quantization channel, and vice versa.

[0084] Finally, the channel switcher matches and determines the input channel feature type according to the above switching logic. At the same time, it sets a 10% hysteresis margin for each determination threshold. That is, it switches to the next channel only when the Doppler frequency shift rises from below the threshold to 110% of the threshold, and switches to the next channel only when it falls from above the threshold to 90% of the threshold. This effectively avoids the ping-pong handover problem caused by frequent fluctuations of the channel near the determination threshold. That is, the channel repeatedly switches between two states, introducing additional handover overhead and performance degradation. Finally, it determines a unique target channel feedback strategy channel or dual-channel joint working mode. During the handover process, the feedback data of the previous moment is retained as a transition to ensure that there is no gap in channel state information on the base station side.

[0085] In this embodiment, precoding selection is performed on the channel state feedback data to obtain the base station-side precoding matrix. This includes: receiving and decoding the channel state feedback data through the base station to obtain channel state decoded data; performing reconstruction algorithm analysis on each logical scenario channel in the dual-channel differential compression feedback strategy to determine the scenario channel association reconstruction algorithm; using the scenario channel association reconstruction algorithm to reconstruct the channel features of the channel state decoded data to obtain the channel key feature vector; and performing precoding selection in the hierarchical precoding matrix codebook pre-stored in the base station based on the channel key feature vector to obtain the base station-side precoding matrix.

[0086] Specifically, the base station first receives channel state feedback data from the target channel feedback strategy channel determined in the previous step via the uplink control channel, i.e., the physical uplink control channel / physical uplink shared channel, which is a dedicated wireless channel for terminals to transmit control information and a small amount of data to the base station. The frame structure, length, and encoding method of this data are completely determined by the target channel type and are strictly synchronized with the terminal side: if the target channel is a differential quantization feedback strategy channel, the feedback data is a sequence of differences between the current and previous channel states after adaptive step-size quantization encoding, the frame header contains a 3-bit quantization step-size index and 2-bit synchronization information, and the frame length is dynamically adjusted according to the channel rank; if the target channel is a base-index feedback strategy channel, the feedback data is a packet The fixed-length index frame, containing a 2-bit channel rank identifier and a 6-bit optimal basis vector number, has a total length of only 8 bits. If the target channel is in dual-channel joint operation mode, the feedback data is a time-division multiplexed hybrid frame divided by dynamic time slots. The first 60% of the time slots transmit the basis index frame, and the last 40% of the time slots transmit the differential quantization difference sequence. The time slot ratio is adjusted in real time with the Doppler frequency shift. The base station first performs cyclic redundancy check on the received data, that is, it realizes transmission error detection and low-density parity check by adding a check code to the end of the data frame. This is a channel coding and decoding with extremely strong error correction capability. Invalid data frames that fail the check are eliminated, and random bit errors and burst bit errors introduced during transmission are corrected to obtain error-free channel state decoded data.

[0087] After acquiring the channel state decoding data, the base station performs reconstruction algorithm analysis. This analysis is not a simple lookup table, but a deep matching analysis based on the compression principle of the dual-channel differential compression feedback strategy, the time-varying characteristics of the channel, and the computing power constraints of the base station. Ultimately, it determines the scenario-channel associated reconstruction algorithm. This algorithm is a set of dedicated inverse transform algorithms that strictly correspond one-to-one with the compression principle of each channel, specifically including three types: The back-projection reconstruction algorithm for differential quantization feedback strategy channels. This algorithm specifically utilizes the strong temporal correlation between adjacent time slots of the fast-changing channel, meaning that the channel state changes little between adjacent moments and has strong continuity. Instead of directly reconstructing the complete channel matrix from scratch, it projects the dequantized channel difference vector onto the channel state space that has been reconstructed at the previous moment and verified as valid through actual transmission. It quickly converges to the current channel state through a maximum of three weighted iterative updates, with a computational load only that of traditional full-channel reconstruction. The reconstruction algorithm achieves 30% efficiency, enabling millisecond-level tracking of instantaneous channel fluctuations in high-speed mobile scenarios. A codebook lookup reconstruction algorithm for the base index feedback strategy channel is implemented based on a hierarchical precoding matrix codebook pre-synchronized by the base station and terminal. It requires no complex matrix operations, only address mapping in the codebook based on the decoded base index to directly retrieve the corresponding channel basis vector, resulting in near-zero computational complexity. This algorithm is particularly well-suited for low-overhead requirements in slowly varying channels. A joint superposition reconstruction algorithm for the hybrid strategy channel combines the advantages of the previous two algorithms. It first reconstructs the slowly varying stable basis components of the channel through codebook lookup, then reconstructs the fast-changing fluctuation components through back-projection, and finally uses the optimal estimation algorithm, which minimizes the mean square error between the estimated and true values, to perform noise correction on the linear superposition result of the two components, balancing reconstruction accuracy and computational efficiency.

[0088] Next, the channel feature reconstruction process is executed. The base station uses the matched target channel association reconstruction algorithm to process the channel state decoding data. First, the channel rank is determined. The channel rank refers to the maximum number of independent parallel data streams that a multi-input multi-output channel can support. Multi-output refers to parallel data transmission achieved by using multiple antennas at both the transmitting and receiving ends. It is determined by the spatial correlation and signal-to-interference-plus-noise ratio of the channel, and directly determines the spatial multiplexing gain that the system can obtain, that is, the ability to improve the system throughput by transmitting multiple data streams in parallel. The rank determination logic of different algorithms is slightly different: the back-projection reconstruction algorithm first synthesizes the dequantized channel difference with the channel matrix of the previous time step to obtain the initial channel matrix of the current time step, and then performs eigenvalue decomposition on its covariance matrix, that is, the process of decomposing the matrix into eigenvalues ​​and eigenvectors. The magnitude of the eigenvalue corresponds to the amount of energy carried by the eigenvector. The number of eigenvalues ​​that are greater than a preset noise threshold, usually 1% of the largest eigenvalue, is taken as the current channel rank; codebook lookup The table reconstruction algorithm directly determines the current channel rank based on the 2-bit channel rank identifier field in the decoded index frame, without additional calculation. The joint superposition reconstruction algorithm performs eigenvalue decomposition on the initial channel matrix after superposition correction to obtain the channel rank, and introduces a moving average filter with one time slot to avoid frequent changes in the channel rank. After the channel rank is determined, the base station performs dimensionality reduction reconstruction on the channel state decoded data according to the current channel rank to obtain the key channel feature vectors: the back projection reconstruction algorithm extracts the top K principal feature vectors corresponding to the current channel rank from the synthesized complete channel matrix, and removes secondary feature components and noise components with an energy share of less than 1%; the codebook lookup table reconstruction algorithm directly extracts the K orthogonal basis vectors corresponding to the rank from the codebook as the key channel feature vectors; the joint superposition reconstruction algorithm extracts the top K principal feature vectors from the superposition corrected channel matrix. The final key channel feature vector is a low-dimensional vector with the same dimension as the channel rank, retaining only the channel information that has the greatest impact on transmission performance.

[0089] Finally, a hierarchical precoding matrix codebook identical to that of the terminal is pre-stored on the base station side. This codebook is a set of precoding matrices pre-designed according to the 3GPPR17 standard and actual channel statistical characteristics, organized hierarchically according to channel rank and precision level. It is divided into 4 rank levels and 3 precision levels. Each precoding matrix corresponds to a set of optimal transmit beamforming weights, which can concentrate the transmit energy in the spatial direction where the terminal is located, while suppressing interference to other users in the same cell to the greatest extent. Based on the channel key feature vector obtained by reconstruction, the inner product value of it and all precoding matrices of the same rank in the codebook is calculated. The larger the inner product value, the higher the matching degree of the precoding matrix with the current channel, and the greater the array gain that can be obtained. That is, the greater the ability to improve the received signal strength through multi-antenna beamforming. Finally, the precoding matrix with the largest inner product value is selected as the precoding matrix on the base station side, and the index of the matrix is ​​stored in the feedback buffer of the base station as a reference benchmark for differential quantization feedback at the next time step.

[0090] In this embodiment, a scenario-channel association reconstruction algorithm is used to reconstruct channel features from channel state decoded data to obtain key channel feature vectors. This includes: the scenario-channel association reconstruction algorithm specifically includes a back-projection reconstruction algorithm for differential quantization feedback strategy channels, a base index feedback strategy channel for codebook lookup table reconstruction algorithms, and a hybrid strategy channel for joint superposition reconstruction algorithms; the scenario-channel association reconstruction algorithm is used to match scenario algorithms with the channel state decoded data to determine the target channel association reconstruction algorithm; the scenario-channel association reconstruction algorithm is used to perform channel rank determination on the channel state decoded data to obtain the current channel rank; and the scenario-channel association reconstruction algorithm is used to reconstruct channel features from the channel state decoded data according to the current channel rank to obtain key channel feature vectors.

[0091] Specifically, the scenario channel correlation reconstruction algorithm is first defined. This algorithm is designed for the three working modes of the dual-channel differential compression feedback strategy, specifically including three types: The back-projection reconstruction algorithm for the differential quantization feedback strategy channel. This algorithm specifically utilizes the temporal correlation of rapidly changing channels, i.e., the channel state change amplitude between adjacent time slots in high-speed mobile scenarios is less than 10%, exhibiting strong continuity design. Its specific process is as follows: First, the quantization difference in the channel state decoding data is dequantized to obtain the channel state difference vector between the current and previous time moments. Then, this difference vector is added element-wise to the complete channel matrix stored in the base station feedback buffer at the previous time moment to obtain the initial channel matrix at the current time moment. Next, the initial channel matrix is ​​projected onto the channel feature space that has been verified and valid at the previous time moment. Noise interference is eliminated through weighted iterative updates. During the iteration process, the weight coefficients are adaptively adjusted according to the channel change rate; the faster the channel change, the more... The higher the weight of the current difference, the more it can converge to the true channel state at the current moment after a maximum of 3 iterations. The computational cost is only 28% of that of the traditional least squares reconstruction algorithm. It can track the instantaneous channel fluctuations in high-speed rail scenarios with a speed of more than 350 km / h with a delay of less than 1 millisecond. The codebook lookup table reconstruction algorithm for the base index feedback strategy channel is based on the hierarchical precoding matrix codebook that is pre-synchronized between the base station and the terminal during the random access phase. The specific process is as follows: First, extract the 6-bit base index field and the 2-bit channel rank identifier field from the channel state decoding data. Then, convert the base index into the physical address of the codebook. Read the precoding base vector at the corresponding address from the on-chip cache of the base station baseband chip through direct memory access. No complex matrix operations are required. The entire process has a delay of only 0.2 microseconds and a computational complexity close to zero. It is particularly suitable for the low power consumption requirements in stationary or low-speed walking scenarios.

[0092] Following this is the joint superposition reconstruction algorithm for the hybrid strategy channel. This algorithm combines the advantages of the previous two algorithms. Its specific process is as follows: First, the base index part and the differential quantization difference part are separated from the time-division multiplexed hybrid decoded data. The slow-varying stable base component with a channel variation period greater than 100 milliseconds is reconstructed through codebook lookup. Then, the fast-varying fluctuating component with a channel variation period less than 100 milliseconds is obtained through back-projection reconstruction. The two components are then linearly superimposed. Finally, the minimum mean square error algorithm is used to minimize the mean square error between the estimated and true values. Noise correction is applied to the superposition result, with the correction coefficient dynamically adjusted according to the current signal-to-interference-plus-noise ratio (SNR). This ensures reconstruction accuracy while keeping the computational load within a certain range. The back-projection algorithm accounts for less than 60% of the total. Next, scenario algorithm matching is performed. The base station queries the channel-algorithm mapping table pre-embedded in the baseband chip algorithm accelerator based on the target channel feedback strategy channel type recorded after the channel switching is completed. This mapping table is a one-to-one hard-coded mapping relationship. The matching process is a pure hardware table lookup operation with a latency of no more than 0.8 microseconds, which does not introduce any additional processing overhead. The matching rules are strict: differential quantization feedback strategy channels correspond to back-projection reconstruction algorithms, base index feedback strategy channels correspond to codebook lookup table reconstruction algorithms, and hybrid strategy channels correspond to joint superposition reconstruction algorithms. Algorithm mismatch will cause the reconstruction error to increase exponentially. Therefore, the mapping table does not support any dynamic modification.

[0093] Subsequently, the channel rank determination is performed on the channel state decoded data using the matched target channel association reconstruction algorithm. Channel rank refers to the maximum number of independent parallel data streams that a multi-input multi-output channel can support, determined by the channel's spatial correlation and signal-to-interference-plus-noise ratio (SINR). It directly determines the spatial multiplexing gain the system can achieve, i.e., the ability to improve system throughput through parallel transmission of multiple data streams. The rank determination logic of different algorithms has been optimized based on the characteristics of their input data: the back-projection reconstruction algorithm calculates the covariance matrix of the synthesized initial channel matrix, and then performs eigenvalue decomposition on the covariance matrix, that is, decomposing the matrix into a set of orthogonal eigenvectors. The corresponding eigenvalues ​​represent the proportion of signal energy carried by the eigenvector. The number of eigenvalues ​​that are greater than a preset noise threshold, usually 1% of the maximum eigenvalue, and the number of eigenvalues ​​that are lower than the threshold and whose components are noise rather than effective signals, is used as the current channel rank. The codebook lookup table reconstruction algorithm directly extracts a 2-bit channel rank identifier field from the decoded index frame to determine the current channel rank without any additional calculations. The joint superposition reconstruction algorithm performs eigenvalue decomposition on the initial channel matrix after superposition correction to obtain the instantaneous channel rank, and introduces a moving average filter for one time slot, that is, the current final channel rank is the mode of the instantaneous ranks of the first 3 time slots.

[0094] Finally, the target channel association reconstruction algorithm is used to perform dimensionality reduction reconstruction of the channel state decoding data according to the current channel rank to obtain the key channel feature vectors. The core purpose of dimensionality reduction reconstruction is to remove secondary feature components and Gaussian white noise components with an energy ratio of less than 1% in the original channel matrix, and retain only the core information that has the greatest impact on transmission performance. The reconstruction process of different algorithms is as follows: the back projection reconstruction algorithm extracts the top K principal feature vectors corresponding to the current channel rank from the converged complete channel matrix, that is, the K feature vectors with the largest eigenvalues. These principal feature vectors carry more than 99% of the channel energy; the codebook lookup table reconstruction algorithm directly extracts the K orthogonal basis vectors of the corresponding rank from the codebook as the key channel feature vectors; the joint superposition reconstruction algorithm extracts the top K principal feature vectors from the superimposed and corrected channel matrix. The final key channel feature vector is a low-dimensional vector with the same dimension as the channel rank. Compared with the 16-dimensional complex dimension of the original 4×4 MIMO channel matrix, the computational load can be reduced.

[0095] In this embodiment, communication transmission scheduling is analyzed based on the base station-side precoding matrix and link environment-aware data to determine the target communication transmission parameters. This includes: obtaining the communication transmission parameter dimensions; performing communication transmission scheduling analysis based on the base station-side precoding matrix and link environment-aware data according to the communication transmission parameter dimensions to obtain the communication transmission parameter selection space; constructing a link communication transmission reward function according to the link communication transmission target; and using the link communication transmission reward function to optimize the communication transmission parameter selection space to determine the target communication transmission parameters.

[0096] Specifically, the first step is to obtain the communication transmission parameter dimension. This dimension refers to the set of all adjustable physical layer and link layer parameters that can directly affect the transmission performance of the communication system and can be dynamically configured by the base station. It includes six dimensions: modulation and coding scheme, which determines the amount of information carried by each symbol and its anti-interference capability (32 levels); transport block size, which determines the amount of data transmitted within each transmission time interval (TTI); physical resource blocks (PRB), representing the number of time-frequency resources allocated to each PRB in a 5G system (12 subcarriers × 14 symbols); downlink transmit power; hybrid automatic repeat request (HRP), which improves transmission reliability by retransmitting erroneous data packets; and the number of spatial multiplexing layers. The number of spatial multiplexing layers is directly determined by the rank of the base station-side precoding matrix obtained in the previous step and is not adjustable. The remaining five parameters are strongly correlated with indicators such as signal-to-interference-plus-noise ratio (SINR), bit error rate (BER), and latency jitter in the link environment-aware data. Based on the rank constraint of the precoding matrix on the base station side and the performance threshold constraint of the link environment-aware data, a layer-by-layer pruning and filtering process is performed on all parameter combinations according to this parameter dimension: First, a unique value for the number of spatial multiplexing layers is determined based on the rank of the precoding matrix, eliminating all parameter combinations that do not match this number of layers; then, the selectable range of MCS is determined based on the SINR in the link environment-aware data; next, the lower limit of the number of HARQ processes is determined based on the BER; finally, the lower limit of the number of PRB allocations is determined based on latency jitter. This ultimately constructs a communication transmission parameter selection space, which is a low-dimensional feasible parameter set after multiple constraints, reducing the computational complexity of optimization.

[0097] Next, based on the differentiated link communication transmission objectives of the current carried services, a link communication transmission reward function is constructed. This function is not a traditional single throughput indicator function, but a multi-objective weighted comprehensive performance evaluation function that can simultaneously consider the three core performance indicators of throughput, latency, and reliability. Its standardized mathematical expression is:

[0098] ;

[0099] in, This represents the reward for link communication transmission, with a value range of [-1, 1]. A larger value indicates better overall performance. This represents the theoretical throughput under the current parameter combination. This represents the Shannon limit throughput under the current channel conditions. This represents the end-to-end delay under the current parameter combination. This represents the maximum tolerable latency for the current business. This represents the theoretical bit error rate under the current parameter combination. This represents the maximum tolerable bit error rate for the current business. , , For dynamic weighting coefficients, satisfying + + =1, which can be adjusted in real time according to the service type: for real-time services such as voice and video conferencing, set to 1. =0.3、 =0.4、 =0.3, prioritizing real-time and reliable transmission; for non-real-time services such as file downloads and video caching, set to... =0.7、 =0.1、 =0.2, prioritizing maximizing system transmission efficiency; for ultra-reliable, low-latency services such as industrial control, set... =0.2、 =0.3、 =0.5, prioritizing reliability.

[0100] Subsequently, the reward function is used to perform hierarchical iterative optimization of the communication transmission parameter selection space. The specific process is as follows: First, 50 initial communication transmission parameter combinations are randomly selected within the parameter selection space using the Latin hypercube sampling method. The reward value of each combination is calculated, and the top 10 parameter combinations with the highest reward values ​​are selected to form the frontier communication transmission parameter set. This set represents the locally better solutions in the current parameter space. Second, the frontier communication transmission parameter set is used as the initial parent population for crossover and mutation expansion. The crossover operation involves performing single-point crossover of corresponding parameter values ​​between two parent parameter combinations with a probability of 0.7, generating offspring parameter combinations that inherit the superior characteristics of their parents. The mutation operation involves randomly changing the value of a parameter with a probability of 0.1, introducing a new solution space to avoid getting stuck in a specific problem. The first step involves generating an initial population of 200 individuals for communication transmission parameters, achieving local optima. The second step uses a link communication transmission reward function to iteratively evaluate and update the initial population. In each generation, the 50 individuals with the highest reward are retained as parents for the next generation, and crossover and mutation are performed to generate new offspring until a preset maximum of 20 iterations is reached or the reward value converges (i.e., the reward increase is less than 0.1% for three consecutive generations), resulting in a converged communication transmission parameter population space. The third step involves globally maximizing the reward function based on this population space to obtain the globally optimal first communication transmission parameter. The fifth step involves neighborhood growth centered on this first communication transmission parameter, generating small parameter combinations with ±1 step size variations in each adjustable parameter dimension, resulting in a total of [number missing]. By combining neighborhood parameters, a fine-grained optimization process is performed to ultimately determine the target communication transmission parameters with the best overall performance.

[0101] In this embodiment, the link communication transmission reward function is used to optimize the communication transmission parameter selection space to determine the target communication transmission parameters. This includes: randomly selecting N communication transmission parameters within the communication transmission parameter selection space; evaluating and optimizing the N communication transmission parameters using the link communication transmission reward function to obtain a leading set of communication transmission parameters; and solving and optimizing the link communication transmission reward function based on the leading set of communication transmission parameters to determine the target communication transmission parameters.

[0102] Specifically, firstly, in the communication transmission parameter selection space, which is the low-dimensional solution space containing only feasible parameter combinations after pruning by precoding matrix rank constraints and link performance threshold constraints, there is a 5-dimensional adjustable parameter space, including modulation and coding scheme, transport block size, number of physical resource blocks allocated, downlink transmit power, and hybrid automatic repeat request. N communication transmission parameter combinations are randomly selected using the Latin hypercube sampling method, with N being 50. The Latin hypercube sampling method is a hierarchical random sampling technique that ensures uniform sampling points in each parameter dimension. Each parameter combination obtained by sampling is a feasible solution that satisfies all constraints, and its parameter values ​​are all within the range determined by the previous pruning. For example, the value range of the modulation and coding scheme MCS is 20-25 corresponding to the signal-to-interference-plus-noise ratio, and the value range of the number of HARQ processes is 4-8 corresponding to the bit error rate.

[0103] Next, a pre-constructed link communication transmission reward function is used to evaluate and optimize each of the 50 parameter combinations. The evaluation process is executed in parallel in the dedicated algorithm accelerator of the base station baseband chip, with the evaluation time for each parameter combination not exceeding 0.2 microseconds. During the evaluation, the theoretical throughput, end-to-end delay, and bit error rate are first calculated based on the parameter combination: the theoretical throughput is determined by the modulation and coding scheme, the number of physical resource blocks, and the number of spatial multiplexing layers; the end-to-end delay is determined by the transport block size, the number of hybrid automatic repeat request processes, and the scheduling cycle; and the theoretical bit error rate is determined by the modulation and coding scheme and the current signal-to-interference-plus-noise ratio. Then, these calculated values ​​are substituted into the reward function to obtain the reward value for each parameter combination. The larger the reward value, the better the overall performance of the parameter combination. Finally, the 50 parameter combinations are sorted from high to low according to their reward values, and the top 10 parameter combinations with the highest reward values ​​are selected to form the frontier communication transmission parameter set. This set represents the local optimal solution in the current parameter space, which not only preserves the best performing individual but also ensures a certain degree of diversity in the solution space, providing a good starting point for subsequent global optimization.

[0104] Subsequently, the link communication transmission reward function is iteratively optimized based on the leading-edge communication transmission parameter set. The specific process is as follows: First, the leading-edge communication transmission parameter set is used as the initial parent population for crossover and mutation expansion. The crossover operation uses a single-point crossover method, that is, two parent parameter combinations are randomly selected, and their values ​​of one parameter dimension are swapped with a probability of 0.7 to generate offspring parameter combinations that inherit the superior characteristics of the parents. The mutation operation randomly changes the value of a parameter with a probability of 0.1, introducing a new solution space to avoid the algorithm getting trapped in local optima, ultimately generating an initial communication transmission parameter population of size 200. Second, the link communication transmission reward function is used to optimize the initial population. The process involves iterative evaluation and updates. In each generation, the 50 individuals with the highest reward are retained as parents for the next generation. Crossover and mutation are then performed to generate new offspring. This iterative process continues until the preset maximum number of iterations (20) is reached or the convergence condition is met (i.e., the highest reward value of the population increases by less than 0.1% for three consecutive generations), resulting in a converged population space for communication transmission parameters. The third step involves selecting the individual with the highest reward value from the converged population space as the first communication transmission parameter, which is considered the globally approximate optimal solution. The fourth step involves fine-tuning the growth of neighboring individuals centered on the first communication transmission parameter. This involves generating small parameter combinations with a step size of ±1 on each adjustable parameter dimension, resulting in a total of [number missing]. For each neighborhood parameter combination, calculate their reward value one by one, and finally select the neighborhood parameter combination with the highest reward value as the target communication transmission parameter to improve the performance of the parameter combination.

[0105] In this embodiment, the target communication transmission parameters are determined by solving and optimizing the link communication transmission reward function based on the leading-edge communication transmission parameter set. This includes: using the leading-edge communication transmission parameter set as the initial parameter solution for crossover and mutation expansion to obtain an initial communication transmission parameter population; iteratively evaluating and updating the initial communication transmission parameter population using the link communication transmission reward function until a preset termination condition is met, thus obtaining a communication transmission parameter population space; and solving and optimizing the link communication transmission reward function based on the communication transmission parameter population space to determine the target communication transmission parameters.

[0106] Specifically, the algorithm first performs crossover and mutation expansion. Crossover preserves the superior parameter characteristics of the parent generation, while mutation introduces new solution space diversity, preventing premature convergence to local optima. The process is as follows: The resulting frontier communication transmission parameter set containing 10 locally optimal solutions is used as the initial parent population. Parent individuals are randomly paired, forming 5 parent-child combinations. For each parent-child combination, a single-point crossover operation is performed. This involves randomly selecting one of five adjustable parameter dimensions: modulation and coding scheme, transport block size, number of physical resource blocks allocated, downlink transmit power, and number of hybrid automatic repeat request processes. The parameter values ​​of the two parent individuals in that dimension are swapped with a probability of 0.7, generating two offspring that inherit some of the superior characteristics of their parents. For each individual, the parameter values ​​for the crossover operation are strictly limited to the feasible range determined by the previous pruning, avoiding invalid parameter combinations. Subsequently, a mutation operation is performed on all generated offspring individuals, which randomly changes the value of a certain parameter dimension with a probability of 0.1. The mutation method is to adjust the parameter within the feasible range by ±1 steps. For example, the MCS level is adjusted from 22 to 21 or 23, and the PRB allocation number is adjusted from 12 to 11 or 13. The mutation operation can break the limitations of the parent genes, introduce new parameter combinations, and avoid the algorithm getting trapped in local optima. Through the above crossover and mutation operations, an initial communication transmission parameter population of size 200 is finally generated. This population retains the excellent performance characteristics of the frontier parameter set and has sufficient solution space diversity.

[0107] Next, iterative evaluation and updates are performed, a process of cyclical evolution. By continuously selecting superior individuals and eliminating inferior ones, the overall performance of the population is gradually improved. Specifically: First, a pre-built link communication transmission reward function is used to evaluate 200 parameter combinations in the current population in parallel. The evaluation process is executed in a dedicated algorithm accelerator on the base station baseband chip. The evaluation time for each parameter combination does not exceed 0.2 microseconds. During the evaluation, the corresponding theoretical throughput, end-to-end latency, and bit error rate are calculated based on the parameter combination, and then substituted into the reward function to obtain the reward value for each individual. Then, an elite retention strategy is adopted for individual selection, that is, the 50 individuals with the highest reward in the current population are retained as the parent population of the next generation. The elite retention strategy can ensure the superiority of the 50 individuals. The parameter characteristics are not lost during evolution, and the population size is significantly reduced, improving iteration efficiency. Then, crossover and mutation operations are performed again on the new parent population to generate a next-generation population of size 200. The above evaluation, selection, and crossover and mutation process is repeated until the preset termination conditions are met. The termination conditions include two: first, reaching the preset maximum number of iterations of 20; and second, meeting the convergence condition, that is, the highest reward value of the population increases by less than 0.1% for three consecutive generations. The convergence condition can terminate the iteration in advance when the population performance no longer improves significantly, avoiding unnecessary computational overhead. After the iteration terminates, a communication transmission parameter population space is obtained, consisting of the 50 individuals with the highest reward. All individuals in this space are feasible solutions with performance close to the global optimum.

[0108] Finally, optimization is performed, which is the process of extracting the final optimal parameter from the converged population space. This involves two steps: First, the individual with the highest reward value is selected from the communication transmission parameter population space as the first communication transmission parameter. This parameter represents the globally approximate optimal solution, indicating the best result obtained by the genetic algorithm's global search. Second, a fine-grained optimization process is performed using neighborhood individual growth. This compensates for the genetic algorithm's insufficient convergence accuracy near the global optimal solution. Specifically, with the first communication transmission parameter as the center, small parameter combinations with ±1 step size are generated in each adjustable parameter dimension. Since there are 5 adjustable parameter dimensions, a total of [number missing] combinations are generated. For each of the 32 neighborhood parameter combinations, the reward value is calculated one by one. Finally, the neighborhood parameter combination with the highest reward value is selected as the target communication transmission parameter. Neighborhood optimization can perform precise search within a very small range of the global optimal solution, further improving the performance of parameter combinations.

[0109] In this embodiment, the target communication transmission parameters are determined by solving and optimizing the link communication transmission reward function based on the communication transmission parameter population space. This includes: globally maximizing the link communication transmission reward function based on the communication transmission parameter population space to obtain the first communication transmission parameter; and performing neighborhood individual growth and optimization with the first communication transmission parameter as the center to determine the target communication transmission parameter.

[0110] Specifically, firstly, the link communication transmission reward function is globally maximized based on the communication transmission parameter population space. Here, the communication transmission parameter population space refers to the set of 50 feasible parameter combinations with the highest reward, selected by an elite retention strategy after 20 rounds of iterative evolution or meeting the convergence condition of the highest reward increase being less than 0.1% for three consecutive generations. Each combination is a 5-dimensional discrete parameter vector containing modulation and coding scheme, transport block size, number of physical resource blocks allocated, downlink transmit power, and number of hybrid automatic repeat request processes. All parameter values ​​strictly conform to the precoding matrix rank constraint and link performance threshold constraint, and there are no invalid parameter combinations. The maximization process is executed in a single-instruction multiple-data parallel manner in the dedicated algorithm accelerator of the base station baseband chip. It traverses all 50 individuals in the population space and recalculates the reward value of each individual using 64-bit double-precision floating-point operations. Then, all individuals are sorted from high to low reward value, and the individual with the highest reward value is selected as the first communication transmission parameter. This parameter is the globally approximate optimal solution obtained by the genetic algorithm's global search, representing the best-performing coarse-grained parameter combination in the current parameter space. However, since the crossover and mutation operations of the genetic algorithm are adjusted using integer step sizes, there may be unexplored parameter adjustment space near the optimal solution that can bring a small performance improvement.

[0111] Next, neighborhood individual growth optimization is performed with the first communication transmission parameter as the center. Neighborhood individual growth involves exploring all possible small parameter changes within the smallest neighborhood of the globally approximate optimal solution to uncover better, more precise solutions missed by the global search of the genetic algorithm. Specifically: First, the search step size for each of the five adjustable parameter dimensions is ±1, meaning each parameter can only be increased or decreased by 1 from its current value. This is because the communication transmission parameters are discrete integers, and adjustments with a step size greater than 1 have already been fully explored in the previous global evolution process. Then, all possible independent neighborhood parameter combinations are generated. Since each dimension has two independent change directions, +1 and -1, a total of [number] combinations are generated. Different combinations of neighborhood parameters.

[0112] Then, the legality of each of these 32 neighborhood combinations is verified, and invalid combinations that exceed the feasible range of parameters are eliminated. For example, if the MCS level of the first communication transmission parameter is 25 and the feasible upper limit of MCS is 25, then the combination of MCS+1 is invalid; if the PRB allocation quantity is 10 and the feasible lower limit of PRB is 10, then the combination of PRB-1 is invalid.

[0113] Finally, the reward value is calculated for each neighborhood combination that passes the validity check, and compared with the reward value of the first communication transmission parameter. The parameter combination with the highest reward value is selected as the target communication transmission parameter. If the reward value of all valid neighborhood combinations is lower than that of the first communication transmission parameter, the first communication transmission parameter is directly used as the target communication transmission parameter.

[0114] In summary, this embodiment deploys a lightweight link sensing module on the terminal side that reuses the idle computing resources of existing RF front-end and baseband chip. It constructs a six-dimensional link performance index set including received signal strength, signal-to-interference-plus-noise ratio, bit error rate, latency jitter, channel multipath spread, and Doppler shift estimation to achieve accurate composite classification of channel characteristics. Furthermore, it designs a dual-channel, differentiated compression feedback strategy combining differential quantization and base indexing. Based on the channel change rate attribute, it dynamically switches or jointly activates the feedback channel, and uses corresponding back-projection, codebook lookup, and joint superposition reconstruction algorithms to achieve efficient and low-complexity reconstruction of channel characteristics. Simultaneously, it constructs a multi-objective weighted link communication transmission reward function, combining global optimization using a genetic algorithm with fine-grained neighborhood search to achieve optimal scheduling of transmission parameters. Ultimately, it achieves an optimal balance between feedback overhead, computational complexity, and system transmission performance, fully adapting to the communication transmission needs of different mobile speeds and service types.

[0115] Example 2:

[0116] like Figure 2 As shown, this embodiment provides a communication transmission control system based on channel state feedback, including:

[0117] Classification module 11 is used to deploy a lightweight link awareness module on the terminal side, collect link environment awareness data through the lightweight link awareness module, classify the link environment awareness data by channel characteristics, and obtain the channel feature type.

[0118] Matching module 12 is used to construct a dual-channel differential compression feedback strategy, and to match and switch the channel feature type using the dual-channel differential compression feedback strategy to determine the target channel feedback strategy channel;

[0119] Encoding module 13 is used to collect channel state feedback data through the target channel feedback strategy channel, perform precoding selection on the channel state feedback data, and obtain the base station side precoding matrix;

[0120] The parsing module 14 is used to perform communication transmission scheduling parsing based on the base station-side precoding matrix and link environment awareness data, determine the target communication transmission parameters, and perform communication closed-loop control through the target communication transmission parameters.

[0121] The classification module 11 is also used to perform the following steps: constructing a link performance index set, which includes received signal strength, signal-to-interference-plus-noise ratio, bit error rate and delay jitter, channel multipath spread and Doppler shift estimation; performing channel feature extraction calculation on the link environment perception data according to the link performance index set to obtain the channel performance index feature vector; setting feature judgment thresholds for the link performance index set according to the link transmission performance requirements to determine the multi-level judgment threshold for channel type; and classifying the channel performance index feature vector based on the multi-level judgment threshold for channel type to obtain the channel feature type.

[0122] Matching module 12 is also used to perform the following steps: the specific channel architecture of the dual-channel differential compression feedback strategy includes a differential quantization feedback strategy channel and a base index feedback strategy channel; the channel switcher is logically connected to each branch channel in the dual-channel differential compression feedback strategy, wherein the channel switcher adopts a MOSFET power switching circuit; channel switching logic is constructed, specifically, the fast-changing channel activates the differential quantization feedback strategy channel, the slow-changing channel activates the base index feedback strategy channel, and the mixed channel simultaneously activates both the differential quantization feedback strategy channel and the base index feedback strategy channel; the channel switcher performs channel matching and switching according to the channel feature type according to the channel switching logic to determine the target channel feedback strategy channel.

[0123] The encoding module 13 is also used to perform the following steps: receiving and decoding channel state feedback data through the base station to obtain channel state decoded data; performing reconstruction algorithm analysis on each logical scenario channel in the dual-channel differential compression feedback strategy to determine the scenario channel association reconstruction algorithm; using the scenario channel association reconstruction algorithm to reconstruct channel features from the channel state decoded data to obtain key channel feature vectors; and performing precoding selection in the hierarchical precoding matrix codebook pre-stored in the base station based on the key channel feature vectors to obtain the base station-side precoding matrix.

[0124] Encoding module 13 is also used to perform the following steps: The scene channel association reconstruction algorithm specifically includes a back projection reconstruction algorithm for the differential quantization feedback strategy channel, a base index feedback strategy channel for the codebook lookup table reconstruction algorithm, and a hybrid strategy channel for the joint superposition reconstruction algorithm; the scene channel association reconstruction algorithm is used to match the channel state decoding data with scene algorithms to determine the target channel association reconstruction algorithm; the scene channel association reconstruction algorithm is used to perform channel rank determination on the channel state decoding data to obtain the current channel rank; the scene channel association reconstruction algorithm is used to reconstruct the channel features of the channel state decoding data according to the current channel rank to obtain the channel key feature vector.

[0125] The parsing module 14 is also used to perform the following steps: obtain the communication transmission parameter dimension, perform communication transmission scheduling parsing based on the base station side precoding matrix and link environment awareness data according to the communication transmission parameter dimension, and obtain the communication transmission parameter selection space; construct the link communication transmission reward function according to the link communication transmission target; use the link communication transmission reward function to optimize the communication transmission parameter selection space and determine the target communication transmission parameters.

[0126] The parsing module 14 is also used to perform the following steps: randomly select N communication transmission parameters within the communication transmission parameter selection space; evaluate and optimize the N communication transmission parameters using the link communication transmission reward function to obtain the leading communication transmission parameter set; and solve and optimize the link communication transmission reward function based on the leading communication transmission parameter set to determine the target communication transmission parameter.

[0127] The parsing module 14 is also used to perform the following steps: cross-mutate and expand the frontier communication transmission parameter set as the initial parameter solution to obtain the initial communication transmission parameter population; iteratively evaluate and update the initial communication transmission parameter population using the link communication transmission reward function until a preset termination condition is met to obtain the communication transmission parameter population space; and solve and optimize the link communication transmission reward function based on the communication transmission parameter population space to determine the target communication transmission parameters.

[0128] The parsing module 14 is also used to perform the following steps: global maximization of the link communication transmission reward function based on the population space of the communication transmission parameters to obtain the first communication transmission parameter; and neighborhood individual growth and optimization with the first communication transmission parameter as the center to determine the target communication transmission parameter.

Claims

1. A communication transmission control method based on channel state feedback, characterized in that, Includes the following steps: S1: Deploy a lightweight link awareness module on the terminal side, collect link environment awareness data through the lightweight link awareness module, and classify the link environment awareness data by channel characteristics to obtain the channel characteristic type; S2: Construct a dual-channel differential compression feedback strategy, and use the dual-channel differential compression feedback strategy to match and switch the channel feature type to determine the target channel feedback strategy channel; S3: Collect channel state feedback data through the target channel feedback strategy channel, perform precoding selection on the channel state feedback data, and obtain the base station side precoding matrix; S4: Based on the precoding matrix and link environment awareness data of the base station side, perform communication transmission scheduling analysis to determine the target communication transmission parameters, and perform closed-loop control of communication through the target communication transmission parameters; Step S2 employs a differentiated compression feedback strategy to match and switch channel feature types using dual channels. The method for determining the target channel feedback strategy channel includes: B1: The specific channel architecture of the dual-channel differential compression feedback strategy includes differential quantization feedback strategy channel and base index feedback strategy channel; The channel switcher is logically connected to each branch channel in the dual-channel differential compression feedback strategy, wherein the channel switcher adopts a MOSFET power switching circuit. B2: Construct channel switching logic, which includes fast-changing channel activating differential quantization feedback strategy channel, slow-changing channel activating base index feedback strategy channel, and hybrid channel activating both differential quantization feedback strategy channel and base index feedback strategy channel simultaneously. B3: The channel switcher performs channel matching and switching based on channel characteristic types according to the channel switching logic to determine the target channel feedback strategy channel; The method for precoding selection of channel state feedback data in step S3 to obtain the base station-side precoding matrix includes: C1: The channel state feedback data is received and decoded by the base station to obtain the channel state decoded data; C2: Analyze the reconstruction algorithm for each logical scenario channel in the dual-channel differential compression feedback strategy to determine the scenario channel association reconstruction algorithm; C3: The scenario channel association reconstruction algorithm is used to reconstruct the channel features of the channel state decoding data to obtain the key channel feature vector; C4: In the base station, a hierarchical precoding matrix codebook is pre-stored. Based on the key feature vectors of the channel, precoding selection is performed in the hierarchical precoding matrix codebook to obtain the base station-side precoding matrix.

2. The communication transmission control method based on channel state feedback according to claim 1, characterized in that, The method for classifying the channel feature types of the link environment-aware data in step S1 includes: A1: Construct a set of link performance metrics, which includes received signal strength, signal-to-interference-plus-noise ratio, bit error rate and delay jitter, channel multipath spread and Doppler shift estimation; A2: Perform channel feature extraction and calculation on the link environment perception data according to the link performance index set to obtain the channel performance index feature vector; A3: Set feature judgment thresholds for the link performance index set according to the link transmission performance requirements, and determine multi-level judgment thresholds for channel types; A4: Based on the multi-level judgment threshold of channel type, the channel performance index feature vector is classified to obtain the channel feature type.

3. The communication transmission control method based on channel state feedback according to claim 1, characterized in that, Step C3 includes: The scene channel association reconstruction algorithm specifically includes a back projection reconstruction algorithm for differential quantization feedback strategy channels, a codebook lookup table reconstruction algorithm for base index feedback strategy channels, and a joint superposition reconstruction algorithm for hybrid strategy channels; A scenario-based channel association reconstruction algorithm is used to match the channel state decoding data with the scenario algorithm to determine the target channel association reconstruction algorithm. The current channel rank is obtained by performing channel rank determination on the channel state decoding data through the scene channel association reconstruction algorithm; The scenario channel association reconstruction algorithm is used to reconstruct the channel features of the channel state decoding data according to the current channel rank, and obtain the key channel feature vector.

4. The communication transmission control method based on channel state feedback according to claim 1, characterized in that, The method for determining the target communication transmission parameters in step S4, based on the base station-side precoding matrix and link environment-aware data, includes: D1: Obtain the communication transmission parameter dimension, and perform communication transmission scheduling analysis based on the base station-side precoding matrix and link environment awareness data according to the communication transmission parameter dimension to obtain the communication transmission parameter selection space; D2: Construct the link communication transmission reward function based on the link communication transmission target; D3: Use the link communication transmission reward function to optimize the communication transmission parameter selection space and determine the target communication transmission parameters.

5. The communication transmission control method based on channel state feedback according to claim 4, characterized in that, Step D3 includes: E1: Randomly select N communication transmission parameters within the communication transmission parameter selection space; E2: The link communication transmission reward function is used to evaluate and optimize N communication transmission parameters to obtain the frontier communication transmission parameter set; E3: Based on the frontier communication transmission parameter set, the link communication transmission reward function is solved and optimized to determine the target communication transmission parameters.

6. The communication transmission control method based on channel state feedback according to claim 5, characterized in that, Step E3 includes: F1: The frontier communication transmission parameter set is used as the initial parameter solution and cross-mutation is performed to expand it, so as to obtain the initial communication transmission parameter population. F2: Use the link communication transmission reward function to iteratively evaluate and update the initial communication transmission parameter population until the preset termination condition is met, and obtain the communication transmission parameter population space. F3: Based on the population space of communication transmission parameters, the link communication transmission reward function is solved and optimized to determine the target communication transmission parameters.

7. The communication transmission control method based on channel state feedback according to claim 6, characterized in that, Step F3 includes: The first communication transmission parameter is obtained by globally maximizing the link communication transmission reward function based on the population space of the communication transmission parameter. The target communication transmission parameters are determined by growing and optimizing neighboring individuals based on the first communication transmission parameter.

8. A communication transmission control system based on channel state feedback, characterized in that, The system for implementing the method according to any one of claims 1 to 7 comprises: The classification module is used to deploy a lightweight link awareness module on the terminal side. The lightweight link awareness module collects link environment awareness data, classifies the link environment awareness data by channel characteristics, and obtains the channel feature type. The matching module is used to construct a dual-channel differential compression feedback strategy. It uses the dual-channel differential compression feedback strategy to match and switch the channel feature type to determine the target channel feedback strategy channel. The encoding module is used to collect channel state feedback data through the target channel feedback strategy channel, perform precoding selection on the channel state feedback data, and obtain the base station side precoding matrix. The parsing module is used to perform communication transmission scheduling parsing based on the base station-side precoding matrix and link environment awareness data, determine the target communication transmission parameters, and perform communication closed-loop control based on the target communication transmission parameters.

Images