Trainable joint channel estimation, detection and decoding method for URLLC-MIMO systems

Abstract

This invention provides a trainable joint channel estimation, detection, and decoding method suitable for URLLC-MIMO systems. The method considers a URLLC-MIMO system where the transmitter maps an LDPC codeword to modulation symbols and distributes them evenly across different transmit antennas. The receiver constructs a joint channel estimation, detection, and decoding problem model and solution framework based on the MAP criterion, then deeply expands it into a neural network, trains its adjustable parameters offline, and finally uses the trained neural network to acquire codeword bits online. Compared to traditional receiving methods, this invention significantly improves system BLER performance, reduces system latency, and saves signaling overhead.

Description

Technical Field

[0001] This invention relates to the field of wireless communication, and in particular to a trainable joint channel estimation, detection and decoding method suitable for URLLC-MIMO systems. Background Technology

[0002] Fifth-generation mobile communication (5G) has spurred the development of new communication scenarios, such as ultra-reliable low-latency communications (URLLC), to support emerging services like telemedicine and industrial automation. To meet the stringent low-latency requirements of URLLC, systems typically employ short-packet transmission schemes. On the other hand, MIMO (Multiple-Input Multiple-Output) technology can improve system transmission reliability by utilizing spatial redundancy. Therefore, short-packet MIMO transmission has gradually gained attention. However, existing research mainly focuses on theoretical performance analysis under finite code lengths, with relatively little research considering receiver algorithms for URLLC-MIMO systems. Furthermore, the limited decoding capability of shorter code-length error-correcting codes poses challenges to receiver design.

[0003] Traditional turbo receivers improve performance by explanatory information exchanged between different modules, a method successfully applied to LDPC (Low Density Parity Check Code) long-code coding systems. However, when considering short packet transmission, short code coding is required. The presence of short cycles in the Tanner graph of LDPC short codes can introduce unfavorable positive feedback during backpropagation (BP) decoding, leading to error propagation in turbo iterations and performance degradation. Furthermore, to ensure efficient short packet transmission, pilot length is often limited, restricting the performance of pilot-based channel estimation and impacting overall system performance. Additionally, the interaction between turbo receiver modules introduces non-negligible delays, which does not meet the requirements of URLLC (Ultra-URLLC). Summary of the Invention

[0004] The technical problem to be solved by this invention is to provide a trainable joint channel estimation, detection and decoding method suitable for URLLC-MIMO systems, so as to avoid performance loss and error propagation caused by module partitioning, significantly improve the system BLER performance, reduce system latency and save signaling overhead.

[0005] The technical solution adopted by the present invention to solve the above technical problems is as follows:

[0006] A trainable joint channel estimation, detection, and decoding method for URLLC-MIMO systems is proposed. Considering a URLLC-MIMO system, the transmitter maps an LDPC codeword to modulation symbols and distributes them evenly across different transmit antennas. The receiver constructs a joint channel estimation, detection, and decoding problem model and solution framework based on the MAP criterion, then deeply expands it into a neural network, trains its adjustable parameters offline, and finally uses the trained neural network to acquire codeword bits online. The specific steps include:

[0007] (1) Construct a joint channel estimation, detection and decoding problem model and solution framework based on the MAP criterion;

[0008] (2) Construct a neural network for solving the deep expansion joint channel estimation, detection and decoding problem, and set the adjustable parameters therein as trainable parameters;

[0009] (3) Offline stage: Generate the dataset required for training, train the neural network, and obtain trainable parameters;

[0010] (4) Online stage: The receiving end uses the trained neural network to make a decision on the output of each layer of the network. If the verification constraint is satisfied, the processing is terminated in advance and the current decision is output; otherwise, the decision corresponding to the output of the last layer of the network is output.

[0011] Furthermore, in the aforementioned trainable joint channel estimation, detection, and decoding method suitable for URLLC-MIMO systems, it is assumed that a URLLC-MIMO system is configured with... root transmitting antenna and The root receiving antenna, the channel coefficient is in The transmission time of each symbol remains unchanged, where and This indicates the number of pilot symbols and the number of data symbols. In the data transmission order, the transmitting end first encodes the information bits into a code with a code length of... LDPC codewords .make This represents the binary parity-check matrix of the LDPC code, where To verify the number of nodes, the sending end then maps the codeword to modulation symbols. And it uses multi-stream multiplexing technology to distribute the data evenly. Transmitted on the root transmitting antenna, where Let be the modulation order. Let the pilot matrix and data matrix be... and ,in The received signal can then be expressed as:

[0012] ,

[0013] in, and These represent the pilot receiver matrix and the data receiver matrix, respectively. Let represent the Gaussian channel matrix, where each element is independent and identically distributed, with a mean of 0 and a variance of 1. Let represent an additive white Gaussian noise matrix, where each element is independent and identically distributed, with a mean of 0 and a variance of . . Notice The Each element corresponds to a modulation symbol. Furthermore, for ease of subsequent explanation, Recorded as , This represents the bit-symbol mapping function. Also, consider its vectorized form. , and They are respectively , and .

[0014] Furthermore, in step (1), the following joint channel estimation, detection, and decoding problem model based on the MAP criterion is constructed:

[0015] Optimization goal:

[0016] Constraints:

[0017] ,

[0018] ,

[0019] in Indicates modulo 2 operation. The dimension is The identity matrix, Represents the LDPC code parity check matrix The Each element.

[0020] Furthermore, in step (1), the original problem is rewritten in the following form:

[0021] Optimization goal:

[0022] Constraints:

[0023] ,

[0024] ,

[0025] in, The introduced penalty factor, Indicates the first The set of variables involved in each verification formula This indicates that the condition of odd cardinality is satisfied. The set of subsets. For ease of subsequent representation, the first constraint is rewritten in matrix form according to the element correspondence rule as follows:

[0026] ,

[0027] in, and These represent the weight matrix and the bias matrix, respectively.

[0028] Furthermore, in step (1), auxiliary variables are introduced. dual variables and penalty factor The rewritten problem is solved based on the Alternating Direction Multiplier Method (ADMM), where the first... Each ADMM iteration includes the following steps:

[0029] (1.1) Calculate intermediate variables :

[0030] ;

[0031] (1.2) Calculate intermediate variables :

[0032] ,

[0033] in, , This indicates taking the largest eigenvalue of the input matrix;

[0034] (1.3) For Update the bit variables sequentially according to the following formula:

[0035] ,

[0036] in, , Representation matrix The List, Representation matrix The One diagonal element, This indicates projecting the input element onto... The interval. Additionally... and The solution depends on the modulation order. Taking QPSK as an example, at this time... , and It can be calculated using the following formula;

[0037] ,

[0038] ,

[0039] in, express The One element, and These represent taking the real and imaginary parts of the input element, respectively.

[0040] (1.4) Update the auxiliary variable according to the following formula. :

[0041] ,

[0042] in, This means projecting each element of the input vector onto... interval;

[0043] (1.5) Update the dual variable according to the following formula. :

[0044] .

[0045] Furthermore, in step (2), the ADMM solution framework is expanded in depth, and the first... The ADMM iteration corresponds to the [number]th [item]. Layered neural network, and penalty factor and as well as Set as trainable parameters , This represents the number of network layers.

[0046] Furthermore, in step (3), the offline phase includes the following steps:

[0047] (3.1) Generate the dataset required for training, with each set of data including the network input. , , , The labels used for training, i.e., the actual transmitted codewords. .

[0048] (3.2) The first Output of layered neural networks and tags The mean squared error between the two sides is used as the loss function for training the neural network to obtain trainable parameters.

[0049] Furthermore, in step (4), during the online phase, the receiving end uses the trained neural network to make a decision on the output of each layer. If the verification constraint is satisfied, the processing is terminated early, and the current decision is output; otherwise, the decision corresponding to the output of the last layer is output. Specifically, for the first... The output of the layer network is determined according to the following formula:

[0050] .

[0051] Compared with the prior art, the present invention has the following advantages and beneficial technical effects:

[0052] (1) This invention provides a trainable joint channel estimation, detection and decoding method for URLLC-MIMO systems, which avoids performance loss and error propagation caused by module partitioning and can significantly improve the BLER performance of the system.

[0053] (2) Compared with traditional receiving methods, the present invention avoids iterative interaction between different modules, reduces system latency, and saves signaling overhead. Attached Figure Description

[0054] Figure 1 The present invention provides a flowchart of a trainable joint channel estimation, detection and decoding method applicable to URLLC-MIMO systems.

[0055] Figure 2 The network structure diagram provided by this invention.

[0056] Figure 3 This is a schematic diagram showing the results of a simulation experiment of the present invention provided in the embodiments. Detailed Implementation

[0057] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0058] This invention provides a trainable joint channel estimation, detection, and decoding method suitable for URLLC-MIMO systems, which can significantly improve the system's BLER performance, reduce system latency, and save signaling overhead.

[0059] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0060] To verify the performance of the trainable joint channel estimation, detection, and decoding method for URLLC-MIMO systems provided in this invention, a URLLC-MIMO system simulation platform needs to be built. In this embodiment, a URLLC-MIMO system is assumed to be configured... root transmitting antenna and The root receiving antenna, the channel coefficient is in The transmission time of each symbol remains unchanged, where and This indicates the number of pilot symbols and the number of data symbols. In the data transmission order, the transmitting end first encodes the information bits into a code with a code length of... LDPC codewords .make This represents the binary parity-check matrix of the LDPC code, where To verify the number of nodes, the sending end then maps the codeword to modulation symbols. And it uses multi-stream multiplexing technology to distribute the data evenly. Transmitted on the root transmitting antenna, where Let be the modulation order. Let the pilot matrix and data matrix be... and ,in The received signal can then be expressed as:

[0061] ,

[0062] in, and These represent the pilot receiver matrix and the data receiver matrix, respectively. Let represent the Gaussian channel matrix, where each element is independent and identically distributed, with a mean of 0 and a variance of 1. Let represent an additive white Gaussian noise matrix, where each element is independent and identically distributed, with a mean of 0 and a variance of . . Notice The Each element corresponds to a modulation symbol. Furthermore, for ease of subsequent explanation, Recorded as , This represents the bit-symbol mapping function. Also, consider its vectorized form. , and They are respectively , and .

[0063] like Figure 1 As shown, this method is performed according to the following steps:

[0064] Step (1): Construct the following joint channel estimation, detection, and decoding problem model based on the MAP criterion:

[0065] Optimization goal:

[0066] Constraints:

[0067] ,

[0068] ,

[0069] in Indicates modulo 2 operation. The dimension is The identity matrix, Represents the LDPC code parity check matrix The Each element.

[0070] To facilitate solving, the original problem is rewritten in the following form:

[0071] Optimization goal:

[0072] Constraints:

[0073] ,

[0074] ,

[0075] in, The introduced penalty factor, Indicates the first The set of variables involved in each verification formula This indicates that the condition of odd cardinality is satisfied. The set of subsets. For ease of subsequent representation, the first constraint is rewritten in matrix form according to the element correspondence rule as follows:

[0076] ,

[0077] in, and These represent the weight matrix and the bias matrix, respectively.

[0078] Next, we introduce auxiliary variables. dual variables and penalty factor The rewritten problem is solved based on the Alternating Direction Multiplier Method (ADMM), where the first... Each ADMM iteration includes the following steps:

[0079] (1.1) Calculate intermediate variables :

[0080] .

[0081] (1.2) Calculate intermediate variables :

[0082] ,

[0083] in, , This indicates taking the largest eigenvalue of the input matrix.

[0084] (1.3) For Update the bit variables sequentially according to the following formula:

[0085] ,

[0086] in, , Representation matrix The List, Representation matrix The One diagonal element, This indicates projecting the input element onto... The interval. Furthermore, under the QPSK modulation considered in this embodiment, and It can be calculated using the following formula;

[0087] ,

[0088] ,

[0089] Under 16-QAM, and It can be calculated using the following formula;

[0090]

[0091]

[0092]

[0093]

[0094] in, express The One element, and These represent taking the real and imaginary parts of the input element, respectively.

[0095] (1.4) Update the auxiliary variable according to the following formula. :

[0096] ,

[0097] in, This means projecting each element of the input vector onto... Interval.

[0098] (1.5) Update the dual variable according to the following formula. :

[0099] .

[0100] Step (2): Deeply expand the ADMM solution framework described above, and then... The ADMM iteration corresponds to the [number]th [item]. Layered neural network, and penalty factor and as well as Set as trainable parameters , This refers to the network layer number. Specifically, the [number]th [layer]. The structure of a layered neural network is as follows: Figure 2 As shown.

[0101] Step (3): The offline phase includes the following steps:

[0102] (3.1) Generate the dataset required for training, with each set of data including the network input. , , , The labels used for training, i.e., the actual transmitted codewords. .

[0103] (3.2) The first Output of layered neural networks and tags The mean squared error between the two sides is used as the loss function for training the neural network to obtain trainable parameters.

[0104] Step (4): In the online phase, the receiving end uses the trained neural network to make a decision on the output of each layer. If the check constraint is satisfied, the processing is terminated early, and the current decision is output; otherwise, the decision corresponding to the output of the last layer is output. More specifically, for the... The output of the layer network is determined according to the following formula:

[0105] .

[0106] By comparing this invention with traditional MAP-Turbo receivers and MMSE-Turbo receivers that consider data-assisted channel estimation, we conducted simulation experiments to evaluate the superiority of this invention. The simulation parameters are shown in the table below:

[0107] Table 1 Simulation Parameter List

[0108]

[0109] With SNR representing the signal-to-noise ratio on the horizontal axis and BLER representing the block error rate on the vertical axis, the simulation results of the specific embodiments described above are as follows: Figure 3 As shown in the figure. Simulation results demonstrate that, under finite pilot overhead, the trainable joint channel estimation, detection, and decoding method proposed in this invention exhibits significant BLER performance advantages compared to traditional MAP-Turbo and MMSE-Turbo receivers that consider data-aided channel estimation.

[0110] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any modifications or equivalent changes made based on the technical essence of the present invention shall still fall within the scope of protection claimed by the present invention.

Claims

1. A trainable joint channel estimation, detection, and decoding method suitable for URLLC-MIMO systems, characterized in that, Specifically, the steps include the following: S1. Construct a joint channel estimation, detection, and decoding problem model and solution framework based on the MAP criterion; where: Optimization goal: ， Constraints: , ， in Indicates modulo 2 operation. , and They are vectorized forms respectively , and , and These represent the pilot receiver matrix and the data receiver matrix, respectively. Represents the Gaussian channel matrix. Represents the pilot matrix, The dimension is The identity matrix, The number of receiving antennas, The number of transmitting antennas, Indicates the number of pilot symbols. Indicates the number of data symbols; For variance, For code length is LDPC codewords, Represents the bit-sign mapping function, Represents the LDPC code parity check matrix The One element, Represents the binary parity check matrix of the LDPC code. To verify the number of nodes; S2. Construct a neural network for solving the deep expansion joint channel estimation, detection and decoding problem, and set the adjustable parameters in it as trainable parameters; S3, Offline Stage: Generate the dataset required for training, train the neural network, and obtain trainable parameters; S4. Online Phase: The receiving end uses the trained neural network to make a decision on the output of each layer of the network. If the verification constraint is satisfied, the processing is terminated early and the current decision is output; otherwise, the decision corresponding to the output of the last layer of the network is output.

2. The trainable joint channel estimation, detection, and decoding method for URLLC-MIMO systems according to claim 1, characterized in that, Before performing step S1, the following is also included: configuring the URLLC-MIMO system. root transmitting antenna and The root receiving antenna, the channel coefficient is in The transmission time of each symbol remains unchanged. During the data transmission phase, after an LDPC codeword is mapped to a modulation symbol, it is evenly distributed to different transmitting antennas for transmission.

3. The trainable joint channel estimation, detection, and decoding method for URLLC-MIMO systems according to claim 2, characterized in that, The process of mapping an LDPC codeword to a modulation symbol and then equally distributing it across different transmit antennas for transmission is as follows: First, the sending end encodes the information bits into a code with a length of LDPC codewords ,make This represents the binary parity-check matrix of the LDPC code, where To verify the number of nodes; Then, the transmitting end maps the codeword into modulation symbols. And it uses multi-stream multiplexing technology to distribute the data evenly. Transmitted on the root transmitting antenna, where The modulation order; Let the pilot matrix and data matrix be... and ,in The received signal can then be expressed as: ， in, and These represent the pilot receiver matrix and the data receiver matrix, respectively. Let represent the Gaussian channel matrix, where each element is independent and identically distributed, with a mean of 0 and a variance of 1; Let represent an additive white Gaussian noise matrix, where each element is independent and identically distributed, with a mean of 0 and a variance of . ; noticed The Each element corresponds to a modulation symbol. ,Will Recorded as , This represents the bit-symbol mapping function.

4. The trainable joint channel estimation, detection, and decoding method for URLLC-MIMO systems according to claim 1, characterized in that, In step S1, the original problem is rewritten in the following form: Optimization goal: , Constraints: , ， in, The introduced penalty factor, Indicates the first The set of variables involved in each verification formula This indicates that the condition of odd cardinality is satisfied. The set of subsets of; The first constraint is rewritten in matrix form according to the element-to-element correspondence rule: ， in, and These represent the weight matrix and the bias matrix, respectively.

5. A trainable joint channel estimation, detection, and decoding method for URLLC-MIMO systems according to claim 4, characterized in that, In step S1, auxiliary variables are introduced. dual variables and penalty factor The rewritten problem is solved based on the Alternating Direction Multiplier Method (ADMM), where the first... Each ADMM iteration includes the following steps: (1) Calculate intermediate variables : ； (2) Calculate intermediate variables : ， in, , This indicates taking the largest eigenvalue of the input matrix; (3) For Update the bit variables sequentially according to the following formula: ， in, , Representation matrix The List, Representation matrix The One diagonal element, This indicates projecting the input element onto... interval, and The solution depends on the modulation order; (4) Update the auxiliary variable according to the following formula. : ， in, This means projecting each element of the input vector onto... interval; (5) Update the dual variable according to the following formula. : 。 6. A trainable joint channel estimation, detection, and decoding method for URLLC-MIMO systems according to claim 5, characterized in that, When using Quadrature Phase Shift Keying (QPSK), at this time , and Calculate using the following formula; ， ， in, express The One element, and These represent taking the real and imaginary parts of the input element, respectively.

7. A trainable joint channel estimation, detection, and decoding method for URLLC-MIMO systems according to claim 5, characterized in that, In step S2, the ADMM solution framework is expanded in depth, and the first... The ADMM iteration corresponds to the [number]th [item]. Layered neural network, and penalty factor and as well as Set as trainable parameters , This represents the number of network layers.

8. A trainable joint channel estimation, detection, and decoding method for URLLC-MIMO systems according to claim 7, characterized in that, In step S3, the offline phase includes the following steps: S301. Generate the dataset required for training. Each set of data includes the network input. , , , The labels used for training, i.e., the actual transmitted codewords. ; S302, the first Output of layered neural networks and tags The mean squared error between the two sides is used as the loss function for training the neural network to obtain trainable parameters.

9. A trainable joint channel estimation, detection, and decoding method for URLLC-MIMO systems according to claim 8, characterized in that, In step S4, during the online phase, the receiving end uses the trained neural network to make a decision on the output of each layer. If the check constraint is satisfied, the processing is terminated early, and the current decision is output; otherwise, the decision corresponding to the output of the last layer is output. Specifically, for the... The output of the layer network is determined according to the following formula: 。

Images