A method, apparatus, and storage medium for constructing a rate-compatible polar code cluster.

Abstract

A method, device, and storage medium for constructing rate-compatible polar code clusters are disclosed. This method constructs a general encoding for a set of polar codewords with various code lengths and rates. First, the encoding construction process of rate-compatible polar codes is transformed into a Markov decision process. Then, for the target polar code cluster, a multi-objective reinforcement learning algorithm is used for encoding construction. Under serial cancellation (SC) decoding conditions, a fast construction algorithm assisted by reconstruction is employed. Under list serial cancellation (SCL) decoding conditions, a joint construction algorithm oriented towards the SCL decoder is used. Finally, a highly reliable general information bit sequence is obtained, improving the error correction capability of the target rate-compatible polar code cluster.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication technology, and relates to channel coding and artificial intelligence technology in wireless communication. It is a coding construction method, device and storage medium for a code rate compatible polar code cluster. Background Technology

[0002] To ensure highly reliable information transmission, wireless communication systems inevitably employ advanced error control techniques, with channel coding being the foundation of wireless communication technology. Polar codes are the first theoretically proven constructive coding scheme with achievable channel capacity, characterized by a regular coding structure, low decoding complexity, and excellent error correction performance. The theoretical basis of polar codes is channel polarization, which generates multiple virtual composite sub-channels with different capacities by recombining and splitting multiple independent physical channels of the same capacity. Based on this, during encoding, high-reliability composite sub-channels are used to transmit bits carrying user information, while low-reliability composite sub-channels carry known, fixed bits. The encoding construction of polar codes involves determining the positions of the composite sub-channels used to carry different information bits during the encoding process; the set of these sub-channel positions is called the information bit sequence.

[0003] In existing wireless communication technologies, diverse transmission services and communication resources often require codewords with various coding lengths and rates. While puncturing or shortening the master code can yield rate-compatible polar codes to meet the needs of different coding lengths and rates, this disrupts the complete codeword structure. The puncturing or shortening patterns are incompatible with the coding construction results, leading to performance loss and reduced error correction capabilities. Reconstructing the information bit sequence from a single codeword can improve performance, but this approach is impractical for real-world communication applications, requiring multiple updates to the information bit sequence, increasing latency and storage costs, and necessitating repeated retrieval and retrieval. Therefore, optimizing the general coding construction results for rate-compatible polar codes with multiple code length patterns is a critical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] The problem this invention aims to solve is that existing technologies, when constructing rate-compatible polar codes, can disrupt the codeword structure, potentially leading to decoding performance loss due to the deletion of information bits, thus affecting codeword reliability and error correction capabilities. Existing improvement schemes are not very practical and difficult to apply in real-world scenarios.

[0005] The technical solution of this invention is as follows: a coding construction method for a rate-compatible polar code cluster. First, the coding construction process of the rate-compatible polar code is transformed into a Markov decision process. On this basis, a multi-objective reinforcement learning algorithm is used to construct the code for the target polar code cluster, and the cases are handled separately. Under the condition of serial cancellation (SC) decoding, a fast construction algorithm with reconstruction assistance is used. Under the condition of list serial cancellation (SCL) decoding, a joint construction algorithm oriented towards the SCL decoder is used. Finally, the general information bit sequence of the target polar code cluster is obtained.

[0006] Furthermore, the coding construction process of rate-compatible polar codes is the process of determining the positions of the synthetic sub-channels used to carry information bits during the coding process. The set of these sub-channel positions is called the information bit sequence. The information bit sequence of rate-compatible polar codes has nested characteristics, thus mapping the construction process to a Markov decision process. For target polar code clusters, the coding construction using a multi-objective reinforcement learning algorithm is as follows:

[0007] In multi-objective reinforcement learning, state s k It is the target information bit sequence in the encoding construction process, i.e., s k : and K is the number of information bits to be encoded, and the complete length of the master code is N = 2^k. n n = 1, 2, ..., information bit sequence A corresponding binary vector set is generated. A positive integer T is the dimension of the code cluster, where integer i ∈ {1, 2, ..., T}, and each binary vector c i This represents the sequence of information bits corresponding to the current position. The encoded length is M i The polar codeword, in the initial state It is a set of vectors containing all zeros, corresponding to the empty set. The action is a positive integer satisfy Each rate-compatible polar code in a polar code family has its own associated reward, which is a vector signal. correspond The return value is Represents polar code clusters The overall error rate performance is related to the information bit sequence. The reward is related to the rate-compatible polar code;

[0008] By interacting with the environment, the agent synchronously optimizes its action strategy ρ according to different objectives, that is:

[0009]

[0010] TQ(·) represents the synthesized state-action vectorization function. It represents its optimal result; Q(·) represents the state-action function for each target object; s and a N Let the current state and action be represented respectively, then the optimal strategy is... Obtain it in the following way:

[0011]

[0012] Based on the weighting criterion for the state-action functions of each target code rate compatible polar code, the calculated Q value of the comprehensive state-action vectorization function TQ(·) is expressed as:

[0013]

[0014] Where, γ i Let represent the positive weight value of the i-th rate-compatible polar code object. The reward value obtained in the k-th step is a vector. The overall reward value is related to the reward value of each target object. Assuming the learning rate α = 1, according to the weighting criterion, the overall reward value is... The calculation is as follows:

[0015]

[0016] bler i,k The block error rate for each target object is obtained based on the reward value of the polar code cluster. Optimal universal information bit sequence

[0017] The present invention also provides an electronic device, the electronic device including a processor and a memory, the memory storing at least one instruction or at least one program, the at least one instruction or at least one program being loaded and executed by the processor, and implementing the above-described encoding construction method for a code rate compatible polar code cluster.

[0018] The present invention also provides a computer-readable storage medium storing at least one instruction or at least one program, wherein the at least one instruction or at least one program is loaded and executed by a processor to implement the above-described method for constructing a code rate compatible polar code cluster.

[0019] In this invention, the target polar code cluster is a given code cluster, and encoding construction is performed on this code cluster; that is, the target polar code cluster is the object of encoding construction. Based on Markov decision processes, this invention employs a multi-objective reinforcement learning algorithm to learn the relationship between rate-compatible polar codes and information bit sequences, for a given polar code cluster. The encoding is constructed to obtain a final optimal general information bit sequence. Complete channel coding.

[0020] This invention proposes a general encoding construction scheme for rate-compatible polar codes with multiple code lengths. The information bit sequence constructed in this invention is applicable to the entire target polar code family, ensuring good overall performance. In contrast, traditional construction methods are only applicable to a single codeword within the code family, resulting in significant performance loss when other codewords are encoded using the same sequence. This invention utilizes multi-objective reinforcement learning to construct encodings for polar codeword sets with multiple code lengths, obtaining a highly reliable general information bit sequence that adapts to target polar codeword sets with various code lengths. This improves the overall error correction capability of the target polar code family and enhances the practicality of the encoding construction scheme. Furthermore, the reinforcement learning algorithm is optimized using the performance estimation characteristics of polar codes, reducing execution complexity. Attached Figure Description

[0021] Figure 1 This is a schematic diagram illustrating the implementation process of a coding construction method for a rate-compatible polar code cluster according to the present invention.

[0022] Figure 2 The diagrams illustrate the transformation of the coding construction process into a Markov decision process. (a) is a diagram illustrating the mapping of the polar code coding construction process to a Markov decision process, and (b) is a diagram illustrating the mapping of the code rate compatible polar code cluster generation process to a Markov decision process.

[0023] Figure 3 This is a flowchart of the fast construction algorithm for reconstruction assistance in this invention.

[0024] Figure 4 This is a flowchart of the joint construction algorithm for the SCL decoder in this invention. Detailed Implementation

[0025] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings. Figure 1 As shown, a coding construction method for a rate-compatible polar code cluster includes: transforming the coding construction process of the rate-compatible polar code into a Markov decision process; for the target polar code cluster, employing a multi-objective reinforcement learning algorithm for coding construction; under conventional serial cancellation decoding conditions, employing a reconstruction-assisted fast construction algorithm; under list serial cancellation (SCL) decoding conditions, employing a joint construction algorithm oriented towards the SCL decoder; and finally obtaining a highly reliable target universal information bit sequence. This invention transforms the coding construction process of the rate-compatible polar code into a Markov decision process, and on this basis, employs a multi-objective reinforcement learning algorithm for coding construction to obtain an information bit sequence. Different construction algorithms are selected to construct the information bit sequence according to different decoding conditions, that is, determining the set of synthetic sub-channels used to carry information bits. The implementation of this invention is described in detail below.

[0026] The target polar code cluster of this invention contains a variety of codewords with different codeword lengths and encoding rates, i.e., code rate compatible polar codes. These codewords are obtained by punching or shortening the full codeword length of the mother code.

[0027] Assume the target polar code cluster is Where T is the positive integer, K is the dimension of the code cluster, K is the number of information bits to be encoded, and the complete length of the mother code is N = 2^k. n (Positive integer n=1 , 2,...). The code rate compatible polar codewords included are... It is obtained through punching or shortening. Here, the punching and shortening algorithms are not required. As a preferred example, quasi-uniform punching and quasi-uniform shortening methods are used here, M. i It's typing The length of , and the corresponding code rate is R. i =K / M i , default M i >M i+1 Bitrate adaptation style Indicates polar code A set of indexes of the positions of the punched or shortened codeword bits.

[0028] This invention first transforms the coding construction process of rate-compatible polar codes into a Markov decision process. Specifically, this coding construction process involves determining the positions of the synthetic sub-channels used to carry information bits during the coding process; the set of these sub-channel positions is called the information bit sequence. During the coding construction process, the information bit sequence of the polar code exhibits nesting characteristics, thus allowing the construction process to be mapped to a Markov decision process (MDP).

[0029] Then, based on MDP, this case is encoded and constructed using multi-objective reinforcement learning, such as... Figure 2 As shown in (a), multi-objective reinforcement learning corresponds to state s in this invention. k It is the target information bit sequence in the encoding construction process, i.e., s k : and Information bit sequence This will generate a corresponding binary vector set. Each binary vector c i This represents the sequence of information bits corresponding to the current position. The encoded length is M i polar codeword, c i It represents the process of learning and training, in the initial state. It is a set of vectors containing all zeros, corresponding to the empty set. Action 'a' is a positive integer Represents the position index of a certain synthesized subchannel, satisfying Every movement For state s k Each has a reward, representing the corresponding state s k The codeword error correction performance. Correspondingly in this invention, each rate-compatible polar code in the polar code cluster has its own associated reward, and the reward signal is a vector. correspond The return value is Represents polar code clusters The overall error rate performance is related to the information bit sequence. Related to the reward for each rate-compatible polar code, the multi-objective learning algorithm of this invention learns the correspondence between the target polar code cluster and the information bit sequence based on the reward value, and finally outputs the optimal information bit sequence for the target polar code cluster.

[0030] For rate-compatible polar codes within a code cluster, there is also a nesting characteristic between them; higher-rate codewords are generated from lower-rate codewords. Assume M... i >M i+1 Then it satisfies Similarly, the generation process of rate-compatible polar code clusters can be mapped to a Markov decision process: such as Figure 2 As shown in (b), state s is the codeword set at this time. Action a is a set of binary codeword bits of a given length, which follows the action 'a'. The reward r represents the code word Error probability performance.

[0031] For target polar code families, this invention combines a multi-objective reinforcement learning algorithm to construct the encoding, thereby obtaining applicable target polar code families. A highly reliable universal information bit sequence. Based on the MDP mapping relationship, the state transition process is executed sequentially. When given (s, a N If the next state s′ is determined, then the information bit sequence can be determined. The construction process. In the multi-objective reinforcement learning process, the agent and the MDP environment interact with each other in various discrete stages. In stage k, the agent observes state s and, according to the policy... Select the corresponding action At the same time, obtain corresponding rewards from the environment. The goal of an agent is to optimize its strategy to maximize overall returns.

[0032] To improve the reliability of the target polar code cluster, the optimization strategy needs to address the following issues:

[0033]

[0034]

[0035] in, Φ and Φ represent the objective functions for the error probability of a single target code rate compatible polar code and the overall error probability of the target polar code cluster, respectively; bler i It is a target rate compatible polar code Block error rate, Is for The preset block error rate threshold.

[0036] To address the problem of improving error correction performance while simultaneously satisfying multiple rate-compatible polar codes, this invention employs a multi-objective reinforcement learning method, providing effective action strategy solutions for tasks with multiple different rewards and continuous decision-making problems.

[0037] By interacting with the environment, the agent synchronously optimizes its action strategy ρ according to different objectives, that is...

[0038]

[0039] TQ(·) represents the synthesized state-action vectorization function. It represents its optimal result; Q(·) represents the state-action function for each target object; s and a N Let the current state and action be represented respectively, then the optimal strategy is... Obtain it in the following way:

[0040] Preferably, the processing is based on the weighted criterion of the state-action function for each target code rate compatible polar code. Other multi-objective function relational criteria can also be used. The Q-value of the comprehensive state-action vectorization function is then expressed as:

[0041]

[0042] γ i This represents the positive weight value of the i-th target rate compatible polar code object. Its definition and calculation are not fixed here. The state-action function value of the i-th target rate compatible codeword is calculated as follows:

[0043]

[0044] Where α represents the learning rate in the multi-objective reinforcement learning algorithm, s′, a N′ These represent the future state and action of the agent during its interaction with the environment. Furthermore, in the encoding construction based on multi-objective reinforcement learning, the reward value obtained at each stage k is a vector, and the overall reward value is related to the reward value of each target object. Assuming a learning rate α = 1, then according to the weighting criterion, the overall reward value... Calculated accordingly

[0045]

[0046] As a preferred example, here we set

[0047] The reward value is an important parameter in multi-objective reinforcement learning algorithms. Selecting an action generates its corresponding reward value, which reflects the quality of the selected action.

[0048] It is worth noting that, under conventional serial cancellation SC decoding conditions, this invention obtains each target object Block error rate in the reward value (bler) i,k The method is not obtained through extensive Monte Carlo numerical simulation statistics, but directly through theoretical calculation estimation. The calculation method is as follows:

[0049]

[0050] Among them, snr d This represents the design signal-to-noise ratio in the Gaussian approximation construction method for polar codes. Indicates the synthesized subchannel The error probability, where j is a positive integer, and is the information bit sequence. In this context, Pr represents the error probability, and u is the bit to be encoded. These are the bits recovered through decoding. Based on the above method, the complexity of obtaining the reward value during the learning process of encoding construction is significantly reduced.

[0051] Under the condition of list-sequence cancellation SCL decoding, in order to simplify the complexity of obtaining rewards during the interaction between the agent and the environment, each target object Return value Block error rate in Bler i,k The maximum likelihood upper bound of the codeword is used for estimation. As an example of optimal selection, the cross-boundary P is adopted here. IB Estimate, i.e.

[0052]

[0053] P IB This represents the upper bound of the maximum likelihood of the codeword, where it is assumed that the noise variance is σ. 2 Transmitting an all-zero codeword in an additive white Gaussian noise channel, A d Let be the number of received Hamming codewords with weight d, where d ≥ 1. Let Q(x) represent the pairwise error probability, and Q(x) be the Gaussian error function. The block error rate estimated by cross-boundary analysis closely matches the block error rate obtained from actual simulation using a list-serial cancellation decoder, especially under high signal-to-noise ratio conditions. The process of estimating the block error rate using the maximum likelihood upper bound only needs to be performed once, avoiding a large number of redundant Monte Carlo simulation operations under list-serial cancellation decoding conditions, thus significantly reducing the complexity of obtaining reward values ​​during the learning interaction process.

[0054] Under conventional serial cancellation decoding conditions, such as Figure 3 The flowchart shown illustrates that this invention is geared towards target code rate compatible polar code clusters. The fast construction algorithm for designing reconstruction assistance is as follows:

[0055] 1.1) Initialize the mother code length N, information bit length K, and target code rate compatible polar code cluster. 1≤i≤T;

[0056] 1.2) Reconstructing each target code rate compatible polar code in the code cluster The corresponding optimal information bit sequence

[0057] 1.3) Determine the initial motion space The initial state information bit sequence s: Let the maximum number of episodes be E;

[0058] 1.4) Randomly initialize TQ(s,a) N );

[0059] 1.5) For each period e (1≤e≤E), repeat the operations from 1.6) to 1.15) below:

[0060] 1.6) Initialize state s;

[0061] 1.7) For stage k of each cycle, repeat the following operation:

[0062] 1.8) Select Action Calculate Returns Calculated s′ is the state updated according to the action; 1.9) For i = 1, 2, ..., T, according to Calculate the corresponding target objects sequentially Q value:

[0063]

[0064] 1.10) Calculate the comprehensive Q-value TQ(s,a) of the corresponding target code cluster. N );

[0065] 1.11) Based on TQ(s,a) N Determine the action

[0066] Update 1.12)

[0067] 1.13) State transition: s←s′;

[0068] 1.14) Determine if the current state s has ended. If not, proceed to step 1.7; if yes, continue to the next step.

[0069] 1.15) Determine if e = E is satisfied. If no, go to step 1.5; if yes, continue to the next step.

[0070] 1.16) Output the constructed optimal information bit sequence

[0071] Under list serial cancellation decoding conditions, such as Figure 4 The flowchart shown represents a target rate-compatible polar code cluster. The joint construction algorithm for an SCL decoder is designed as follows:

[0072] 2.1) Initialize the mother code length N, information bit length K, and target code rate compatible polar code cluster.

[0073] 1≤i≤T;

[0074] 2.2) Determine the initial motion space. here Indicates target rate compatible polar code The set of position indices of the punched or shortened codeword bits, pre-defined for each target codeword. The length relationship is M i >M i+1 When, satisfy Let the maximum number of episodes be E;

[0075] 2.3) Randomly initialize TQ(s,a) N );

[0076] 2.4) For each period e, repeat the following operation:

[0077] 2.5) Initialize state s;

[0078] 2.6) For stage k of each cycle, repeat the following operation:

[0079] 2.7) Select Action Calculate Returns get

[0080] 2.8) For i = 1, 2, ..., T, calculate the corresponding target objects sequentially. Q value:

[0081]

[0082] 2.9) Calculate the comprehensive Q-value TQ(s,a) of the corresponding target code cluster. N );

[0083] 2.10) Based on TQ(s,a) N Determine the action

[0084] Update 2.11)

[0085] 2.12) State transition: s←s′;

[0086] 2.13) Determine if state s has ended. If not, proceed to step 2.6; if yes, continue to the next step.

[0087] 2.14) Determine if e = E is satisfied. If no, go to step 2.4; if yes, continue to the next step.

[0088] 2.15) Output the constructed information bit sequence

[0089] The method of this invention solves the problem of error correction performance loss caused by the destruction of codeword structure and the possible deletion of valid information bits during encoding construction in the prior art by constructing a code rate compatible polar code through the following overall solution:

[0090] 1) Optimize the entire code cluster, i.e., codewords of various code lengths;

[0091] 2) Multi-objective reinforcement learning methods were adopted;

[0092] 3) The coding characteristics of polar codes were utilized during the learning process.

[0093] As an example, the method of the present invention also provides an electronic device, which includes a processor and a memory. The memory stores at least one instruction or at least one program. The at least one instruction or at least one program is loaded and executed by the processor, and when executed, it implements the above-described encoding construction method for a code rate compatible polar code cluster.

[0094] As another embodiment, the present invention also provides a computer-readable storage medium storing at least one instruction or at least one program, wherein the at least one instruction or at least one program is loaded and executed by a processor to implement the above-described method for constructing a code rate compatible polar code cluster.

[0095] The above description is merely a preferred embodiment of the present invention to facilitate understanding by those skilled in the art, but it is not intended to limit the method of the present invention. Any changes, equivalent substitutions, and improvements made within the core design and principles related to the present invention should be included within the protection scope of the present invention.

Claims

1. A coding construction method for rate-compatible polar code clusters, characterized in that: First, the coding construction process of rate-compatible polar codes is transformed into a Markov decision process. Based on this, a multi-objective reinforcement learning algorithm is used to construct the code for the target polar code cluster, and the cases are handled separately. Under the condition of serial cancellation (SC) decoding, a fast construction algorithm with reconstruction assistance is used. Under the condition of list serial cancellation (SCL) decoding, a joint construction algorithm oriented towards the SCL decoder is used. Finally, the general information bit sequence of the target polar code cluster is obtained. Specifically, based on Markov decision processes, a multi-objective reinforcement learning algorithm is used for encoding construction to obtain the target polarization code cluster. The general information bit sequence, It contains multiple codewords with different codeword lengths and encoding rates. , yes Code word length, bitrate matching style express The set of position indices of the punched or shortened codeword bits. The algorithm for construction is selected based on different situations, and the specific handling of each situation is as follows: Under SC decoding conditions, the block error rate in the reward value of each target object is obtained through theoretical calculation and estimation. The calculation method is as follows: in, This represents the design signal-to-noise ratio in the Gaussian approximation construction method for polar codes. Indicates the synthesized subchannel The error probability, where j is a positive integer. One of the elements, Pr, represents the error probability. These are the bits to be encoded. These are the bits recovered through decoding; Under SCL decoding conditions, the reward value for each target object Block error rate Obtained using the maximum likelihood upper bound of codewords, and employing cross-boundary estimation. estimate, This represents a maximum likelihood upper bound for word processing: Where, it is assumed that the noise variance is Transmitting a codeword of all zeros in an additive white Gaussian noise channel The weight of the Hamming received is The number of code words, , Represents the probability of pairwise errors. It is the Gaussian error function; The target polar code code cluster is encoded and constructed using a multi-objective reinforcement learning algorithm as follows: In multi-objective reinforcement learning, state It is the sequence of target information bits in the encoding construction process, that is , , and K is the number of information bits to be encoded, and the full length of the mother code is... , Information bit sequence A corresponding binary vector set is generated. A positive integer T is the dimension of the code cluster, and an integer... Each binary vector This represents the sequence of information bits corresponding to the current position. The length of the encoded output is Polar codewords, in the initial state It is a set of vectors containing all zeros, corresponding to the empty set. The action is a positive integer. ,satisfy , Each rate-compatible polar code in a polar code cluster has its own associated reward, and the reward signal is a vector. = ,correspond The return value is , Represents polar code clusters The overall error rate performance is related to the information bit sequence. The reward is related to the rate-compatible polar code; By interacting with the environment, the agent synchronously optimizes its action strategy according to different objectives. ,Right now: Represents the integrated state-action vectorization function. This indicates its optimal result; This represents the state-action function for each target object; and Let the current state and action be represented respectively, then the optimal strategy is... Obtain it in the following way: ; Based on the weighting criterion of the state-action function for each target code rate compatible polar code, the comprehensive state-action vectorization function is processed. The calculated Q value is expressed as: , in, Let represent the positive weight value of the i-th rate-compatible polar code object. The reward value obtained in the k-th step is a vector, and the overall reward value is related to the reward value of each target object. Let the learning rate be... At that time, based on the weighted criterion, the overall return value The calculation is as follows: The block error rate for each target object is obtained based on the reward value of the polar code cluster. Optimal universal information bit sequence .

2. The coding construction method for a rate-compatible polar code cluster according to claim 1, characterized in that... Under serial cancellation SC decoding conditions, target rate compatible polar code clusters The fast construction algorithm for designing a refactoring aid is as follows: 1.1) Initialize the mother code length N, information bit length K, and target code rate compatible polar code cluster. , ; 1.2) Reconstructing the rate-compatible polar codes for each target code rate in the code cluster The corresponding optimal information bit sequence ; 1.3) Determine the initial motion space The initial state information bit sequence Let the maximum number of cycles be E; 1.4) Random initialization ; 1.5) For each period e, Repeat steps 1.6) to 1.15); 1.6) Initialization State ; 1.7) For stage k of each cycle, repeat steps 1.8) to 1.14): 1.8) Select Action Calculate the return , get ; 1.9) For ,according to Calculate the corresponding target objects sequentially Q value: ; 1.10) Calculate the comprehensive Q value of the corresponding target code cluster. ; 1.11) Based on Determine the action ; Update 1.12) , ; 1.13) State transition: ; 1.14) Determine if state s has ended. If not, proceed to step 1.7); if yes, continue to the next step. 1.15) Determine if e=E. If no, go to step 1.5); if yes, continue to the next step. 1.16) Output the constructed information bit sequence .

3. The coding construction method for a rate-compatible polar code cluster according to claim 1, characterized in that... Under SCL decoding conditions, the target code rate compatible polar code set is oriented towards The joint construction algorithm for an SCL decoder is designed as follows: 2.1) Initialize and set the mother code length N, information bit length K, and target code rate compatible polar code cluster. , ; 2.2) Determine the initial motion space. , here , Indicates target rate compatible polar code The set of position indices of the punched or shortened codeword bits, pre-defined for each target codeword. The length relationship is When, satisfy Let the maximum number of cycles be E; 2.3) Random initialization ; 2.4) For each period e, repeat the following operation: 2.5) Initialization State ; 2.6) For stage k of each cycle, repeat the following operation: 2.7) Select Action Calculate the return , get ; 2.8) For Calculate the corresponding target objects in sequence Q value: ; 2.9) Calculate the comprehensive Q value of the corresponding target code cluster. ; 2.10) Based on Determine the action ; Update 2.11) , ; 2.12) State transition: ; 2.13) Determine if state s has ended. If not, proceed to step 2.6); if yes, continue to the next step. 2.14) Determine if e=E. If no, go to step 2.4); if yes, continue to the next step. 2.15) Output the constructed information bit sequence .

4. An electronic device, characterized in that: The electronic device includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by the processor, and when executed, implements the coding construction method of the code rate compatible polar code cluster as described in any one of claims 1-3.

5. A computer-readable storage medium, characterized in that... The computer-readable storage medium stores at least one instruction or at least one program, which is loaded and executed by a processor to implement the coding construction method of the code rate compatible polar code cluster as described in any one of claims 1-3.

Images