Channel adaptive hybrid automatic request retransmission method and system

Abstract

The invention discloses a channel adaptive hybrid automatic request retransmission method and system. The method comprises the following steps: training a recurrent neural network prediction model by using a self-supervision mechanism to obtain a channel quality prediction model; predicting channel quality parameters based on the channel quality prediction model; comparing the prediction result of the channel quality parameter with a preset threshold value, judging the channel quality according to the comparison result, and determining the coding and transmission mode of the data according to the channel quality for transmission; according to the method, the coding and transmission modes of the data can be determined according to the prediction result of the channel quality, so that the requirements of high reliability and low delay are met.

Description

Technical Field

[0001] The present invention relates to the technical field of wireless communication of Internet of Things, and in particular to a channel adaptive hybrid automatic request retransmission method and system. Background Art

[0002] With the rapid development of semiconductor technology, computer technology and wireless communication technology, the Internet of Things has been widely used in many fields such as national defense security, industrial control, agricultural production, etc. In the field of wireless communication in the substation of smart grid, the Internet of Things technology also plays an important role. In recent years, a large number of distributed power sources, electric vehicles, and energy storage have been connected to the power grid, and the grid dispatching and operation mode has begun to shift to coordinated control of source, grid, load and storage, which has spawned distributed photovoltaic regulation, precise load control and other control-related services. Such services require high-reliability and low-latency communication technology to carry them, which puts higher requirements on the wireless communication technology in the substation.

[0003] Existing wireless communication technologies rely on mechanisms such as channel coding and hybrid automatic retransmission requests to effectively improve transmission reliability. However, due to the limited processing and communication capabilities of IoT terminals, these mechanisms inevitably increase processing and communication delays. On the one hand, for wireless terminals with very high channel quality, non-coded transmission can already achieve good reliability. Coded transmission not only fails to improve reliability, but also increases processing delays. On the other hand, for wireless terminals with poor channel quality, the data after channel coding is still likely to need to be retransmitted, which will further increase the delay. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to provide a channel adaptive hybrid automatic retransmission request method and system that can reduce processing and communication delays.

[0005] Technical solution: A channel adaptive hybrid automatic request retransmission method according to the present invention comprises:

[0006] The channel quality prediction model is obtained by training the recurrent neural network prediction model using a self-supervision mechanism;

[0007] Predicting channel quality parameters based on a channel quality prediction model;

[0008] The prediction result of the channel quality parameter is compared with the preset threshold, the channel quality is determined according to the comparison result, and the data encoding and transmission mode are determined according to the channel quality for transmission.

[0009] Furthermore, the method of using a self-supervision mechanism to train a recurrent neural network prediction model to obtain a channel quality prediction model includes:

[0010] Obtain long-term historical channel quality data in the target scenario, and use the historical channel quality data to construct training samples for the recurrent neural network prediction model;

[0011] The sequence of channel quality history data of several consecutive previous moments in the training sample is used as input, and the channel quality history data of the next moment is used as output, and the recurrent neural network prediction model is trained using a self-supervision mechanism;

[0012] During the training process of the recurrent neural network prediction model, the cost function of the recurrent neural network prediction model is obtained in each iteration process. When the cost function of the recurrent neural network prediction model is less than a preset threshold, the training is terminated to obtain a channel quality prediction model.

[0013] Furthermore, the self-supervision mechanism includes the following settings:

[0014] Calculate the supervision signal, which is used to supervise the training process of the recurrent neural network prediction model. The supervision signal calculation steps are as follows:

[0015] Calculated based on the current channel quality parameter value:

[0016] E b / N0=f1(CQ(t))

[0017] Among them, E b is the signal energy per bit, N0 is the noise power density, f1 is used to map the channel quality parameter CQ(t) at the current time t to the signal-to-noise ratio, and f1 is determined according to the channel quality type to be predicted;

[0018] According to E b / N0 calculates the bit error rate:

[0019] BER = f2(E b / N0)

[0020] Where BER is the bit error rate, f2 is the bit error rate function, and the bit error rate function is determined according to the modulation / demodulation method used;

[0021] Calculate the packet reception rate:

[0022] PRR = (1-BER) Nbit

[0023] Among them, PRR is the packet reception rate, Nbit is the number of bits in the packet;

[0024] Calculate the supervision signal:

[0025] e=PRR

[0026] Among them, e is the supervision signal.

[0027] Furthermore, the training of the recurrent neural network prediction model includes:

[0028] Initialize the recurrent neural network prediction model, specifically:

[0029] Set the number of input layer neurons in the recurrent neural network prediction model to p1, the number of hidden layer neurons to p2, and the number of output layer neurons to p3. The number of input layer neurons and the number of output layer neurons are determined by the number of inputs and the number of outputs. The number of hidden layer neurons p2 can be calculated as follows:

[0030]

[0031] Among them, a and b are parameters;

[0032] Initialize the weights, specifically setting the weights from the input layer neurons to the hidden layer neurons to v ij , represents the weight from the i-th input layer neuron to the j-th hidden layer neuron; the weight from the hidden layer neuron to the output layer neuron is ω jk , represents the weight from the jth hidden layer neuron to the kth output layer neuron; the weight from the previous hidden layer neuron to the current hidden layer neuron is u jl , represents the weight from the lth hidden layer neuron at the previous moment to the jth hidden layer neuron at the current moment; the Sigmoid function is selected as the activation function of the hidden layer neuron;

[0033] Initialize the neuron nodes and set the jth hidden layer neuron to:

[0034] h(j)=0

[0035] Initialize the kth output layer neuron as:

[0036] y(k)=0

[0037] The target cost function is set as follows:

[0038]

[0039] Among them, e is the supervision signal, p is the model output, H is the binary cross entropy, α is the penalty coefficient, W is the parameter set of the neural network used, ⊙ is the XOR logical operation, and the binary cross entropy formula is as follows:

[0040]

[0041] Among them, p is the probability value vector, p m is the mth element of the probability value vector, ranging from [0,1]; y is the discrete label vector, y mis the mth element of the discrete label vector, with a value of 0 or 1, and n is the length of the vector;

[0042] The training samples are input into the model, and the output of the output layer neurons is calculated as follows:

[0043] The input to the jth hidden layer neuron is:

[0044]

[0045] The output of the jth hidden layer neuron is:

[0046] h j (t) = f(net j (t))

[0047] The input to the kth output layer neuron is:

[0048]

[0049] The output of the kth output layer neuron is:

[0050] y k (t) = f(net k (t))

[0051] Where, f represents the Sigmoid function;

[0052] Set the cost back propagation to the gradient descent algorithm, the learning rate is η, η>0, the momentum factor is λ, 0<λ<1, and the update formula of the connection weight is as follows:

[0053]

[0054] in:

[0055]

[0056] Furthermore, predicting the channel quality parameter based on the channel quality prediction model includes:

[0057] Construct an empty channel quality parameter sequence CQ(N)={};

[0058] Based on the time window T W The channel quality parameters are counted to obtain N channel quality parameter sequences [CQ(t-(N-1)×T W ),...,CQ(tT W ),CQ(t)], t is the current time;

[0059] The channel quality parameter sequence is input into the channel quality prediction model to calculate the predicted value of the channel quality at the next moment

[0060] Further, comparing the prediction result of the channel quality parameter with a preset threshold, determining the channel quality according to the comparison result, and determining the encoding and transmission mode of the data for transmission according to the channel quality, includes:

[0061] Setting a preset threshold Th1 and a preset threshold Th2 for determining a corresponding channel quality parameter, wherein Th1 is greater than Th2;

[0062] When the channel quality parameter prediction value When the channel quality is determined to be high, the data is directly transmitted after verification encoding;

[0063] When the channel quality parameter prediction value When the channel quality is judged to be average, the hybrid automatic request repeat transmission method is used to perform error correction coding on the data before transmission;

[0064] When the quality parameter prediction value When the channel quality is determined to be poor, a hybrid automatic request retransmission transmission method with repetition coding is used to encode and transmit the data.

[0065] Further, the data is directly transmitted after verification encoding, including:

[0066] The sender verifies and encodes the data before sending it;

[0067] When the receiving end successfully decodes, the receiving end sends a confirmation signal to the sending end to indicate that it has been received correctly and no retransmission is required; when the receiving end fails to decode, the receiving end sends an error signal to the sending end to indicate that the reception has failed, and the sending end repeats sending the corresponding data.

[0068] Furthermore, the hybrid automatic repeat request transmission method is used to transmit data after error correction coding, including:

[0069] The sender performs error correction encoding on the data before sending it;

[0070] The receiving end receives the code and caches it, and decodes it according to the received error correction code. After the receiving end successfully decodes, it sends a confirmation signal to the sending end to indicate that it is correctly received and no longer needs to be retransmitted; after the receiving end fails to decode, the receiving end sends an error signal to the sending end to indicate that the reception failed, and the sending end repeats the corresponding data.

[0071] Furthermore, the hybrid automatic request retransmission transmission method using repetition coding encodes and transmits data, including:

[0072] The sending end performs error correction coding on the data;

[0073] The sender copies the same code and sends it together;

[0074] The receiving end receives the code and caches it; after successfully decoding the first code, the receiving end discards the repeated code and sends a confirmation signal to the sending end to indicate that it is correctly received and no longer needs to be retransmitted; after the receiving end fails to decode the first code, the receiving end performs a Zeiss merger on the two repeated codes; after successfully decoding the merged two codes, the receiving end sends a confirmation signal to the sending end to indicate that it is correctly received and no longer needs to be retransmitted; after failing to decode the merged two codes, the receiving end sends an error signal to the sending end to indicate that the reception failed, and the sending end repeatedly sends the corresponding data.

[0075] Based on the same inventive concept, the present invention also provides a channel adaptive hybrid automatic request retransmission system, comprising:

[0076] A training unit, used for training the recurrent neural network prediction model using a self-supervision mechanism to obtain a channel quality prediction model;

[0077] A prediction unit, configured to predict a channel quality parameter based on a channel quality prediction model;

[0078] The transmission unit is used to compare the prediction result of the channel quality parameter with a preset threshold, determine the channel quality according to the comparison result, and determine the encoding and transmission mode of the data for transmission according to the channel quality.

[0079] Furthermore, the method of using a self-supervision mechanism to train a recurrent neural network prediction model to obtain a channel quality prediction model includes:

[0080] Obtain long-term historical channel quality data in the target scenario, and use the historical channel quality data to construct training samples for the recurrent neural network prediction model;

[0081] The sequence of channel quality history data of several consecutive previous moments in the training sample is used as input, and the channel quality history data of the next moment is used as output, and the recurrent neural network prediction model is trained using a self-supervision mechanism;

[0082] During the training process of the recurrent neural network prediction model, the cost function of the recurrent neural network prediction model is obtained in each iteration process. When the cost function of the recurrent neural network prediction model is less than a preset threshold, the training is terminated to obtain a channel quality prediction model.

[0083] Furthermore, the self-supervision mechanism includes the following settings:

[0084] Calculate the supervision signal, which is used to supervise the training process of the recurrent neural network prediction model. The supervision signal calculation steps are as follows:

[0085] Calculated based on the current channel quality parameter value:

[0086] E b / N0=f1(CQ(t))

[0087] Among them, E b is the signal energy per bit, N0 is the noise power density, f1 is used to map the channel quality parameter CQ(t) at the current time t to the signal-to-noise ratio, and f1 is determined according to the channel quality type to be predicted;

[0088] According to E b / N0 calculates the bit error rate:

[0089] BER = f2(E b / N0)

[0090] Where BER is the bit error rate, f2 is the bit error rate function, and the bit error rate function is determined according to the modulation / demodulation method used;

[0091] Calculate the packet reception rate:

[0092] PRR = (1-BER) Nbit

[0093] Among them, PRR is the packet reception rate, Nbit is the number of bits in the packet;

[0094] Calculate the supervision signal:

[0095] e=PRR

[0096] Among them, e is the supervision signal.

[0097] Furthermore, the training of the recurrent neural network prediction model includes:

[0098] Initialize the recurrent neural network prediction model, specifically:

[0099] Set the number of input layer neurons in the recurrent neural network prediction model to p1, the number of hidden layer neurons to p2, and the number of output layer neurons to p3. The number of input layer neurons and the number of output layer neurons are determined by the number of inputs and the number of outputs. The number of hidden layer neurons p2 can be calculated as follows:

[0100]

[0101] Among them, a and b are parameters;

[0102] Initialize the weights, specifically setting the weights from the input layer neurons to the hidden layer neurons to v ij , represents the weight from the i-th input layer neuron to the j-th hidden layer neuron; the weight from the hidden layer neuron to the output layer neuron is ω jk, represents the weight from the jth hidden layer neuron to the kth output layer neuron; the weight from the previous hidden layer neuron to the current hidden layer neuron is u jl , represents the weight from the lth hidden layer neuron at the previous moment to the jth hidden layer neuron at the current moment; the Sigmoid function is selected as the activation function of the hidden layer neuron;

[0103] Initialize the neuron nodes and set the jth hidden layer neuron to:

[0104] h(j)=0

[0105] Initialize the kth output layer neuron as:

[0106] y(k)=0

[0107] The target cost function is set as follows:

[0108]

[0109] Among them, e is the supervision signal, p is the model output, H is the binary cross entropy, α is the penalty coefficient, W is the parameter set of the neural network used, ⊙ is the XOR logical operation, and the binary cross entropy formula is as follows:

[0110]

[0111] Among them, p is the probability value vector, p m is the mth element of the probability value vector, ranging from [0,1]; y is the discrete label vector, y m is the mth element of the discrete label vector, with a value of 0 or 1, and n is the length of the vector;

[0112] The training samples are input into the model, and the output of the output layer neurons is calculated as follows:

[0113] The input to the jth hidden layer neuron is:

[0114]

[0115] The output of the jth hidden layer neuron is:

[0116] h j (t) = f(net j (t))

[0117] The input to the kth output layer neuron is:

[0118]

[0119] The output of the kth output layer neuron is:

[0120] yk (t) = f(net k (t))

[0121] Where, f represents the Sigmoid function;

[0122] Set the cost back propagation to the gradient descent algorithm, the learning rate is η, η>0, the momentum factor is λ, 0<λ<1, and the update formula of the connection weight is as follows:

[0123]

[0124] in:

[0125]

[0126] Furthermore, predicting the channel quality parameter based on the channel quality prediction model includes:

[0127] Construct an empty channel quality parameter sequence CQ(N)={};

[0128] Based on the time window T W The channel quality parameters are counted to obtain N channel quality parameter sequences [CQ(t-(N-1)×T W ),...,CQ(tT W ),CQ(t)], t is the current time;

[0129] The channel quality parameter sequence is input into the channel quality prediction model to calculate the predicted value of the channel quality at the next moment

[0130]

[0131] Further, comparing the prediction result of the channel quality parameter with a preset threshold, determining the channel quality according to the comparison result, and determining the encoding and transmission mode of the data for transmission according to the channel quality, includes:

[0132] Setting a preset threshold Th1 and a preset threshold Th2 for determining a corresponding channel quality parameter, wherein Th1 is greater than Th2;

[0133] When the channel quality parameter prediction value When the channel quality is determined to be high, the data is directly transmitted after verification encoding;

[0134] When the channel quality parameter prediction value When the channel quality is judged to be average, the hybrid automatic request repeat transmission method is used to perform error correction coding on the data before transmission;

[0135] When the channel quality parameter prediction value When the channel quality is determined to be poor, a hybrid automatic request retransmission transmission method with repetition coding is used to encode and transmit the data.

[0136] Further, the data is directly transmitted after verification encoding, including:

[0137] The sender verifies and encodes the data before sending it;

[0138] When the receiving end successfully decodes, the receiving end sends a confirmation signal to the sending end to indicate that it has been received correctly and no retransmission is required; when the receiving end fails to decode, the receiving end sends an error signal to the sending end to indicate that the reception has failed, and the sending end repeats sending the corresponding data.

[0139] Furthermore, the hybrid automatic repeat request transmission method is used to transmit data after error correction coding, including:

[0140] The sender performs error correction encoding on the data before sending it;

[0141] The receiving end receives the code and caches it, and decodes it according to the received error correction code; after the receiving end successfully decodes, it sends a confirmation signal to the sending end to indicate that it has been correctly received and no longer needs to be retransmitted; after the receiving end fails to decode, the receiving end sends an error signal to the sending end to indicate that the reception has failed, and the sending end repeats sending the corresponding data.

[0142] Furthermore, the hybrid automatic request retransmission transmission method using repetition coding encodes and transmits data, including:

[0143] The sending end performs error correction coding on the data;

[0144] The sender copies the same code and sends it together;

[0145] The receiving end receives the code and caches it; after successfully decoding the first code, the receiving end discards the repeated code and sends a confirmation signal to the sending end to indicate that it is correctly received and no longer needs to be retransmitted; after the receiving end fails to decode the first code, the receiving end performs a Zeiss merger on the two repeated codes; after successfully decoding the merged two codes, the receiving end sends a confirmation signal to the sending end to indicate that it is correctly received and no longer needs to be retransmitted; after failing to decode the merged two codes, the receiving end sends an error signal to the sending end to indicate that the reception failed, and the sending end repeatedly sends the corresponding data.

[0146] Based on the same inventive concept, the present invention also provides a computer device, comprising: one or more processors; the processor is used to store one or more programs; when the one or more programs are executed by the one or more processors, the channel adaptive hybrid automatic request retransmission method according to any one of the above items is implemented.

[0147] Based on the same inventive concept, the present invention further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed, the channel-adaptive hybrid automatic retransmission request method according to any one of the above items is implemented.

[0148] Beneficial effects: Compared with the prior art, the present invention adopts a channel-adaptive hybrid automatic request retransmission method for high-reliability, low-latency services, and schedules the adopted transmission mode according to the prediction results of channel quality, thereby minimizing processing and communication delays while meeting reliability requirements; the method proposed in the present invention first predicts channel quality parameters based on a self-supervised recurrent neural network to obtain an evaluation of the channel quality; secondly, based on the prediction results of the channel quality, the data encoding and transmission methods are determined, thereby achieving the effect of high reliability and low latency. BRIEF DESCRIPTION OF THE DRAWINGS

[0149] Figure 1 is a flow chart of a method according to an embodiment of the present invention;

[0150] Figure 2 A schematic diagram of self-supervised training and prediction of a channel quality prediction model according to an embodiment of the present invention;

[0151] Figure 3 A schematic diagram of a transmission mode decision and encoding process according to an embodiment of the present invention;

[0152] Figure 4 A schematic diagram of the probability of first transmission success of different methods according to an embodiment of the present invention;

[0153] Figure 5 A schematic diagram of average retransmission times of different methods according to an embodiment of the present invention;

[0154] Figure 6 FIG. 4 is a schematic diagram of frame error rates of different methods according to an embodiment of the present invention. DETAILED DESCRIPTION

[0155] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0156] As attached Figure 1 As shown, a channel adaptive hybrid automatic request retransmission method of this embodiment includes:

[0157] S1: Use the self-supervision mechanism to train the recurrent neural network prediction model to obtain the channel quality prediction model;

[0158] S2: predicting channel quality parameters based on a channel quality prediction model;

[0159] S3: Compare the prediction result of the channel quality parameter with a preset threshold, determine the channel quality according to the comparison result, and determine the encoding and transmission mode of the data for transmission according to the channel quality.

[0160] Specifically, this embodiment selects the signal to noise ratio (SNR) as the channel quality parameter for prediction.

[0161] The step S1 of using a self-supervisory mechanism to train the recurrent neural network prediction model to obtain a channel quality prediction model includes:

[0162] Obtain long-term channel quality measurement data in the target scenario and construct training samples for the recurrent neural network prediction model;

[0163] Taking N consecutive channel quality history sequences as input and the channel quality at the next moment as output, the recurrent neural network prediction model is trained using a self-supervised method.

[0164] When the cost function of the prediction model meets the requirements, the training is ended and the prediction model parameters are obtained.

[0165] The process of the operation phase of the channel quality prediction model in step S2 includes:

[0166] Construct an empty signal-to-noise ratio parameter sequence S(N)={};

[0167] Every time window T W The SNR parameters are counted to obtain a sequence containing N SNR parameters [SNR(t-(N-1)×T W ),...S,NR(tT W )S,NR(t)], where t is the current time;

[0168] Calculate the SNR prediction value at the next moment based on the channel quality measurement model

[0169] Step S1 includes training of a channel quality prediction model for SNR, with the following initial settings:

[0170] Calculate the supervision signal: This supervision signal is used to supervise the model training process. The calculation steps are as follows: First, calculate E according to the SNR prediction value b / N0:

[0171]

[0172] Among them, E b is the signal energy per bit, N0 is the noise power density, R is the symbol rate, and BW is the channel bandwidth;

[0173] Then, according to E b / N0 calculates the bit error rate:

[0174]

[0175] Where BER is the bit error rate, the signal modulation method is DSSS-OQPSK, and Q is the complementary cumulative distribution function;

[0176] Next, calculate the packet reception rate:

[0177] PRR = (1-BER) Nbit

[0178] Among them, PRR is the packet reception rate, Nbit is the number of bits;

[0179] Finally, the supervision signal is calculated:

[0180] e=PRR

[0181] Among them, e is the supervision signal, and PRR is used as the supervision signal;

[0182] Parameter setting: Let the number of neurons in the input layer be p1, the number of neurons in the hidden layer be p2, and the number of neurons in the output layer be p3. The number of neurons in the input layer and the output layer is determined by the number of inputs and outputs. The number of neurons in the hidden layer, p2, can be calculated as follows:

[0183]

[0184] The target cost function is set as follows:

[0185]

[0186] Among them, e is the supervision signal, p is the model output, H is the binary cross entropy, λ is the penalty coefficient, W is the parameter set of the neural network used, and ⊙ is the XOR logical operation. The binary cross entropy formula is as follows:

[0187]

[0188] Among them, p is the probability value vector, p m is the mth element of the probability value vector, ranging from [0,1]. y is the discrete label vector, y m It is the mth element of the discrete label vector, with a value of 0 or 1, and n is the length of the vector.

[0189] Weight initialization: Set the weight from the input layer to the hidden layer to v ij , represents the weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer; the weight from the hidden layer to the output layer is ω jk, represents the weight from the jth neuron in the hidden layer to the kth neuron in the output layer; the weight from the hidden layer at the previous moment to the hidden layer at the current moment is u jl , represents the weight from the lth neuron in the hidden layer at the previous moment to the jth neuron in the hidden layer at the current moment; the Sigmoid function is selected as the activation function of the hidden layer;

[0190] Neuron node initialization, set the jth hidden layer neuron initialization setting to:

[0191] h(j)=0

[0192] The kth output layer neuron is initialized as:

[0193] y(k)=0

[0194] Step S1 includes training a channel quality prediction model for SNR:

[0195] Set training threshold;

[0196] The training samples are input into the model, and the output of the output layer neurons is calculated as follows:

[0197] The input to the jth hidden layer neuron is:

[0198]

[0199] The output of the jth hidden layer neuron is:

[0200] h j (t) = f(net j (t))

[0201] The input to the kth output layer neuron is:

[0202]

[0203] The output of the kth output layer neuron is:

[0204] y k (t) = f(net k (t))

[0205] Where f represents the Sigmoid function

[0206] Set the cost back propagation to the gradient descent algorithm, the learning rate to η (η>0), and the momentum factor to λ (0<λ<1) to prevent the prediction model from falling into the local optimum. The update formula of the connection weight is as follows:

[0207]

[0208] in:

[0209]

[0210] The self-supervised training and prediction of the channel quality prediction model of this embodiment are as follows Figure 2 shown.

[0211] Determining the data encoding and transmission mode according to the channel quality in step S3 includes:

[0212] Determine the quality of the current channel and confirm the transmission mode. The specific steps are as follows:

[0213] When the signal-to-noise ratio prediction value When the channel quality is high, the data is directly transmitted after verification encoding, as shown in the schematic diagram. Figure 3 As shown in the figure, Th1 is a threshold of SNR corresponding to a lower bit error rate. The specific transmission steps are as follows:

[0214] 1) The sender performs a simple checksum encoding on the data before sending it;

[0215] 2) When the receiving end successfully decodes, the receiving end sends a confirmation signal to the sending end to indicate that the data is received correctly and no retransmission is required; when the receiving end fails to decode, the receiving end sends an error signal to the sending end to indicate that the data is received unsuccessfully and the sending end repeats the sending of the corresponding data.

[0216] When the signal-to-noise ratio prediction value When the channel quality is average, the hybrid automatic request retransmission method is used to transmit the data after error correction coding, as shown in the schematic diagram. Figure 3 The specific transmission steps are as follows:

[0217] 1) The sending end performs error correction encoding on the data before sending it;

[0218] 2) The receiving end receives the code and caches it, and decodes it according to the received error correction code. After the receiving end successfully decodes, it sends a confirmation signal to the sending end to indicate that it has been received correctly and no retransmission is required; after the receiving end fails to decode, it sends an error signal to the sending end to indicate that the reception has failed, and the sending end repeats the corresponding data.

[0219] When the signal-to-noise ratio prediction value When the channel quality is poor, the hybrid automatic request retransmission transmission method with repeated coding is used to encode and transmit the data, as shown in the schematic diagram. Figure 3 The specific transmission steps are as follows:

[0220] 1) The sender performs error correction coding on the data;

[0221] 2) The sender copies the same code and sends it together;

[0222] 3) The receiving end receives the code and caches it. After the receiving end successfully decodes the first code segment, it discards the repeated code segment and sends a confirmation signal to the sending end to indicate that it is correctly received and no longer needs to be retransmitted; after the receiving end fails to decode the first code segment, the receiving end performs Zeiss merging of the two repeated codes; after successfully decoding the two combined codes, the receiving end sends a confirmation signal to the sending end to indicate that it is correctly received and no longer needs to be retransmitted; after failing to decode the two combined codes, the receiving end sends an error signal to the sending end to indicate that the reception failed, and the sending end repeats the corresponding data.

[0223] In order to verify the reliability of the proposed method, polar code is used as the error correction code for verification. The specific steps are as follows:

[0224] The specific steps for constructing polarization codes for polarization channels are as follows:

[0225] For length N = 2 n The polar code of (n≥0) is calculated by the following formula:

[0226]

[0227] in, is the bit sequence after polar code encoding, is the uncoded bit sequence, G N is the generated sequence. N Calculated by the following formula:

[0228]

[0229] Among them, B N is the bit reversal of the uncoded bit sequence, is the nth Kronecker power of F2. Among them,

[0230] For the transfer probability of polarization code of polarization channel, the specific steps are as follows:

[0231] For a specific modulation and demodulation mode, there is a fixed relationship between the SNR and the bit error rate of channel W:

[0232]

[0233] Among them, B M (·) is the bit error rate function of a specific modulation and demodulation mode, E b is the signal energy per bit, N0 is the noise power density, R b is the amount of data transmission, B N is the noise bandwidth.

[0234] For length N = 2 nThe channel capacities of each polarization channel are:

[0235]

[0236]

[0237] Among them, 2 n =1,

[0238] In order to verify the reliability of the proposed method, the following simulation settings are performed: the polar code length is set to 64 bits, the code rate is set to 0.5, the maximum number of retransmissions is set to two, and the modulation and demodulation mode is set to BPSK. It is noted that if more than one bit is wrong in a frame of data (64 bits), it cannot be successfully decoded. The probability of successful first transmission, the number of data retransmissions, and the frame error rate of the proposed method, the existing BPSK modulation and demodulation method, and the existing polar code encoding are obtained as follows: Figure 4 , Figure 5 and Figure 6 shown.

[0239] It can be seen that polar code coded transmission has a gain of about 10dB compared to direct transmission, while the proposed method has a gain of about 4dB compared to polar code transmission. The proposed method is superior to the existing methods in the probability of successful first transmission, the number of data retransmissions, and the frame error rate.

[0240] Based on the same inventive concept, this embodiment also provides a channel adaptive hybrid automatic retransmission request system, including:

[0241] A training unit, used for training the recurrent neural network prediction model using a self-supervision mechanism to obtain a channel quality prediction model;

[0242] A prediction unit, configured to predict a channel quality parameter based on a channel quality prediction model;

[0243] The transmission unit is used to compare the prediction result of the channel quality parameter with a preset threshold, determine the channel quality according to the comparison result, and determine the encoding and transmission mode of the data for transmission according to the channel quality.

[0244] Furthermore, the method of using a self-supervision mechanism to train a recurrent neural network prediction model to obtain a channel quality prediction model includes:

[0245] Obtain long-term historical channel quality data in the target scenario, and use the historical channel quality data to construct training samples for the recurrent neural network prediction model;

[0246] The sequence of channel quality history data of several consecutive previous moments in the training sample is used as input, and the channel quality history data of the next moment is used as output, and the recurrent neural network prediction model is trained using a self-supervision mechanism;

[0247] During the training process of the recurrent neural network prediction model, the cost function of the recurrent neural network prediction model is obtained in each iteration process. When the cost function of the recurrent neural network prediction model is less than a preset threshold, the training is terminated to obtain a channel quality prediction model.

[0248] Furthermore, the self-supervision mechanism includes the following settings:

[0249] Calculate the supervision signal, which is used to supervise the training process of the recurrent neural network prediction model. The supervision signal calculation steps are as follows:

[0250] Calculated based on the current channel quality parameter value:

[0251] E b / N0=f1(CQ(t))

[0252] Among them, E b is the signal energy per bit, N0 is the noise power density, f1 is used to map the channel quality parameter CQ(t) at the current time t to the signal-to-noise ratio, and f1 is determined according to the channel quality type to be predicted;

[0253] According to E b / N0 calculates the bit error rate:

[0254] BER = f2(E b / N0)

[0255] Where BER is the bit error rate, f2 is the bit error rate function, and the bit error rate function is determined according to the modulation / demodulation method used;

[0256] Calculate the packet reception rate:

[0257] PRR = (1-BER) Nbit

[0258] Among them, PRR is the packet reception rate, Nbit is the number of bits in the packet;

[0259] Calculate the supervision signal:

[0260] e=PRR

[0261] Among them, e is the supervision signal.

[0262] Furthermore, the training of the recurrent neural network prediction model includes:

[0263] Initialize the recurrent neural network prediction model, specifically:

[0264] Set the number of input layer neurons in the recurrent neural network prediction model to p1, the number of hidden layer neurons to p2, and the number of output layer neurons to p3. The number of input layer neurons and the number of output layer neurons are determined by the number of inputs and the number of outputs. The number of hidden layer neurons p2 can be calculated as follows:

[0265]

[0266] Among them, a and b are parameters;

[0267] Initialize the weights, specifically setting the weights from the input layer neurons to the hidden layer neurons to v ij , represents the weight from the i-th input layer neuron to the j-th hidden layer neuron; the weight from the hidden layer neuron to the output layer neuron is ω jk , represents the weight from the jth hidden layer neuron to the kth output layer neuron; the weight from the previous hidden layer neuron to the current hidden layer neuron is u jl , represents the weight from the lth hidden layer neuron at the previous moment to the jth hidden layer neuron at the current moment; the Sigmoid function is selected as the activation function of the hidden layer neuron;

[0268] Initialize the neuron nodes and set the jth hidden layer neuron to:

[0269] h(j)=0

[0270] Initialize the kth output layer neuron as:

[0271] y(k)=0

[0272] The target cost function is set as follows:

[0273]

[0274] Among them, e is the supervision signal, p is the model output, H is the binary cross entropy, α is the penalty coefficient, W is the parameter set of the neural network used, ⊙ is the XOR logical operation, and the binary cross entropy formula is as follows:

[0275]

[0276] Among them, p is the probability value vector, p m is the mth element of the probability value vector, ranging from [0,1]; y is the discrete label vector, y m is the mth element of the discrete label vector, with a value of 0 or 1, and n is the length of the vector;

[0277] The training samples are input into the model, and the output of the output layer neurons is calculated as follows:

[0278] The input to the jth hidden layer neuron is:

[0279]

[0280] The output of the jth hidden layer neuron is:

[0281] h j (t) = f(net j (t))

[0282] The input to the kth output layer neuron is:

[0283]

[0284] The output of the kth output layer neuron is:

[0285] y k (t) = f(net k (t))

[0286] Where, f represents the Sigmoid function;

[0287] Set the cost back propagation to the gradient descent algorithm, the learning rate is η, η>0, the momentum factor is λ, 0<λ<1, and the update formula of the connection weight is as follows:

[0288]

[0289] in:

[0290]

[0291] Furthermore, predicting the channel quality parameter based on the channel quality prediction model includes:

[0292] Construct an empty channel quality parameter sequence CQ(N)={};

[0293] Based on the time window T W The channel quality parameters are counted to obtain N channel quality parameter sequences [CQ(t-(N-1)×T W ),...,CQ(tT W ),CQ(t)], t is the current time;

[0294] The channel quality parameter sequence is input into the channel quality prediction model to calculate the predicted value of the channel quality at the next moment

[0295]

[0296] Further, comparing the prediction result of the channel quality parameter with a preset threshold, determining the channel quality according to the comparison result, and determining the encoding and transmission mode of the data for transmission according to the channel quality, includes:

[0297] Setting a preset threshold Th1 and a preset threshold Th2 for determining a corresponding channel quality parameter, wherein Th1 is greater than Th2;

[0298] When the channel quality parameter prediction value When the channel quality is determined to be high, the data is directly transmitted after verification encoding;

[0299] When the channel quality parameter prediction value When the channel quality is judged to be average, the hybrid automatic request repeat transmission method is used to perform error correction coding on the data before transmission;

[0300] When the channel quality parameter prediction value When the channel quality is determined to be poor, a hybrid automatic request retransmission transmission method with repetition coding is used to encode and transmit the data.

[0301] Further, the data is directly transmitted after verification encoding, including:

[0302] The sender verifies and encodes the data before sending it;

[0303] When the receiving end successfully decodes, the receiving end sends a confirmation signal to the sending end to indicate that it has been received correctly and no retransmission is required; when the receiving end fails to decode, the receiving end sends an error signal to the sending end to indicate that the reception has failed, and the sending end repeats sending the corresponding data.

[0304] Furthermore, the hybrid automatic repeat request transmission method is used to transmit data after error correction coding, including:

[0305] The sender performs error correction encoding on the data before sending it;

[0306] The receiving end receives the code and caches it, and decodes it according to the received error correction code; after the receiving end successfully decodes, it sends a confirmation signal to the sending end to indicate that it has been correctly received and no longer needs to be retransmitted; after the receiving end fails to decode, the receiving end sends an error signal to the sending end to indicate that the reception has failed, and the sending end repeats sending the corresponding data.

[0307] Furthermore, the hybrid automatic request retransmission transmission method using repetition coding encodes and transmits data, including:

[0308] The sending end performs error correction coding on the data;

[0309] The sender copies the same code and sends it together;

[0310] The receiving end receives the code and caches it; after successfully decoding the first code, the receiving end discards the repeated code and sends a confirmation signal to the sending end to indicate that it is correctly received and no longer needs to be retransmitted; after the receiving end fails to decode the first code, the receiving end performs a Zeiss merger on the two repeated codes; after successfully decoding the merged two codes, the receiving end sends a confirmation signal to the sending end to indicate that it is correctly received and no longer needs to be retransmitted; after failing to decode the merged two codes, the receiving end sends an error signal to the sending end to indicate that the reception failed, and the sending end repeatedly sends the corresponding data.

[0311] Based on the same inventive concept, this embodiment also provides a computer device, including: one or more processors; the processor is used to store one or more programs; when the one or more programs are executed by the one or more processors, the channel adaptive hybrid automatic request retransmission method described in any one of the above items is implemented.

[0312] Based on the same inventive concept, this embodiment further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed, the channel-adaptive hybrid automatic retransmission request method according to any one of the above items is implemented.

Claims

1. A channel adaptive hybrid automatic retransmission request method, characterized in that: include: The channel quality prediction model is obtained by training the recurrent neural network prediction model using a self-supervision mechanism; Predicting channel quality parameters based on a channel quality prediction model; The prediction result of the channel quality parameter is compared with the preset threshold, the channel quality is determined according to the comparison result, and the data encoding and transmission mode are determined according to the channel quality for transmission.

2. The channel adaptive hybrid automatic retransmission request method according to claim 1, characterized in that: The method of using a self-supervisory mechanism to train a recurrent neural network prediction model to obtain a channel quality prediction model includes: Obtain long-term historical channel quality data in the target scenario, and use the historical channel quality data to construct training samples for the recurrent neural network prediction model; The sequence of channel quality history data of several consecutive previous moments in the training sample is used as input, and the channel quality history data of the next moment is used as output, and the recurrent neural network prediction model is trained using a self-supervision mechanism; During the training process of the recurrent neural network prediction model, the cost function of the recurrent neural network prediction model is obtained in each iteration process. When the cost function of the recurrent neural network prediction model is less than a preset threshold, the training is terminated to obtain a channel quality prediction model.

3. The channel adaptive hybrid automatic retransmission request method according to claim 1, characterized in that: The self-supervision mechanism includes the following settings: Calculate the supervision signal, which is used to supervise the training process of the recurrent neural network prediction model. The supervision signal calculation steps are as follows: Calculated based on the current channel quality parameter value: E b / N0=f1(CQ(t)) Among them, E b is the signal energy per bit, N0 is the noise power density, f1 is used to map the channel quality parameter CQ(t) at the current time t to the signal-to-noise ratio, and f1 is determined according to the channel quality type to be predicted; According to E b / N0 calculates the bit error rate: BER=f2(E b / N0) Where BER is the bit error rate, f2 is the bit error rate function, and the bit error rate function is determined according to the modulation / demodulation method used; Calculate the packet reception rate: PRR=(1-BER) Nbit Among them, PRR is the packet reception rate, Nbit is the number of bits in the packet; Calculate the supervision signal: e=PRR Among them, e is the supervision signal.

4. The channel adaptive hybrid automatic retransmission request method according to claim 2, characterized in that: The training of the recurrent neural network prediction model includes: Initialize the recurrent neural network prediction model, specifically: Set the number of input layer neurons in the recurrent neural network prediction model to p1, the number of hidden layer neurons to p2, and the number of output layer neurons to p3. The number of input layer neurons and the number of output layer neurons are determined by the number of inputs and the number of outputs. The number of hidden layer neurons p2 can be calculated as follows: Among them, a and b are parameters; Initialize the weights, specifically setting the weights from the input layer neurons to the hidden layer neurons to v ij , represents the weight from the i-th input layer neuron to the j-th hidden layer neuron; the weight from the hidden layer neuron to the output layer neuron is ω jk , represents the weight from the jth hidden layer neuron to the kth output layer neuron; the weight from the previous hidden layer neuron to the current hidden layer neuron is u jl , represents the weight from the lth hidden layer neuron at the previous moment to the jth hidden layer neuron at the current moment; the Sigmoid function is selected as the activation function of the hidden layer neuron; Initialize the neuron nodes and set the jth hidden layer neuron to: h(j)=0 Initialize the kth output layer neuron as: y(k)=0 The target cost function is set as follows: Among them, e is the supervision signal, p is the model output, H is the binary cross entropy, α is the penalty coefficient, W is the parameter set of the neural network used, ⊙ is the XOR logical operation, and the binary cross entropy formula is as follows: Among them, p is the probability value vector, p m is the mth element of the probability value vector, with a value range of [0,1]; y is the discrete label vector, y m is the mth element of the discrete label vector, with a value of 0 or 1, and n is the length of the vector; The training samples are input into the model, and the output of the output layer neurons is calculated as follows: The input to the jth hidden layer neuron is: The output of the jth hidden layer neuron is: h j (t)=f(net j (t)) The input to the kth output layer neuron is: The output of the kth output layer neuron is: y k (t)=f(net k (t)) Where, f represents the Sigmoid function; Set the cost back propagation to the gradient descent algorithm, the learning rate is η, η>0, the momentum factor is λ, 0<λ<1, and the update formula of the connection weight is as follows: in:

5. The channel adaptive hybrid automatic repeat request method according to claim 1, characterized in that: The channel quality parameters are predicted based on the channel quality prediction model, including: Construct an empty channel quality parameter sequence CQ(N)={}; Based on the time window T W The channel quality parameters are counted to obtain N channel quality parameter sequences [CQ(t-(N-1)×T W ),...,CQ(tT W ),CQ(t)], t is the current time; The channel quality parameter sequence is input into the channel quality prediction model to calculate the predicted value of the channel quality at the next moment 6. The channel adaptive hybrid automatic repeat request method according to claim 1, characterized in that: The method of comparing the prediction result of the channel quality parameter with a preset threshold, determining the channel quality according to the comparison result, and determining the encoding and transmission mode of the data for transmission according to the channel quality includes: Setting a preset threshold Th1 and a preset threshold Th2 for determining a corresponding channel quality parameter, wherein Th1 is greater than Th2; When the channel quality parameter prediction value When the channel quality is determined to be high, the data is directly transmitted after verification encoding; When the channel quality parameter prediction value When the channel quality is judged to be average, the hybrid automatic request repeat transmission method is used to perform error correction coding on the data before transmission; When the channel quality parameter prediction value When the channel quality is determined to be poor, a hybrid automatic request retransmission transmission method with repetition coding is used to encode and transmit the data.

7. The channel adaptive hybrid automatic repeat request method according to claim 6, characterized in that: The step of directly transmitting the data after verifying and encoding the data comprises: The sender verifies and encodes the data before sending it; When the receiving end successfully decodes, the receiving end sends a confirmation signal to the sending end to indicate that it has been received correctly and no retransmission is required; when the receiving end fails to decode, the receiving end sends an error signal to the sending end to indicate that the reception has failed, and the sending end repeats sending the corresponding data.

8. The channel adaptive hybrid automatic repeat request method according to claim 6, characterized in that: The hybrid automatic retransmission request transmission method is used to transmit data after error correction coding, including: The sender performs error correction encoding on the data before sending it; The receiving end receives the code and caches it, and decodes it according to the received error correction code; after the receiving end successfully decodes, it sends a confirmation signal to the sending end to indicate that it has been correctly received and no longer needs to be retransmitted; after the receiving end fails to decode, the receiving end sends an error signal to the sending end to indicate that the reception has failed, and the sending end repeats sending the corresponding data.

9. The channel adaptive hybrid automatic repeat request method according to claim 6, characterized in that: The hybrid automatic request retransmission transmission method using repetition coding encodes and transmits data, comprising: The sending end performs error correction coding on the data; The sender copies the same code and sends it together; The receiving end receives the code and caches it; after successfully decoding the first code, the receiving end discards the repeated code and sends a confirmation signal to the sending end to indicate that it is correctly received and no longer needs to be retransmitted; after the receiving end fails to decode the first code, the receiving end performs a Zeiss merger on the two repeated codes; after successfully decoding the merged two codes, the receiving end sends a confirmation signal to the sending end to indicate that it is correctly received and no longer needs to be retransmitted; after failing to decode the merged two codes, the receiving end sends an error signal to the sending end to indicate that the reception failed, and the sending end repeatedly sends the corresponding data.

10. A channel adaptive hybrid automatic repeat request system, characterized in that: include: A training unit, used for training the recurrent neural network prediction model using a self-supervision mechanism to obtain a channel quality prediction model; A prediction unit, configured to predict a channel quality parameter based on a channel quality prediction model; The transmission unit is used to compare the prediction result of the channel quality parameter with a preset threshold, determine the channel quality according to the comparison result, and determine the encoding and transmission mode of the data for transmission according to the channel quality.

11. The channel adaptive hybrid automatic repeat request system according to claim 10, characterized in that: The method of using a self-supervisory mechanism to train a recurrent neural network prediction model to obtain a channel quality prediction model includes: Obtain long-term historical channel quality data in the target scenario, and use the historical channel quality data to construct training samples for the recurrent neural network prediction model; The sequence of channel quality history data of several consecutive previous moments in the training sample is used as input, and the channel quality history data of the next moment is used as output, and the recurrent neural network prediction model is trained using a self-supervision mechanism; During the training process of the recurrent neural network prediction model, the cost function of the recurrent neural network prediction model is obtained in each iteration process. When the cost function of the recurrent neural network prediction model is less than a preset threshold, the training is terminated to obtain a channel quality prediction model.

12. The channel adaptive hybrid automatic repeat request system according to claim 10, characterized in that: The self-supervision mechanism includes the following settings: Calculate the supervision signal, which is used to supervise the training process of the recurrent neural network prediction model. The supervision signal calculation steps are as follows: Calculated based on the current channel quality parameter value: E b / N0=f1(CQ(t)) Among them, E b is the signal energy per bit, N0 is the noise power density, f1 is used to map the channel quality parameter CQ(t) at the current time t to the signal-to-noise ratio, and f1 is determined according to the channel quality type to be predicted; According to E b / N0 calculates the bit error rate: BER=f2(E b / N0) Where BER is the bit error rate, f2 is the bit error rate function, and the bit error rate function is determined according to the modulation / demodulation method used; Calculate the packet reception rate: PRR=(1-BER) Nbit Among them, PRR is the packet reception rate, Nbit is the number of bits in the packet; Calculate the supervision signal: e=PRR Among them, e is the supervision signal.

13. The channel adaptive hybrid automatic repeat request system according to claim 11, characterized in that: The training of the recurrent neural network prediction model includes: Initialize the recurrent neural network prediction model, specifically: Set the number of input layer neurons in the recurrent neural network prediction model to p1, the number of hidden layer neurons to p2, and the number of output layer neurons to p3. The number of input layer neurons and the number of output layer neurons are determined by the number of inputs and the number of outputs. The number of hidden layer neurons p2 can be calculated as follows: Among them, a and b are parameters; Initialize the weights, specifically setting the weights from the input layer neurons to the hidden layer neurons to v ij , represents the weight from the i-th input layer neuron to the j-th hidden layer neuron; the weight from the hidden layer neuron to the output layer neuron is ω jk , represents the weight from the jth hidden layer neuron to the kth output layer neuron; the weight from the previous hidden layer neuron to the current hidden layer neuron is u jl , represents the weight from the lth hidden layer neuron at the previous moment to the jth hidden layer neuron at the current moment; the Sigmoid function is selected as the activation function of the hidden layer neuron; Initialize the neuron nodes and set the jth hidden layer neuron to: h(j)=0 Initialize the kth output layer neuron as: y(k)=0 The target cost function is set as follows: Among them, e is the supervision signal, p is the model output, H is the binary cross entropy, α is the penalty coefficient, W is the parameter set of the neural network used, ⊙ is the XOR logical operation, and the binary cross entropy formula is as follows: Among them, p is the probability value vector, p m is the mth element of the probability value vector, with a value range of [0,1]; y is the discrete label vector, y m is the mth element of the discrete label vector, with a value of 0 or 1, and n is the length of the vector; The training samples are input into the model, and the output of the output layer neurons is calculated as follows: The input to the jth hidden layer neuron is: The output of the jth hidden layer neuron is: h j (t)=f(net j (t)) The input to the kth output layer neuron is: The output of the kth output layer neuron is: y k (t)=f(net k (t)) Where, f represents the Sigmoid function; Set the cost back propagation to the gradient descent algorithm, the learning rate is η, η>0, the momentum factor is λ, 0<λ<1, and the update formula of the connection weight is as follows: in:

14. The channel adaptive hybrid automatic repeat request system according to claim 10, characterized in that: The channel quality parameters are predicted based on the channel quality prediction model, including: Construct an empty channel quality parameter sequence CQ(N)={}; Based on the time window T W The channel quality parameters are counted to obtain N channel quality parameter sequences [CQ(t-(N-1)×T W ),...,CQ(tT W ),CQ(t)], t is the current time; The channel quality parameter sequence is input into the channel quality prediction model to calculate the predicted value of the channel quality at the next moment 15. The channel adaptive hybrid automatic repeat request system according to claim 10, characterized in that: The method of comparing the prediction result of the channel quality parameter with a preset threshold, determining the channel quality according to the comparison result, and determining the encoding and transmission mode of the data for transmission according to the channel quality includes: Setting a preset threshold Th1 and a preset threshold Th2 for determining a corresponding channel quality parameter, wherein Th1 is greater than Th2; When the channel quality parameter prediction value When the channel quality is determined to be high, the data is directly transmitted after verification encoding; When the channel quality parameter prediction value When the channel quality is judged to be average, the hybrid automatic request repeat transmission method is used to perform error correction coding on the data before transmission; When the channel quality parameter prediction value When the channel quality is determined to be poor, a hybrid automatic request retransmission transmission method with repetition coding is used to encode and transmit the data.

16. The channel adaptive hybrid automatic repeat request system according to claim 15, characterized in that: The step of directly transmitting the data after verifying and encoding the data comprises: The sender verifies and encodes the data before sending it; When the receiving end successfully decodes, the receiving end sends a confirmation signal to the sending end to indicate that it has been received correctly and no retransmission is required; when the receiving end fails to decode, the receiving end sends an error signal to the sending end to indicate that the reception has failed, and the sending end repeats sending the corresponding data.

17. The channel adaptive hybrid automatic repeat request system according to claim 15, characterized in that: The hybrid automatic retransmission request transmission method is used to transmit data after error correction coding, including: The sender performs error correction encoding on the data before sending it; The receiving end receives the code and caches it, and decodes it according to the received error correction code; after the receiving end successfully decodes, it sends a confirmation signal to the sending end to indicate that it has been correctly received and no longer needs to be retransmitted; after the receiving end fails to decode, the receiving end sends an error signal to the sending end to indicate that the reception has failed, and the sending end repeats sending the corresponding data.

18. The channel adaptive hybrid automatic repeat request system according to claim 15, characterized in that: The hybrid automatic request retransmission transmission method using repetition coding encodes and transmits data, comprising: The sending end performs error correction coding on the data; The sender copies the same code and sends it together; The receiving end receives the code and caches it; after successfully decoding the first code, the receiving end discards the repeated code and sends a confirmation signal to the sending end to indicate that it is correctly received and no longer needs to be retransmitted; after the receiving end fails to decode the first code, the receiving end performs a Zeiss merger on the two repeated codes; after successfully decoding the merged two codes, the receiving end sends a confirmation signal to the sending end to indicate that it is correctly received and no longer needs to be retransmitted; after failing to decode the merged two codes, the receiving end sends an error signal to the sending end to indicate that the reception failed, and the sending end repeatedly sends the corresponding data.

19. A computer device, characterized in that: include: one or more processors; The processor is used to store one or more programs; When the one or more programs are executed by the one or more processors, the channel adaptive hybrid automatic retransmission request method according to any one of claims 1 to 9 is implemented.

20. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed, the channel adaptive hybrid automatic request retransmission method as claimed in any one of claims 1 to 9 is implemented.

Images