HARQ (Hybrid Automatic Repeat Request) retransmission decision-making method, device, equipment, medium and product

Abstract

Disclosed are an HARQ retransmission decision method, apparatus, device, medium and product, the method comprising: based on constellation point coordinates and a constellation point and Gray code mapping graph, a receiver determining an LLR value matrix thereof according to an SNR value of a transmission signal; inputting the LLR value matrix into a trained HARQ network module, outputting a retransmission probability, and deciding whether retransmission is needed or not according to a set retransmission threshold value; if retransmission is needed, the sender resends the data, the receiver receives the data again, the matrix is calculated and input into the module to obtain a new retransmission probability, the operation is cycled until the decision is no or the maximum number of retransmission times is reached, and the receiver determines a final signal according to the repeated retransmission probability; if retransmission is not needed, the final signal is determined according to the repeated retransmission probability. After the data transmission is finished, calculating the network loss according to the current transmission times, the maximum retransmission times and the SNR value of each time of transmission; and training the HARQ network module to determine parameters based on the loss. According to the invention, the image transmission effect is improved.

Description

Technical Field

[0001] The present application relates to the field of communications, and in particular to a HARQ retransmission decision method, device, equipment, medium and product. Background Art

[0002] With the rapid development of wireless communication technology, wireless image transmission systems have wide and critical applications in many fields, such as security monitoring, telemedicine, drone aerial photography, and intelligent transportation. However, existing wireless image transmission technology faces many severe challenges, which seriously restricts its performance and application expansion.

[0003] Traditional wireless image transmission solutions usually rely on established channel coding and modulation and demodulation technologies, combined with the Hybrid Automatic Repeat Request (HARQ) mechanism to ensure the reliability of data transmission. However, in the complex and changeable actual wireless environment, its drawbacks are becoming more and more prominent.

[0004] On the one hand, traditional HARQ systems have significant limitations in retransmission decision-making. When faced with complex situations such as time-varying channel fading and noise interference, it is difficult to accurately adjust the retransmission strategy dynamically based on the real-time channel status. For example, at the moment when the channel condition deteriorates sharply, it is impossible to quickly determine whether retransmission is urgently needed and at what probability to start retransmission. This often leads to either excessive retransmissions resulting in a waste of computing resources, or too few retransmissions resulting in serious damage to the image quality of the receiver and loss of key information, which greatly affects the user experience and the overall transmission efficiency of the system.

[0005] On the other hand, from the perspective of end-to-end transmission, autoencoder technology has been introduced into the field of wireless image transmission in recent years, aiming to achieve efficient compression and reconstruction of images, hoping to optimize the transmission process by learning the intrinsic characteristics of the data. The autoencoder consists of a pair of complementary network architectures, which constitute the encoder and decoder respectively. The encoder is responsible for compressing the input image into a low-dimensional feature representation, and the decoder tries to restore the original image based on the feature representation. This system is also called the Deep Joint Source-Channel Coding (DJSCC) system.

[0006] Although current autoencoders can exhibit certain advantages, such as being able to adaptively optimize encoding according to data characteristics. However, it is difficult to reconstruct good-quality images under low signal-to-noise ratio (SNR), and it is also difficult to achieve ideal performance when working in coordination with existing Hybrid Automatic Repeat reQuest (HARQ) mechanisms. The reason is that there is a lack of an effective fusion strategy between the two. After the output of the autoencoder directly enters the traditional HARQ system, the HARQ system cannot fully utilize the feature information extracted by the autoencoder, making it difficult to efficiently calculate data reliability and guide subsequent retransmissions when facing transmission errors or poor channels, and unable to fully exploit the advantages of both, ultimately resulting in a bottleneck in the performance improvement of the overall wireless image transmission system.

[0007] To improve the quality of image transmission in the DJSCC system under low SNR, the literature "DAVID G, E al. Deep Joint Source-Channel Coding with Feedback for Wireless Image Transmission[J]. IEEE Transactions on Communications, 2020, 68(5): 2661 - 2675." proposed a system with feedback, namely DJSCC-f. This system alleviates this problem by transmitting the image multiple times, but this approach requires additional codecs and combiners for each training, greatly increasing the training complexity and cost. When the number of layers is too large, it may even lead to problems such as a decline in image transmission effect. Moreover, during data transmission, it is a single transmission, and some semantic information will be lost during the transmission process, especially under low SNR channel conditions, resulting in a further decline in the image transmission effect. Summary of the Invention

[0008] The objective of this application is to provide a HARQ retransmission decision method, device, equipment, medium, and product to solve the problem of the decline in image transmission effect.

[0009] To achieve the above objective, this application provides the following solutions.

[0010] In the first aspect, this application provides a HARQ retransmission decision method, including the following steps.

[0011] Based on the constellation point coordinates and the constellation point - Gray code mapping diagram, according to the SNR value of the transmitted signal, perform Log-Likelihood Ratio (LLR) calculation on the received transmitted signal by the receiver to determine the LLR value matrix.

[0012] Input the LLR value matrix into the trained HARQ network module to output the retransmission probability; the HARQ network module is constructed based on the constellation point coordinates and the constellation point - Gray code mapping diagram.

[0013] Determine whether the received transmission signal needs to be retransmitted according to the retransmission probability and the set retransmission threshold.

[0014] If so, determine the finally received transmission signal according to the retransmission probabilities of multiple retransmissions.

[0015] If not, calculate the network loss according to the current transmission times, the set maximum retransmission times, and the SNR value of each transmission signal.

[0016] Train the HARQ network module based on the network loss to determine the trained HARQ network module.

[0017] In a second aspect, the present application provides a HARQ retransmission decision device, including the following modules.

[0018] LLR value matrix determination module, configured to perform LLR calculation on the transmission signal received by the receiver according to the SNR value of the transmission signal, based on the constellation point coordinates and the constellation point and Gray code mapping diagram, to determine the LLR value matrix.

[0019] Retransmission probability output module, configured to input the LLR value matrix into the trained HARQ network module and output the retransmission probability; the HARQ network module is constructed based on the constellation point coordinates and the constellation point and the Gray code mapping diagram.

[0020] Judgment module, configured to determine whether the received transmission signal needs to be retransmitted according to the retransmission probability and the set retransmission threshold.

[0021] Finally received transmission signal determination module, configured to determine the finally received transmission signal according to the retransmission probabilities of multiple retransmissions.

[0022] Network loss calculation module, configured to calculate the network loss according to the current transmission times, the set maximum retransmission times, and the SNR value of each transmission signal.

[0023] Training module, configured to train the HARQ network module based on the network loss to determine the trained HARQ network module.

[0024] In a third aspect, the present application provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the computer program to implement the HARQ retransmission decision method described in any one of the above.

[0025] In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the HARQ retransmission decision method described in any one of the above is implemented.

[0026] Fifth aspect, the present application provides a computer program product, including a computer program which, when executed by a processor, implements the HARQ retransmission decision method described in any one of the above.

[0027] According to the specific embodiments provided by the present application, the following technical effects are disclosed: The present application accurately determines the retransmission probability based on information such as the log-likelihood ratio (LLR), signal-to-noise ratio (SNR), and the current number of transmissions, and retransmits and combines the transmitted data based on the trained HARQ network module, thereby reducing the loss of semantic information of the image under low SNR, improving the effect of the reconstructed image of the DJSCC system, achieving a harmonious combination of HARQ and the DJSCC system, and improving the image transmission effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0029] Figure 1 It is a schematic flow chart of the HARQ retransmission decision method provided by the present application.

[0030] Figure 2 It is an example diagram of the constellation point coordinates, Gray code, and constellation point mapping dictionary under 64-QAM provided by the present application.

[0031] Figure 3 It is a structural diagram of combining QAM communication provided by the present application.

[0032] Figure 4 It is a structural diagram of the HARQ network module provided by the present application.

[0033] Figure 5 It is a comparison diagram of simulation experiments for the relationship between PSNR and SNR of the present application and the original model.

[0034] Figure 6 It is a comparison diagram of simulation experiments for the relationship between SSIM and SNR of the present application and the original model. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0035] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0036] To make the above objects, features, and advantages of the present application more obvious and understandable, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0037] The embodiment of the present application provides a HARQ retransmission decision method, which is executed by a computer device. Specifically, it can be executed alone by a computer device such as a terminal or a server, or jointly executed by a terminal and a server. In the embodiment of the present application, as Figure 1 shown, the method includes the following steps.

[0038] S1: Based on the constellation point coordinates and the constellation point and Gray code mapping diagram, according to the SNR value of the transmission signal, perform LLR calculation on the transmission signal received by the receiving party to determine the LLR value matrix.

[0039] S2: Input the LLR value matrix into the trained HARQ network module to output the retransmission probability; the HARQ network module is constructed based on the constellation point coordinates and the constellation point and the Gray code mapping diagram.

[0040] S3: According to the retransmission probability and the set retransmission threshold, decide whether the received transmission signal needs to be retransmitted. If so, execute S4; if not, execute S5.

[0041] S4: Instruct the sender to retransmit the transmission signal, and the receiver to re-receive the transmission signal and calculate the LLR value matrix and send it to the HARQ network module to output the retransmission probability until the receiver decides no or reaches the maximum number of retransmissions. Then, instruct the receiver to determine the finally received transmission signal according to the retransmission probabilities of multiple retransmissions.

[0042] S5: Determine the finally received transmission signal according to the retransmission probabilities of multiple retransmissions, and after the current transmission signal is transmitted, calculate the network loss according to the current transmission times, the set maximum number of retransmissions, and the SNR value of each transmission signal.

[0043] S6: Based on the network loss, train the HARQ network module to determine the trained HARQ network module.

[0044] In an exemplary embodiment, before S1, it further includes: training the HARQ network module using the CIFAR-10 image dataset. The images in the CIFAR-10 image dataset have a size of 32×32 and include 50,000 training images and 10,000 test images.

[0045] Calculate the number of bits corresponding to each symbol signal according to the input modulation order M (M≥2), and generate a Gray code array with a length of log2M through operations such as bitwise operations. For each integer n, n∈[0, M-1], first calculate its corresponding Gray code, and the generation method of the Gray code can be expressed as follows.

[0046] a n = n⊕(n >> 1) (1)

[0047] where, ⊕ represents bitwise exclusive OR operation, and >> represents right shift operation.

[0048] In this way, a Gray code sequence a = [a0, a1,... a M-1 can be obtained.

[0049] Then calculate the constellation point coordinates. In the two-dimensional plane, the coordinates of the constellation points can be calculated through the Gray code mapping diagram, as Figure 2 shown.

[0050] Assume the dimension of the constellation points Then the constellation point coordinates can be calculated through the following steps: n gary = a n , where n∈{0, 1,... M-1}.

[0051] Subsequently, calculate the horizontal axis (I axis) and vertical axis (Q axis) coordinates based on the Gray code index.

[0052] (x, y) = divmod(n gary , D)(2)

[0053] Let x and y be the indices of the horizontal axis and vertical axis respectively, then the coordinates of the constellation points are:

[0054] Ax = 2x + 1 - D(3)

[0055] Ay = 2y + 1 - D (4)

[0056] Generate an integer value ranging from -(D - 1) to D - 1, representing the coordinates of the constellation points.

[0057] Finally, in order to standardize the constellation points C, scale them to within the unit circle range, and a set of constellation points c = {c1, c2,..., c m|m = M}, can be expressed as follows.

[0058]

[0059] Next, create a mapping dictionary from constellation points to Gray codes, associate the coordinates of each constellation point with its corresponding Gray code, and finally return the constellation diagram coordinates and the mapping dictionary from constellation points to Gray codes. This mapping dictionary is the constellation point coordinates, the mapping diagram of constellation points and Gray codes, and these information are used as the known information for both communication parties.

[0060] The sender encodes and quantizes the data: As Figure 3 shown, the input image is subjected to feature extraction and size compression through an Encoder module. The output of this Encoder module is the image data Z.

[0061] The image data Z is sent to a Quantizer module for hard quantization to simulate the modulation process. For each input z n ∈ Z, the constellation point set is c = {c1, c2, …, c m |m = M}, calculate the distance between the input z n and all constellation points in the set C, and quantize it to the nearest constellation point, which can be expressed as:

[0062]

[0063] Among them, indicates that the encoded data is modulated into constellation point information, z n (j) is the original value of the input data; c k (j) is the coordinate of the j-th constellation point; c k is the constellation point coordinate (in general); j is the number of input data, that is, calculate the distance between each input symbol and all constellation points and replace the original input symbol with the constellation point with the minimum distance; the quantized transmitted data is denoted as

[0064] The transmitted data passes through the channel: The transmitted data enters an Additive White Gaussian Noise (AWGN) channel with a random SNR in the range of [0, 5] dB, which can be expressed as follows.

[0065]

[0066] Among them, ξ is a Gaussian distributed noise vector with a mean of 0 and a variance of σ 2 , and the data Y after passing through the channel is obtained through this formula.

[0067] The HARQ network module receives data and makes a decision: At the receiver, the HARQ network module receives the data after passing through the AWGN channel and performs verification. For each symbol signal y in the received data Y n , each carries log2M bits of information. Therefore, log2M LLR values need to be calculated for each symbol signal.

[0068] Let the bit LLR value corresponding to the nth received symbol signal be LLR n,k , k = 1, 2,..., log2M, and its calculation formula is as follows.

[0069]

[0070] where P(y n |c) is the probability density function of the received symbol signal y n under the condition that the transmitted symbol is c, c is the coordinate of a single constellation point, is the set of constellation points where the kth bit is 1; is the set of constellation points where the kth bit is 0.

[0071] In the AWGN channel, this probability density function is a Gaussian distribution. Therefore, in the AWGN channel, the LLR can be expressed as follows.

[0072]

[0073] From this, the LLR value matrix L of the received data can be obtained.

[0074] In an exemplary embodiment, S1 can be replaced by the following steps.

[0075] S11: Determine the noise intensity according to the SNR value of the transmission signal;

[0076] S12: Determine the LLR value matrix according to the symbol signal in the transmission signal, the noise intensity, and the Gray code mapping diagram; the symbol signal includes the complex tensor [B, N] or real tensor [B, N, 2] format of the constellation points, B is the batch size of the transmission data, N is the complex value saved in complex form, and the value of N depends on the transmission data of the sender.

[0077] In an exemplary embodiment, S2 can be replaced by the following steps.

[0078] As Figure 4As shown, the trained HARQ network module includes an input layer, a feature extraction network, and a decision network; the feature extraction network includes a first fully connected layer, a first activation layer, a first batch normalization layer, and a first dropout layer connected in sequence; the decision network includes a feature splicing layer, a second fully connected layer, a second activation layer, a second batch normalization layer, a second dropout layer, a third fully connected layer, a third activation layer, and a normalization layer connected in sequence; the first dropout layer and the second dropout layer are Dropout layers.

[0079] S21: Input the LLR value matrix into the feature extraction network to extract the feature vector.

[0080] S22: Calculate the retransmission penalty factor according to the maximum allowable retransmission times and the current transmission times.

[0081] S23: Input the feature vector, SNR value, and the retransmission penalty factor into the decision network, splice them, and output the retransmission probability.

[0082] In practical applications, taking the calculated LLR value matrix L and the SNR value of the current channel as the input of the HARQ network module, the process of extracting the feature vector F can be expressed as follows.

[0083] h1 = Dropout(BatchNorm(ReLU(LW1 + b1))) (10)

[0084] Among them, W1 is the weight matrix of the feature extraction layer, and b1 is the bias matrix of the feature extraction layer.

[0085] Furthermore, the retransmission penalty factor π is calculated to control the retransmission times, and its calculation can be expressed as follows.

[0086]

[0087] Among them, n is the current transmission times, and N max is the maximum allowable retransmission times.

[0088] Furthermore, the calculation of the decision network output can be expressed as follows:

[0089] h1 = Dropout(BatchNorm(ReLU(F c W2 + b2))) (12)

[0090] h2 = ReLU(h1W3 + b3) (13)

[0091] p = σ(h2W4 + b4) (14)

[0092] Among them, h1 and h2 are intermediate values, and F c=[F; SNR; π] is the concatenated feature vector, W2, W3, W4 are the weight matrices of each layer, σ(·) is the Sigmoid activation function, P∈[0,1] is the final retransmission probability; RELU(·) is the rectified linear unit (RELU) activation function; BatchNorm(·) is batch normalization; Dropout(·) is an illustration of the random dropout function.

[0093] In an exemplary embodiment, S3 may be replaced by the following steps.

[0094] S31: Determine whether the retransmission probability is greater than or equal to the set retransmission threshold, if so, execute S32, if not, execute S33.

[0095] S32: Determine that the current transmission signal is unreliable, send a NACK signal to the sender, and trigger a retransmission mechanism;

[0096] S33: Determine that the transmission signal is reliable and send an ACK signal to the sending method.

[0097] In practical applications, according to a given retransmission threshold θ (the default value is 0.5), the decision of the HARQ network module can be expressed as:

[0098]

[0099] If signal = negative ACKnowledgment (NACK), the data transmission is considered unreliable and the retransmission mechanism is triggered. If signal = ACKnowledgment (ACK), the data is considered reliable and no retransmission is performed.

[0100] When the data is unreliable, that is, signal = NACK, the HARQ network module will store the data and send a NACK signal to the sender.

[0101] In an exemplary embodiment, S4 may be replaced by the following steps.

[0102] S41: Determine a weighted weight of a transmission signal for each retransmission according to the retransmission probabilities of multiple retransmissions.

[0103] S42: normalizing the weighted weights to determine normalized weights.

[0104] S43: Determine the transmission signal finally received according to each retransmitted transmission signal and the corresponding normalized weight.

[0105] In practical applications, after receiving the NACK signal, the sender will re-encode and modulate the data transmitted this time and transmit it through the channel. The HARQ network module at the receiver will re-execute the above decision process until the decision condition (signal = ACK) is met or the set maximum number of retransmissions is reached.

[0106] For the results of multiple retransmissions \(Y = [Y_1, Y_2, \ldots, Y N \), where \(N\) is the set maximum number of retransmissions, the corresponding \(w N =(1 - P N ) is used as the weight for weighted combination, denoted as \(w = [w_1, w_2, \ldots, w N \). The softmax function is used to perform the normalization operation on it. For the \(i\)-th weight \(w i \), its normalized result \(w i '\) can be expressed as:

[0107]

[0108] Finally, by multiplying the weights with the corresponding transmitted data, the weighted combined signal can be expressed as follows.

[0109]

[0110] In an exemplary embodiment, the HARQ network module sends to the decoder module for image restoration and reconstruction, obtains the restored and reconstructed image, and calculates the loss based on the current number of retransmissions, the set maximum number of retransmissions, the SNR value of each transmitted signal, and the distortion degree between the reconstructed image and the real image to train the neural network, which can be expressed as follows.

[0111]

[0112] Where, is the total loss, is the loss of the original DJSCC system. For the HARQ network module, its loss can be expressed as follows.

[0113]

[0114] Where, is the loss function; \(n\) is the current number of transmissions; \(N max is the maximum number of retransmissions; \(N\) is the actual number of retransmissions; \(p n is the predicted probability of the \(n\)-th transmission; \(\sigma(\cdot)\) is the sigmoid function; \(SNR n is the signal-to-noise ratio of the \(n\)-th transmission.

[0115] Through an end-to-end neural network architecture, namely the HARQ network module, this application realizes an adaptive retransmission decision strategy. This application not only considers the quality characteristics of signals, but also introduces a penalty factor to balance the number of retransmissions, which can effectively balance the transmission efficiency and reliability and optimize the transmission performance of the communication system. The present invention can be applied to the optimization of the HARQ mechanism of various wireless communication systems modulated by Quadrature Amplitude Modulation (QAM), such as Figure 3 as shown, which has important practical value.

[0116] Based on the technical solution of this application, the data simulation experiment is as follows.

[0117] To verify the effectiveness of this application, a simulation experiment is carried out and compared with the original model DJSCC-Q in terms of the image semantic metrics Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM).

[0118] In the simulation experiment, both the HARQ network module of this application and the original model are trained using the CIFAR-10 image dataset. The QAM modulation order M = 64 is set, the default retransmission threshold σ = 0.5, the maximum number of retransmissions max_retrans = 5, and the model training epoch is set to 150 rounds. The signal-to-noise ratio for each round is set to a random number between SNR = [0, 5] dB, and the experiment is carried out on the AWGN channel.

[0119] The system calculates the number of bits corresponding to each symbol according to the input modulation order, and generates a Gray code array of length through operations such as bitwise operations, and creates a constellation diagram coordinate, constellation point, and Gray code mapping diagram.

[0120] For each epoch, 64 pictures in the dataset are selected as the same batch, and the dimension of each picture is [32, 32, 3]. They are sent to the Encoder module for encoding, and the Encoder module outputs a feature vector Z with a dimension of [512, 2]. This feature vector is in the form of real numbers of complex values.

[0121] Then it is sent to the Quantizer module for hard quantization. For each batch of inputs z n ∈Z, calculate the distance between each z n and all constellation points in the set C, and quantize it to the nearest constellation point. The quantized data is represented as whose dimension is still [512, 2]. The quantization process only quantizes the discrete complex values to the specified constellation point coordinates.

[0122] Next, is sent into an AWGN channel with SNR = [0, 5] dB, and the data after passing through the channel is y. At the receiving end, after the HARQ network module receives the data y, according to Figure 4 the shown process, after obtaining the retransmission probability p and signal, the transmitted data is retransmitted again until the decision condition is met. For each transmission result y n and the corresponding weight w n are combined, and the resulting matrix still has a dimension of [512, 2].

[0123] Next, is sent into the Decoder module for image restoration and reconstruction to obtain the restored and reconstructed image. The dimension of each image is restored to [32, 32, 3]. Finally, the distortion between the restored image and the original image is calculated for the neural network to train.

[0124] Figures 5 - 6 To compare the results of this application and the original model DJSCC-Q under the configuration where the computer is Windows 10, Intel Core i5-13400F CPU, RTX3060TI GPU, 32GB RAM, and the training epoch is 150 rounds, the abscissa of both is SNR, and the ordinates are PSNR and SSIM respectively; Figures 5 - 6 The upper curve in is this application with the HARQ network module added, and the lower curve is the original model.

[0125] For this application from 0 to 5 signal-to-noise ratio: in terms of PSNR, the improvement compared to the original model metrics ranges from 18.59% to 13.06%; in terms of SSIM, the improvement compared to the original model metrics ranges from 7.79% to 3.18%. This shows that under the same low signal-to-noise ratio, this application can obtain a restored image with higher semantic metrics compared to the original model.

[0126] Based on the same inventive concept, the embodiment of this application also provides a HARQ retransmission decision device for implementing the above-mentioned HARQ retransmission decision method. The implementation solution provided by this device to solve the problem is similar to the implementation solution described in the above method. Therefore, the specific limitations in one or more embodiments of the HARQ retransmission decision device provided below can refer to the limitations on the HARQ retransmission decision method in the above text, and will not be repeated here.

[0127] In an exemplary embodiment, a HARQ retransmission decision device is provided, which includes the following modules.

[0128] The LLR value matrix determination module is used to perform LLR calculation on the transmitted signal received by the receiving party based on the constellation point coordinates, the constellation point and Gray code mapping diagram, and according to the SNR value of the transmitted signal, so as to determine the LLR value matrix.

[0129] The retransmission probability output module is used to input the LLR value matrix into the trained HARQ network module and output the retransmission probability; the HARQ network module is constructed based on the constellation point coordinates and the constellation point and Gray code mapping diagram.

[0130] The judgment module is used to decide whether the received transmitted signal needs to be retransmitted according to the retransmission probability and the set retransmission threshold.

[0131] The finally received transmitted signal determination module is used to determine the finally received transmitted signal according to the retransmission probabilities of multiple retransmissions.

[0132] The network loss calculation module is used to calculate the network loss according to the current transmission times, the set maximum retransmission times and the SNR value of each transmitted signal.

[0133] The training module is used to train the HARQ network module based on the network loss to determine the trained HARQ network module.

[0134] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store HARQ retransmission decision data. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals through a network connection. When the computer program is executed by the processor, a HARQ retransmission decision method is implemented.

[0135] In an exemplary embodiment, a computer device is provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the above method is implemented.

[0136] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program which, when executed by a processor, implements the above method.

[0137] In an exemplary embodiment, a computer program product is provided, including a computer program which, when executed by a processor, implements the above method.

[0138] Those of ordinary skill in the art can understand that all or part of the processes in the above method embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above method embodiments. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memories can include read-only memory (ROM), magnetic tapes, floppy disks, flash memories, optical memories, high-density embedded non-volatile memories, resistive random-access memories (ReRAM), magnetoresistive random-access memories (MRAM), ferroelectric random-access memories (FRAM), phase change memories (PCM), graphene memories, etc. Volatile memories can include random-access memories (RAM) or external cache memories, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random-access memory (SRAM) or dynamic random-access memory (DRAM), etc.

[0139] In this application, all actions of obtaining signals, information, or data are carried out on the premise of complying with the corresponding data protection regulations and policies of the country where it is located and obtaining authorization from the owner of the corresponding device.

[0140] The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.

[0141] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0142] In this article, specific examples are used to elaborate on the principles and implementation manners of the present application. The descriptions of the above embodiments are only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A HARQ retransmission decision method, characterized in that: The HARQ retransmission decision method comprises: Based on the coordinates of the constellation points and the mapping diagram of the constellation points and the Gray code, according to the SNR value of the transmission signal, the LLR calculation is performed on the transmission signal received by the receiver to determine the LLR value matrix; Inputting the LLR value matrix into a trained HARQ network module and outputting a retransmission probability; the HARQ network module is constructed based on the constellation point coordinates, the constellation points and the Gray code mapping diagram; Determining whether the received transmission signal needs to be retransmitted according to the retransmission probability and a set retransmission threshold; If yes, the sender is ordered to resend the transmission signal, the receiver receives the transmission signal again and calculates the LLR value matrix and sends it to the HARQ network module, outputs the retransmission probability, until the receiver decides whether to receive the transmission signal or reaches the maximum number of retransmissions, and the receiver determines the final received transmission signal based on the retransmission probabilities of multiple retransmissions; If not, determine the final received transmission signal based on the retransmission probability of multiple retransmissions, and after the current transmission signal is transmitted, calculate the network loss based on the current transmission number, the set maximum retransmission number and the SNR value of each transmission signal; Based on the network loss, a HARQ network module is trained to determine a trained HARQ network module.

2. The HARQ retransmission decision method according to claim 1, characterized in that: Based on the constellation point coordinates, the constellation point and the Gray code mapping diagram, and according to the SNR value of the transmission signal, the LLR calculation is performed on the transmission signal received by the receiver to determine the LLR value matrix, which specifically includes: Determining noise intensity according to the SNR value of the transmission signal; An LLR value matrix is ​​determined according to the symbol signal in the transmission signal, the noise intensity and the Gray code mapping diagram; the symbol signal includes a complex tensor [B, N] or a real tensor [B, N, 2] format of constellation points, B is the batch size of the transmission data, N is a complex value saved in complex form, and the value of N depends on the transmission data of the sender.

3. The HARQ retransmission decision method according to claim 1, characterized in that: Inputting the LLR value matrix into the trained HARQ network module and outputting the retransmission probability specifically includes: The trained HARQ network module includes an input layer, a feature extraction network and a decision network; the feature extraction network includes a first fully connected layer, a first activation layer, a first batch of normalization layers and a first elimination layer connected in sequence; the decision network includes a feature concatenation layer, a second fully connected layer, a second activation layer, a second batch of normalization layers, a second elimination layer, a third fully connected layer, a third activation layer and a normalization layer connected in sequence; Inputting the LLR value matrix into the feature extraction network to extract feature vectors; Calculate the retransmission penalty factor based on the maximum allowed retransmission times and the current transmission times; The feature vector, SNR value and retransmission penalty factor are input into the decision network, and are concatenated to output the retransmission probability.

4. The HARQ retransmission decision method according to claim 1, characterized in that: According to the retransmission probability and the set retransmission threshold, deciding whether the received transmission signal needs to be retransmitted specifically includes: Determine whether the retransmission probability is greater than or equal to a set retransmission threshold; If yes, it determines that the transmission signal is unreliable, sends a NACK signal to the sender, and triggers a retransmission mechanism; If not, it is determined that the transmission signal is reliable and an ACK signal is sent to the sending method.

5. The HARQ retransmission decision method according to claim 1, characterized in that: Determine the final received transmission signal based on the retransmission probabilities of multiple retransmissions, specifically including: Determining a weighted weight of a transmission signal for each retransmission according to the retransmission probabilities of the multiple retransmissions; Normalizing the weighted weights to determine normalized weights; The final received transmission signal is determined according to each retransmitted transmission signal and the corresponding normalized weight.

6. The HARQ retransmission decision method according to claim 1, characterized in that: The network loss is calculated based on the current number of transmissions, the maximum number of retransmissions, and the SNR value of each transmission signal, including: use Calculate network loss; where, is the loss function; n is the current number of transmissions; Nmax is the maximum number of retransmissions; N is the actual number of retransmissions; p n is the predicted probability of the nth transmission; σ(·) is the sigmoid function; SNR n is the signal-to-noise ratio of the nth transmission.

7. A HARQ retransmission decision device, characterized in that: The HARQ retransmission decision device comprises: An LLR value matrix determination module is used to perform LLR calculation on the transmission signal received by the receiving party based on the constellation point coordinates, the constellation point and the Gray code mapping diagram, and the SNR value of the transmission signal to determine the LLR value matrix; A retransmission probability output module, used to input the LLR value matrix into the trained HARQ network module and output the retransmission probability; the HARQ network module is constructed based on the constellation point coordinates, the constellation points and the Gray code mapping diagram; A judgment module, used to decide whether the received transmission signal needs to be retransmitted according to the retransmission probability and a set retransmission threshold; A finally received transmission signal determination module, used to determine the finally received transmission signal according to the retransmission probability of multiple retransmissions; A network loss calculation module is used to calculate the network loss based on the current number of transmissions, the set maximum number of retransmissions, and the SNR value of each transmission signal; The training module is used to train the HARQ network module based on the network loss and determine the trained HARQ network module.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the HARQ retransmission decision method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the HARQ retransmission decision method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the HARQ retransmission decision method according to any one of claims 1 to 6 is implemented.

Images