Trainable joint channel estimation, detection and decoding method suitable for urllc-MIMO system
By constructing a neural network model based on MAP criteria for channel estimation and decoding of the URLLC-MIMO system, the performance losses and error propagation problems caused by module division in short packet transmission are solved, and lower latency and efficient channel estimation are achieved. Significantly improves the BLER performance of the system.
Patent Information
- Application Number
- PCT/CN2023/095913
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-05-10
- Filing Date
- 2023-05-23
- Publication Date
- 2024-11-14
AI Technical Summary
The existing URLLC-MIMO system has performance losses and error propagation caused by module division in short packet transmission, and the traditional channel estimation method cannot effectively improve system performance and cannot meet the low latency requirements of URLLC.
Using the joint channel estimation, detection and decoding method based on MAP criteria, by building neural network models and conducting offline training, trainable parameters are generated to realize online channel estimation and decoding, avoiding iterative interaction between modules, and reducing delay and signaling. Overhead.
It significantly improves the BLER performance of the system, reduces system delay and signaling overhead, and improves the overall performance of the URLLC-MIMO system.
Smart Images

Figure CN2023095913_14112024_PF_FP_ABST
Abstract
Description
Trainable joint channel estimation, detection and decoding method for URLLC-MIMO systems Technical Field
[0001] The present invention relates to the field of wireless communications, and in particular to a trainable joint channel estimation, detection and decoding method applicable to a URLLC-MIMO system. Background Art
[0002] Fifth-generation mobile communications (5G) are driving a new communication scenario—Ultra-Reliable Low-Latency Communications (URLLC)—to support emerging services such as telemedicine and industrial automation. To meet the stringent low-latency requirements of URLLC, systems typically opt for short packet transmission. MIMO (Multiple-Input Multiple-Output) technology, on the other hand, can improve system transmission reliability by leveraging spatial redundancy. Consequently, short packet MIMO transmission is gaining increasing attention. However, existing research primarily focuses on theoretical performance analysis under limited code lengths, and relatively little research has considered reception algorithms for URLLC-MIMO systems. Furthermore, the limited decoding capabilities of error-correcting codes with shorter code lengths pose challenges to receiver design.
[0003] Traditional turbo receivers improve performance by exchanging additional information between different modules, and have been 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 will cause unfavorable positive feedback in BP decoding, resulting in error propagation in turbo iterations and impaired performance. On the other hand, in order to ensure the transmission efficiency of short packets, the pilot length is often limited, which also limits the performance of pilot-based channel estimation, thereby affecting the overall system performance. In addition, the interaction between turbo receiver modules will cause non-negligible delay, which does not meet the requirements of URLLC.
[0004] Summary of the Invention
[0005] The technical problem to be solved by the present invention is as follows: The present invention aims to provide a trainable joint channel estimation, detection and decoding method suitable for URLLC-MIMO system, so as to avoid the performance loss and error propagation caused by module partitioning, significantly improve the system BLER performance, reduce system delay and save signaling overhead.
[0006] The technical solution adopted by the present invention to solve the above technical problems is:
[0007] 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 then divides them equally among 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 the adjustable parameters offline, and finally uses the trained neural network to obtain codeword bits online. The specific steps include:
[0008] (1) Construct a joint channel estimation, detection and decoding problem model and solution framework based on the MAP criterion;
[0009] (2) Build a neural network with a deep unfolding framework for solving the joint channel estimation, detection, and decoding problems, and set the adjustable parameters as trainable parameters;
[0010] (3) Offline stage: Generate the data set required for training, train the neural network, and obtain trainable parameters;
[0011] (4) Online stage: The receiving end uses the trained neural network to judge the output of each layer of the network. If the verification constraint is met, the processing is terminated early and the current judgment is output. Otherwise, the judgment corresponding to the output of the last layer of the network is output.
[0012] Furthermore, in the trainable joint channel estimation, detection and decoding method applicable to the URLLC-MIMO system, it is assumed that a URLLC-MIMO system is configured with N t Transmitting antennas and N r Root receiving antenna, channel coefficient is T P +T D The transmission time of the symbols remains unchanged, where T P and T D Indicates the number of pilot symbols and the number of data symbols. In the data transmission order, the transmitter first encodes the information bits into an LDPC codeword with a code length of N b=[b1,b2,…,b N ] T Let H∈[0,1] M×N represents the binary check matrix of the LDPC code, where M is the number of check nodes. Then, the transmitter maps the codeword to the modulation symbol x=[x1,x2,…,x N / Q ] T , and use multi-stream multiplexing technology to divide it equally into N t The pilot matrix and data matrix are recorded as and Where TD =N / (Q×N t ), the received signal can be expressed as: [Y P ,Y D ]=G[P,D]+N,
[0013] in, and represent the pilot reception matrix and data reception matrix respectively. represents a Gaussian channel matrix, where each element is independent and identically distributed, with mean 0 and variance 1. Represents an additive white Gaussian noise matrix, where each element is independent and identically distributed, with mean 0 and variance σ 2 . Note that the (k,t)th element of D corresponds to the modulation symbol In addition, for the convenience of subsequent representation, D is recorded as f(b), f(·) represents the bit-symbol mapping function. Also record the vectorized form of Y P , Y D and G are y P ,y D and g.
[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] Where [·]2 represents the modulo 2 operation, Indicates dimension N r ×N r The identity matrix, H ji represents the (j,i)th element of the LDPC code check matrix H.
[0018] Furthermore, in step (1), the original problem is rewritten as follows:
[0019] Optimization goal:
[0020] Constraints:
[0021] Among them, α>0 is the penalty factor introduced, represents the set of variables involved in the j-th check formula, Indicates that the potential is odd To facilitate subsequent representation, the first constraint is rewritten into the following matrix form according to the element correspondence rule: Ab≥θ,
[0022] in, and denote the weight matrix and bias matrix respectively.
[0023] Furthermore, in step (1), an auxiliary variable is introduced Dual variable and penalty factor μ>0, the rewritten problem is solved based on the alternating direction method of multipliers (ADMM), where the nth ADMM iteration includes the following steps:
[0024] (1.1) Calculate intermediate variables
[0025] (1.2) Calculate the intermediate variable S n-1 :
[0026] in, λ max ( · ) means taking the maximum eigenvalue of the input matrix;
[0027] (1.3) For k = 1, 2, … ,N t ,t=1,2,…,T D , update the bit variables in sequence according to the following formula:
[0028] in, The matrix A represents the List, Represents the matrix A T A's i-th t,k,q diagonal elements, ∏ [0,1] (·) means projecting the input elements into the interval [0,1]. In addition, and The solution is related to the modulation order. Taking QPSK as an example, Q=2, and It can be calculated as follows;
[0029] in, Indicates S n-1 The (k, t)th element of , Re{·} and Im{·} respectively represent the real and imaginary parts of the input element;
[0030] (1.4) Update the auxiliary variable z according to the following formula n :
[0031] in, Indicates that each element of the input vector is projected to the interval [0,∞];
[0032] (1.5) Update the dual variable η according to the following formula n : η n =η n-1 +Ab n +z n -θ.
[0033] Furthermore, in step (2), the ADMM solution framework described in claim 5 is deeply expanded, the nth ADMM iteration corresponds to the lth layer of the neural network, and the penalty factors α, μ and λ are set as trainable parameters L is the number of network layers.
[0034] Furthermore, in step (3), the offline phase includes the following steps:
[0035] (3.1) Generate the data set required for training. Each set of data includes the network input P, Y P , Y D , G and the label used for training, that is, the real sent codeword b.
[0036] (3.2) The output b of the L-th layer neural network L The mean square error between label b and label b is used as the loss function for training, and the neural network is trained to obtain trainable parameters.
[0037] Furthermore, in step (4), in the online phase, the receiving end uses the trained neural network to judge the output of each layer of the network. If the check constraint is satisfied, the processing is terminated early and the current judgment is output. Otherwise, the judgment corresponding to the output of the last layer of the network is output. For the output of the lth layer of the network, the judgment is made according to the following formula:
[0038] Compared with the prior art, the present invention has the following advantages and beneficial technical effects:
[0039] (1) The present invention provides a trainable joint channel estimation, detection and decoding method for URLLC-MIMO system, which avoids the performance loss and error propagation caused by module partitioning and can significantly improve the BLER performance of the system.
[0040] (2) Compared with the traditional receiving method, the present invention avoids iterative interaction between different modules, reduces system delay, and saves signaling overhead. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] FIG1 is a flow chart of a trainable joint channel estimation, detection and decoding method applicable to a URLLC-MIMO system provided by the present invention.
[0042] FIG2 is a diagram of a network structure provided by the present invention.
[0043] FIG3 is a schematic diagram of the results of a simulation experiment of the present invention provided in the examples. DETAILED DESCRIPTION
[0044] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:
[0045] The present invention provides a trainable joint channel estimation, detection and decoding method applicable to a URLLC-MIMO system, which can significantly improve the system BLER performance, reduce system delay and save signaling overhead.
[0046] In order to better understand the above technical solution, the above technical solution will be described in detail below with reference to the accompanying drawings and specific implementation methods.
[0047] In order to verify the performance results of the trainable joint channel estimation, detection and decoding method for URLLC-MIMO system provided by the present invention, it is necessary to build a URLLC-MIMO system simulation platform. In this embodiment, it is assumed that a URLLC-MIMO system is configured with N t Transmitting antennas and N r Root receiving antenna, channel coefficient is T P +T D The transmission time of the symbols remains unchanged, where T P and T D Indicates the number of pilot symbols and the number of data symbols. In the data transmission order, the transmitter first encodes the information bits into an LDPC codeword with a code length of N b=[b1,b2,…,b N ] T Let H∈[0,1] M×N represents the binary check matrix of the LDPC code, where M is the number of check nodes. Then, the transmitter maps the codeword to the modulation symbol x=[x1,x2,…,x N / Q ] T , and use multi-stream multiplexing technology to divide it equally into N t The pilot matrix and data matrix are recorded as and Where T D =N / (Q×N t ), the received signal can be expressed as: [Y P ,Y D ]=G[P,D]+N,
[0048] in, and represent the pilot reception matrix and data reception matrix respectively. represents a Gaussian channel matrix, where each element is independent and identically distributed, with mean 0 and variance 1. Represents an additive white Gaussian noise matrix, where each element is independent and identically distributed, with mean 0 and variance σ 2 . Note that the (k,t)th element of D corresponds to the modulation symbol In addition, for the convenience of subsequent representation, D is recorded as f(b), f(·) represents the bit-symbol mapping function. Also record the vectorized form of Y P , Y D and G are y P ,y D and g.
[0049] As shown in Figure 1, this method is carried out according to the following steps:
[0050] Step (1): Construct the following joint channel estimation, detection and decoding problem model based on the MAP criterion:
[0051] Optimization goal:
[0052] Constraints:
[0053] Where [·]2 represents the modulo 2 operation, Indicates dimension N r ×N r The identity matrix, H ji represents the (j,i)th element of the LDPC code check matrix H.
[0054] To facilitate the solution, the original problem is rewritten as follows:
[0055] Optimization goal:
[0056] Constraints:
[0057] Among them, α>0 is the penalty factor introduced, represents the set of variables involved in the j-th check formula, Indicates that the potential is odd To facilitate subsequent representation, the first constraint is rewritten into the following matrix form according to the element correspondence rule: Ab≥θ,
[0058] in, and denote the weight matrix and bias matrix respectively.
[0059] Next, we introduce auxiliary variables Dual variable and penalty factor μ>0, the rewritten problem is solved based on the alternating direction method of multipliers (ADMM), where the nth ADMM iteration includes the following steps:
[0060] (1.1) Calculate intermediate variables
[0061] (1.2) Calculate the intermediate variable S n-1 :
[0062] in, λ max (·) means taking the maximum eigenvalue of the input matrix.
[0063] (1.3) for k = 1, 2, ..., N t ,t=1,2,…,T D , update the bit variables in sequence according to the following formula:
[0064] in, represents the i-th t,k,q List, Represents the matrix A T A's i-th t,k,q diagonal elements, ∏ [0,1] (·) indicates that the input element is projected to the interval [0,1]. In addition, under the QPSK modulation considered in this embodiment, and It can be calculated as follows;
[0065] Under 16-QAM (quadrature amplitude modulation), and It can be calculated as follows;
[0066] in, Indicates S n-1 The (k, t)th element of Re{ ·} and Im{·} respectively represent taking the real part and imaginary part of the input element.
[0067] (1.4) Update the auxiliary variable z according to the following formula n :
[0068] in, It means that each element of the input vector is projected to the interval [0,∞].
[0069] (1.5) Update the dual variable η according to the following formula n : η n =η n-1 +Ab n +z n -θ.
[0070] Step (2): Deeply expand the ADMM solution framework, correspond the nth ADMM iteration to the lth layer of the neural network, and set the penalty factors α, μ and λ as trainable parameters L is the number of network layers. Specifically, the structure of the l-th layer neural network is shown in Figure 2.
[0071] Step (3): The offline phase includes the following steps:
[0072] (3.1) Generate the data set required for training. Each set of data includes the network input P, Y P , Y D , G and the label used for training, that is, the real sent codeword b.
[0073] (3.2) The output b of the L-th layer neural network L The mean square error between label b and label b is used as the loss function for training, and the neural network is trained to obtain trainable parameters.
[0074] Step (4): In the online phase, the receiver uses the trained neural network to make judgments on the output of each layer of the network. If the checksum constraint is satisfied, the processing is terminated early and the current judgment is output. Otherwise, the judgment corresponding to the output of the last layer of the network is output. More specifically, for the output of the lth layer of the network, the judgment is made according to the following formula:
[0075] We conducted simulation experiments on this embodiment to evaluate the superiority of the present invention, using a MAP-Turbo receiver and an MMSE-Turbo receiver that consider data-assisted channel estimation in traditional methods as comparison objects. The simulation parameters are shown in the following table:
[0076] Table 1 Simulation parameter list
[0077] The horizontal axis (SNR) represents the signal-to-noise ratio, and the vertical axis (BLER) represents the block error rate. The simulation results of the above-mentioned specific embodiment are shown in Figure 3. The simulation results show that under limited pilot overhead, the trainable joint channel estimation, detection, and decoding method proposed in this invention has significant BLER performance advantages over traditional MAP-Turbo receivers and MMSE-Turbo receivers that consider data-aided channel estimation.
[0078] The above description is merely a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any modification or equivalent variation 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 for URLLC-MIMO systems, It is characterized in that The specific steps include: S1. Construct a joint channel estimation, detection and decoding problem model and solution framework based on the MAP criterion; S2, building a neural network with a deep unfolding joint channel estimation, detection and decoding problem solving framework, and setting the adjustable parameters therein as trainable parameters; S3, offline stage: Generate the data set required for training, train the neural network, and obtain trainable parameters; S4, online stage: The receiving end uses the trained neural network to judge the output of each layer of the network. If the verification constraint is met, the processing is terminated early and the current judgment is output. Otherwise, the judgment corresponding to the output of the last layer of the network is output.
2. A trainable joint channel estimation, detection and decoding method suitable for URLLC-MIMO system according to claim 1, It is characterized in that Before executing step S1, the method further includes: configuring N in the URLLC-MIMO system t Transmitting antennas and N r Root receiving antenna, channel coefficient is T P +T D The transmission time of a symbol remains unchanged, where T P and T D Indicates the number of pilot symbols and the number of data symbols. During the data transmission phase, an LDPC codeword is mapped to modulation symbols and then equally divided and sent to different transmitting antennas.
3. A trainable joint channel estimation, detection and decoding method suitable for URLLC-MIMO system according to claim 2, It is characterized in that After mapping an LDPC codeword into modulation symbols, the symbols are equally divided and sent to different transmitting antennas, as follows: First, the transmitter encodes the information bits into an LDPC codeword b=[b 1 ,b 2 ,…,b N ] T , let H∈[0,1] M×N represents the binary check matrix of the LDPC code, where M is the number of check nodes; The transmitter then maps the codeword into modulation symbols And use multi-stream multiplexing technology to divide it equally into N t The signal is transmitted on the root transmitting antenna, where Q is the modulation order; The pilot matrix and data matrix are denoted as and Where T D =N / (Q×N t ), the received signal can be expressed as: [Y P ,Y D ]=G[P,D]+N, in, and denote the pilot reception matrix and the data reception matrix respectively; represents a Gaussian channel matrix, where each element is independent and identically distributed, with a mean of 0 and a variance of 1; Represents an additive white Gaussian noise matrix, where each element is independent and identically distributed, with a mean of 0 and a variance of σ 2 ; Note that the (k,t)th element of D corresponds to the modulation symbol Let D be denoted as f(b), and f(·) denotes the bit-to-symbol mapping function.
4. A trainable joint channel estimation, detection and decoding method suitable for URLLC-MIMO system according to claim 3, It is characterized in that In step S1, the following joint channel estimation, detection and decoding problem model based on the MAP criterion is constructed: Optimization goal: Constraints: in[·] 2 represents the modulo 2 operation, y P ,y D and g are the vectorized forms of Y P , Y D and G, Indicates that the dimension is N r ×N r The identity matrix, H ji represents the (j,i)th element of the LDPC code check matrix H.
5. A trainable joint channel estimation, detection and decoding method suitable for URLLC-MIMO system according to claim 4, It is characterized in that In step S1, the original problem is rewritten as follows: Optimization goal: Constraints: Among them, α>0 is the penalty factor introduced, represents the set of variables involved in the jth check formula, represents the condition that the potential is odd The set of subsets of ; According to the element correspondence rule, rewrite the first constraint into the following matrix form: Ab ≥ θ, in, and denote the weight matrix and bias matrix respectively.
6. A trainable joint channel estimation, detection and decoding method suitable for URLLC-MIMO system according to claim 5, It is characterized in that In step S1, an auxiliary variable is introduced Dual Variable and penalty factor μ>0, the rewritten problem is solved based on the alternating direction multiplier method ADMM, where the nth ADMM iteration includes the following steps: (1) Calculate intermediate variables (2) Calculate the intermediate variable S n-1 : in, λ max (·) means taking the maximum eigenvalue of the input matrix; (3) For k = 1, 2, ..., N t ,t=1,2,…,T D , update the bit variables in sequence according to the following formula: in, represents the i-th t,k,q List, Represents the matrix A T A's i t,k,q diagonal elements, Π [0,1] (·) means projecting the input elements to the interval [0,1], and The solution of is related to the modulation order; (4) Update the auxiliary variable z according to the following formula n : in, It means that each element of the input vector is projected to the interval [0,∞]; (5) Update the dual variable η according to the following formula n : or n =the n-1 +Ab n +z n -θ.
7. A trainable joint channel estimation, detection and decoding method suitable for URLLC-MIMO system according to claim 6, It is characterized in that When using the quadrature phase shift keying QPSK method, Q = 2, and Calculate as follows; in, Indicates S n-1 The (k, t)th element of , Re{·} and Im{·} represent the real and imaginary parts of the input elements, respectively.
8. A trainable joint channel estimation, detection and decoding method suitable for URLLC-MIMO system according to claim 6, It is characterized in that In step S2, the ADMM solution framework is deeply expanded, the nth ADMM iteration corresponds to the lth layer of the neural network, and the penalty factors α, μ and λ are set as trainable parameters L is the number of network layers.
9. A trainable joint channel estimation, detection and decoding method suitable for URLLC-MIMO system according to claim 8, It is characterized in that In step S3, the offline phase includes the following steps: S301, generate the data set required for training, each set of data includes the network input P, Y P , Y D , G and the label used for training, that is, the real sent codeword b; S302, the output b of the Lth layer neural network L The mean square error between label a and label b is used as the loss function for training, and the neural network is trained to obtain trainable parameters.
10. A trainable joint channel estimation, detection and decoding method suitable for URLLC-MIMO system according to claim 9, It is characterized in that In step S4, in the online stage, the receiving end uses the trained neural network to judge the output of each layer of the network. If the verification constraint is met, the processing is terminated in advance and the current judgment is output. Otherwise, the judgment corresponding to the output of the last layer of the network is output. For the output of the lth layer of the network, the judgment is made according to the following formula:
Citation Information
Patent Citations
MIMO-OFDM system channel estimation method based on deep neural network
CN113472706A
Symbol detection method based on deep learning in MIMO communication system
CN115664893A
Cited By
Cellular-free large-scale MIMO system multi-user joint decoding method based on model-driven deep learning
CN121125017A
Ultrahigh code rate demodulation method and system based on neural network
CN121690929A
