A CBG-HARQ method based on compressed retransmission

Abstract

The present invention discloses a CBG-HARQ method based on compression retransmission, which belongs to the field of data transmission; specifically: during the first transmission, the transmitting end encodes the N CBs contained in the entire CBG to obtain N codewords for transmission, and the receiving end decodes them. For the CBG with an erroneous CB, a 1-bit NAK signal is fed back to request retransmission; during retransmission, the transmitting end compresses the original entire CBG and transmits it; the receiving end calculates the soft information LLR of the retransmitted compressed codeword, and generates a Tanner graph according to the erroneous codeword position index and compression matrix; for each check node in the graph, the external information transmitted to the associated variable node is calculated based on the check relationship, and the variable node synchronously updates the LLR of the erroneous codeword and inputs it into the decoder for decoding. If an erroneous codeword is corrected, the check relationship is updated, and the next iteration is performed until all codewords are correctly decoded, or until no erroneous codeword is corrected after one iteration. The present invention does not increase the feedback overhead, and greatly improves the transmission efficiency of CBG-HARQ.

Description

Technical Field

[0001] The present invention belongs to the field of data transmission, and relates to the transmission efficiency optimization problem of HARQ (hybrid automatic repeat request) based on CBG (code block group) at 5G high data rate, and specifically is a CBG-HARQ method based on compressed retransmission. Background Art

[0002] In recent years, research on various data transmission technologies has been very active. With the introduction of the fifth generation of mobile communication technology, high-speed, low-latency and large-capacity data transmission has been achieved. HARQ is a technology that combines forward error correction and automatic retransmission request. As a key technology to ensure data transmission reliability, it is widely used in various communication scenarios and plays an important role in data transmission.

[0003] The existing 5G system adopts the HARQ scheme based on CBG retransmission. In the 5G physical layer, the information bits in the TB (transport block) are first divided into CBs (code blocks) of equal length, and then multiple CBs in the TB are grouped into CBGs, such as Figure 1 As shown. Since the transmitter performs channel coding on each CB, the receiver can know whether each CB is correct after decoding. In CBG-HARQ, feedback and retransmission are based on CBG, that is, the receiver performs 1-bit ACK / NAK (acknowledgement character / negativeacknowledgment) feedback on each CBG to indicate whether the CBG is transmitted correctly. Similarly, if the transmitter receives NAK feedback, the entire CBG is retransmitted.

[0004] 5G uses CBG-HARQ because of the limitation of feedback signaling overhead. For a CBG containing N CBs, if you want to clearly feedback which CBs are wrong, at least N bits of signaling are required. This HARQ based on CB feedback (CB-HARQ) only needs to retransmit the wrong CB, but it will result in a very large feedback overhead. At the same time, a large amount of feedback on the correctly transmitted CB is meaningless and will also cause a waste of resources.

[0005] The disadvantage of CBG-HARQ is the low retransmission efficiency: since only 1 bit of feedback signaling is used, the transmitter does not know the specific erroneous CB, so all CBs in the CBG must be retransmitted, which results in the retransmitted signal containing many redundant correct CBs. At the high data rate of 5G, a CBG may contain dozens to hundreds of CBs. When the channel conditions are good, such as when the BLER (block error rate) is 0.1, only a few to dozens of CBs actually need to be retransmitted. Therefore, the redundancy in the existing CBG-HARQ of 5G is very large, resulting in low resource utilization of retransmission.

[0006] Most of the existing efficient HARQs are based on CB feedback; for CBG-HARQ, some research results have been achieved to avoid retransmission of the entire CBG through limited multi-bit feedback signaling, such as retransmission based on outer code coding. However, these methods all make a trade-off between feedback overhead and retransmission efficiency, and currently there is no method that can achieve the efficiency of CB-HARQ through 1-bit feedback signaling.

[0007] According to common sense, for sources with large redundancy, in order to improve transmission efficiency, they should be compressed before transmission. Under the CBG-HARQ mechanism, the retransmitted CBG contains a lot of redundancy, and the entropy of the retransmitted CBG is much smaller than the number of retransmitted bits. This is inspired by the following: after receiving 1-bit NAK feedback, first compress the entire CBG into several new CBs, and then transmit them. The receiving end then soft-decompresses the original CBG according to the received signal, which can greatly improve the transmission efficiency of CBG-HARQ. Summary of the invention

[0008] In order to improve the transmission efficiency of CBG-HARQ and reasonably reduce the redundancy in retransmission without increasing the feedback overhead, the present invention proposes a CBG-HARQ method based on compressed retransmission.

[0009] The CBG-HARQ method based on compressed retransmission has the following specific steps:

[0010] Step 1: The transmitter encodes the N CBs contained in the entire CBG to obtain N codewords c1,…,c N Transmitted to the receiving end, the receiving end decodes the received signal, stores all codewords for CBG with erroneous CB, and feeds back a 1-bit NAK signal to request retransmission;

[0011] Soft information LLR (Log-Likelihood Ratio) of all codewords including correct codewords and erroneous codewords;

[0012] Step 2: The sender receives the NAK signal and compresses the original CBG before retransmitting it;

[0013] The sender compresses the entire CBG into a new CB through the compression matrix S and then transmits it through the channel;

[0014] The compression relationship is:

[0015] in, are the original N code words of CBG, is the compressed N c compressed codewords, each of which is a linear combination of the original codewords, and the combination relationship is given by the compression matrix Sure, N represents the binary Galois Field c -N matrix.

[0016] Step 3: The receiving end receives the retransmitted compressed codeword and calculates the soft information LLR of the retransmitted compressed codeword, which is recorded as

[0017] Step 4: Generate a Tanner graph including variable nodes corresponding to the error codeword and check nodes corresponding to the retransmitted compressed codeword according to the error codeword position index and the compression matrix S;

[0018] The variable node is initialized to The check nodes are initialized to

[0019] Among them, b1, b2, …, b B Indicates the position index of the error codeword, b1, b2, ..., b B ∈{1,2,…,N}; LLR of the error codeword stored by the receiving end;

[0020] Step 5: For each check node in the Tanner graph, calculate the external information transmitted to the associated variable node based on the check relationship;

[0021] When generating compressed codewords from the compression matrix, there is an overall check relationship:

[0022] Among them, S A is the compressed matrix S of a1, a2, …, a A Column, C A is the a1, a2, …, a of the original codeword matrix C A OK, S B is the compressed matrix S of b1, b2, …, b B Column, C B is the b1, b2, ..., b of the original codeword matrix CB OK.

[0023] a1,a2,…,a A Indicates the position index of the decoded codeword, a1, a2, ..., a A ∈{1,2,…,N}, A is the number of correctly decoded codewords; b1,b2,…,b B The position index of the codeword indicating the error, b1, b2, ..., b B ∈{1,2,…,N}, B is the number of incorrectly translated code words; A+B=N.

[0024] From the j-th row of the overall check relation, the check relation corresponding to the j-th check node is:

[0025] Among them, w j W = S A C A The vector of the jth row of is known at the receiving end, s j,i represents the element in the jth row and ith column of the compressed matrix S, is the vector of the original codeword;

[0026] Therefore, for the jth check node, based on this check relationship, the variable node b associated with it is calculated i External information transmitted The specific formula for calculating external information is:

[0027]

[0028] in, is the index of the error codeword associated with the jth check node, λ (k) represents the kth element of vector λ, Represents vector w j The kth element of .

[0029] Step 6: After receiving external information, the variable nodes in the Tanner graph synchronously update the LLR of the error codeword;

[0030] Error code word The formula for updating LLR is:

[0031] in, Is the error code word The index of the associated compressed codeword.

[0032] Step 7: Input the updated LLR into the decoder for decoding to determine whether the error codewords are corrected. If all error codewords are corrected, the transmission is successful; otherwise, proceed to step 8;

[0033] Step 8. Continue to determine whether some erroneous codewords have been corrected. If so, update the verification relationship, that is, delete the correctly translated variable nodes and corresponding edges in the Tanner graph, and return to step 5. Otherwise, if no erroneous codewords have been corrected, a 1-bit NAK signal is fed back to request retransmission until the maximum number of retransmissions is reached.

[0034] The maximum number of retransmissions is set manually according to actual needs, and 3 is selected in this embodiment.

[0035] The advantages of the present invention are:

[0036] 1) A CBG-HARQ method based on compressed retransmission. Compared with the existing CBG-HARQ, the present invention greatly reduces the redundancy in retransmission, avoids resource waste, and significantly improves the overall transmission efficiency.

[0037] 2) A CBG-HARQ method based on compressed retransmission. Compared with the existing CB-HARQ, the present invention only requires 1 bit of feedback overhead, and it can be seen from the soft decompression algorithm that a compressed codeword in the present invention can be associated with multiple error codewords, making full use of the merging of soft information, and can achieve higher transmission efficiency than CB-HARQ. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a schematic diagram of the CBG structure;

[0039] Figure 2 A schematic diagram for comparing the processes of the existing retransmission mechanism and the compressed retransmission mechanism;

[0040] Figure 3 A flow chart of a CBG-HARQ method based on compressed retransmission according to the present invention;

[0041] Figure 4 This is the message passing Tanner graph used for soft decompression in the present invention. DETAILED DESCRIPTION

[0042] In order to further illustrate the implementation method of the present invention, an implementation example is given below. This example is only intended to illustrate the principle of the present invention and does not represent any limitation of the present invention.

[0043] For the 5G high-speed data transmission scenario, the present invention proposes a CBG-HARQ method based on compressed retransmission, which can greatly improve the transmission efficiency without increasing the feedback overhead; the compressed retransmission and soft decompression method is applicable to all channel decoding that requires soft information, and the present invention does not specifically limit the specific channel model, modulation method, and channel coding method used.

[0044] The present invention provides a CBG-HARQ method based on compression retransmission, for the transmission of a CBG, by Figure 2 The comparison of the retransmission mechanism shown in FIG. 1 shows that the first transmission is the same as the existing HARQ scheme. The transmitter sends a complete CBG, and the receiver calculates soft information (usually LLR (Log-Likelihood Ratio)) based on the received signal and then decodes it. If there is a CB decoding error in the CBG, the receiver stores the correct codeword and the LLR of the error codeword, and feeds back a 1-bit NAK signal to the transmitter to request retransmission. The retransmission adopts the method of compressed retransmission and soft decompression.

[0045] like Figure 3 As shown, the specific steps are as follows:

[0046] Step 1: During the first transmission, the sender sends the N codewords c1,…,c contained in the entire CBG. N , the receiving end decodes the received signal, stores all codewords for the CBG with an erroneous CB, and feeds back a 1-bit NAK signal to request retransmission;

[0047] Soft information LLR (Log-Likelihood Ratio) of all codewords including correct codewords and erroneous codewords;

[0048] Step 2: The sender receives the NAK signal and compresses the original CBG before retransmitting it;

[0049] In the existing 5G, retransmission is performed based on the entire CBG, that is, all N codewords corresponding to the CBG are sent again during retransmission, such as Figure 2 (a) As shown; the retransmission method of the present application is different from the prior art solution. During retransmission, the transmitting end no longer sends all N codewords, but sends compressed N c New codeword like Figure 2 (b) When retransmitting, the transmitter converts the original N codewords c1, c2, …, c N Compress to N c Compressed codeword (Codewords are represented by row vectors) and then transmitted through the channel; the compression relationship is:

[0050] in, are the original N code words of CBG, is the compressed N c compressed codewords, each of which is a linear combination of the original codewords, and the combination relationship is given by the compression matrix Sure, N represents the binary Galois Fieldc -N matrix.

[0051] Send compressed codewords on retransmission Due to N c <N, so the number of compressed codewords is less than the number of original codewords, reducing the resources required for retransmission;

[0052] Step 3: The receiving end receives the retransmitted compressed codeword and calculates the soft information LLR of the retransmitted compressed codeword, which is recorded as

[0053] Step 4: Generate a Tanner graph including variable nodes corresponding to the error codeword and check nodes corresponding to the retransmitted compressed codeword according to the error codeword position index and the compression matrix S;

[0054] like Figure 4 As shown, the variable node is initialized to The check nodes are initialized to is the wrong original codeword; b1, b2, …, b B Indicates the position index of the error codeword, b1, b2, ..., b B ∈{1,2,…,N}; LLR of the error codeword stored by the receiving end;

[0055] Step 5: For each check node in the Tanner graph, perform soft decompression based on the check relationship and calculate the external information transmitted to the associated variable node;

[0056] When generating compressed codewords from the compression matrix, there is an overall check relationship Extract the jth compressed codeword from it The specified check relationship is used to obtain the calculation formula for calculating the external information of the j-th check node according to the check relationship;

[0057] The overall verification relationship is:

[0058] Among them, S A is the compressed matrix S of a1, a2, …, a A Column, C A is the a1, a2, …, a of the original codeword matrix C A OK, S B is the compressed matrix S of b1, b2, …, b B Column, C B is the b1, b2, ..., b of the original codeword matrix C B OK.

[0059] a1,a2,…,a A Indicates the position index of the decoded codeword, a1, a2, ..., aA ∈{1,2,…,N}, A is the number of correctly decoded codewords; b1,b2,…,b B The position index of the codeword indicating the error, b1, b2, ..., b B ∈{1,2,…,N}, B is the number of incorrectly translated code words; A+B=N.

[0060] From the jth row of the overall check relation, we can get the check relation corresponding to the jth check node:

[0061] Among them, w j W = S A C A The vector of the jth row of s is known at the receiving end, s j,i represents the element in the jth row and ith column of the compressed matrix S, is the vector of the original codeword; therefore, for the jth check node, based on this check relationship, And other associated variable nodes Calculate the variable node b associated with it i External information transmitted The specific formula for calculating external information is:

[0062]

[0063] in, is the index of the error codeword associated with the jth check node, λ (k) represents the kth element of vector λ, Represents vector w j The kth element of .

[0064] Step 6: After receiving the external information, the variable node in the Tanner graph synchronously updates the LLR of the error codeword;

[0065] The i-th variable node corresponds to the error codeword Update LLR according to the received external information, i = 1, 2, ..., B;

[0066] Error code word The formula for updating LLR is:

[0067] in, Is the error code word The index of the associated compressed codeword.

[0068] Step 7: Input the updated LLR into the decoder for decoding to determine whether the error codewords are corrected. If all error codewords are corrected, the transmission is successful; otherwise, proceed to step 8;

[0069] Step 8. Continue to determine whether some erroneous codewords have been corrected. If so, update the verification relationship, that is, delete the correctly translated variable nodes and corresponding edges in the Tanner graph of the message transmission, and return to step 5; otherwise, if no erroneous codewords have been corrected, a 1-bit NAK signal is fed back to request retransmission until the maximum number of retransmissions is reached.

[0070] The iteration termination criteria are: (1) All codewords are decoded correctly, ACK is fed back, and the transmission is completed; (2) After one iteration, no erroneous codewords are corrected, NAK is fed back and the next retransmission is requested. The above scheme is still used for the next retransmission until the maximum number of retransmissions is reached. The maximum number of retransmissions is set manually according to actual needs, and the maximum number of retransmissions is usually 3.

[0071] Example:

[0072] The CBG selected in this embodiment includes N=100 CBs, which are encoded using LDPC codes to become N original codewords c1, c2, ..., c N (Codewords are represented by row vectors.) The codewords are transmitted through an AWGN channel after being modulated by QPSK. When the signal-to-noise ratio is 1.4 dB, the BLER of a single transmission of the codeword is p=0.58.

[0073] All N original codewords are sent in the first transmission. The receiving end calculates LLR based on the received signal and then decodes the CBs separately. If all CBs in the CBG are decoded correctly, a 1-bit ACK is fed back and the transmission ends; if there is a CB decoding error in the CBG, the receiving end stores the LLR of the correct codeword and the error codeword, and feeds back a 1-bit NAK signal to request retransmission, which uses the method of compressed retransmission and soft decompression.

[0074] The specific steps are as follows:

[0075] Step 1: During retransmission, the sender compresses the entire CBG into some new CBs through a compression matrix S, and then transmits them through the channel;

[0076] The original N codewords c1, c2, ..., c N Compress to N c Compressed codeword (Codewords are represented by row vectors), the compression relationship is:

[0077] in, are the original N code words of CBG, is the compressed N c compressed codewords, each of which is a linear combination of the original codewords, and the combination relationship is given by the compression matrix Sure, N represents the binary Galois Field c ×N matrix;

[0078] In this embodiment, the compression matrix S used is a 34×100 matrix, that is, N c =34, S is generated by a pseudo-random algorithm, satisfying that the weight of each column is 2 and the weight of each row is 6 or 5;

[0079] Send 34 compressed codewords during retransmission Compared with retransmitting 100 original codewords, the resources required for retransmission are greatly reduced;

[0080] Step 2: The receiving end receives the retransmitted compressed codeword and calculates the LLR of the compressed codeword, which is recorded as

[0081] Step 3: Generate a Tanner graph for message passing based on the untranslated codewords and the compression matrix S;

[0082] Step 4: Update the LLR of the error codeword through the message passing algorithm;

[0083] In the message passing algorithm, the variable node is initialized as The check nodes are initialized to

[0084] Then, the jth check node is based on Verify relationships and other variable nodes Calculate the external information transmitted to the variable node, j = 1, 2, ..., N c ;

[0085] When executing this step, the overall verification relationship is: Therefore, the check relationship specified by the jth compressed codeword is: Among them, w j W = S A C A The jth row of s is known at the receiving end. j,i Represents the element in the j-th row and i-th column of matrix S.

[0086] Then, the i-th variable node updates the LLR after receiving the external information, i=1,2,…,B;

[0087] Step 5: Use the updated LLR to input the decoder for decoding. If an erroneous codeword is corrected, update the check relation.

[0088] The update verification relationship is reflected in the Tanner graph of message passing by deleting the variable nodes and corresponding edges of the translated pairs;

[0089] Step 6: Repeat steps 4 and 5 until the iteration ends.

Claims

1. A CBG-HARQ method based on compressed retransmission, characterized in that: The specific steps are as follows: Step 1: The transmitter encodes the N CBs contained in the entire CBG to obtain N codewords c 1, …,c N Transmitted to the receiving end, the receiving end decodes the received signal, stores all codewords for CBG with erroneous CB, and feeds back a 1-bit NAK signal to request retransmission; Step 2: The transmitter receives the NAK signal, compresses the original codeword in the entire CBG into a new compressed codeword through the compression matrix S, and then retransmits it through the channel; Step 3: The receiving end receives the retransmitted compressed codeword and calculates the soft information LLR of the retransmitted compressed codeword, which is recorded as Step 4: Generate a Tanner graph including variable nodes corresponding to the error codeword and check nodes corresponding to the retransmitted compressed codeword according to the error codeword position index and the compression matrix S; Step 5: For each check node in the Tanner graph, calculate the external information transmitted to the associated variable node based on the check relationship; The overall verification relationship is: Among them, S A is the compressed matrix S a1,a 2, …,a A Column, C A is a1,a of the original codeword matrix C 2, …,a A OK, S B is the b1,b of the compressed matrix S 2, …,b B Column, C B is the b1,b of the original codeword matrix C 2, …,b B OK; a1,a 2, …,a A Indicates the position index of the decoded codeword, a1,a 2, …,a A ∈{1,2,…,N}, A is the number of correctly decoded codewords; b1,b2,…,b B Indicates the position index of the wrong codeword, b1,b 2, …,b B ∈{1,2,…,N}, B is the number of untranslated codewords; A+B=N; From the jth row of the overall check relation, the check relation corresponding to the jth check node is: Among them, w j W = S A C A The vector of the jth row of is known at the receiving end, s j,i represents the element in the jth row and ith column of the compressed matrix S, is the vector of the original codeword; Therefore, for the jth check node, based on this check relationship, the variable node b associated with it is calculated i External information transmitted The specific formula for calculating external information is: in, is the index of the error codeword associated with the jth check node, λ (k) represents the kth element of vector λ, Represents vector w j The kth element of ; Step 6: After receiving external information, the variable nodes in the Tanner graph synchronously update the LLR of the error codeword; Step 7: Input the updated LLR into the decoder for decoding to determine whether the error codewords are corrected. If all error codewords are corrected, the transmission is successful; otherwise, proceed to step 8; Step 8. Continue to determine whether some erroneous codewords have been corrected. If so, update the verification relationship, that is, delete the correctly translated variable nodes and corresponding edges in the Tanner graph, and return to step 5. Otherwise, if no erroneous codewords have been corrected, a 1-bit NAK signal is fed back to request retransmission until the maximum number of retransmissions is reached.

2. A CBG-HARQ method based on compressed retransmission according to claim 1, characterized in that: In the step 1, all the stored codewords include the soft information LLRs of the correct codewords and the erroneous codewords.

3. The CBG-HARQ method based on compressed retransmission according to claim 1, characterized in that: The compression relationship is: in, are the original N code words of CBG, is the compressed N c compressed codewords, each of which is a linear combination of the original codewords, and the combination relationship is given by the compression matrix Sure, N represents the binary Galois Field c A matrix of size N × N.

4. The CBG-HARQ method based on compressed retransmission according to claim 1, characterized in that: The variable nodes in the Tanner graph are initialized as The check nodes are initialized to in, It is the LLR of the erroneous codeword stored by the receiving end.

5. The CBG-HARQ method based on compressed retransmission according to claim 1, characterized in that: The error code word The formula for updating LLR is: in, Is the error code word The index of the associated compressed codeword.

6. The CBG-HARQ method based on compressed retransmission according to claim 1, characterized in that: The maximum number of retransmissions is set manually according to actual needs.

Images