A large-scale passive address random access method based on decoupling structure

Abstract

The application discloses a large-scale passive address random access method based on decoupling structure. In a wireless network, a multi-antenna base station simultaneously serves a large number of single-antenna mobile devices. In any time slot, only part of the devices are activated and have information to be sent to the base station, and the other devices are in a dormant state. Based on the decoupling structure of the passive address random access protocol, all activated devices use the same and known codebook of the base station, cut the respective long information sequence into multiple sub-blocks, and directly map the code word to the base station through the uplink without adding redundancy. The base station recovers the original information sequence by using a matrix information geometry auxiliary joint code word detection and splicing method. The application provides an efficient large-scale random access method for a wireless network.

Description

Technical Field

[0001] This invention relates to the field of wireless communication, and in particular to a large-scale passive address-free random access method based on a decoupled structure. Background Technology

[0002] In recent years, the number of mobile devices and the data traffic they generate globally has exploded. In this context, next-generation wireless networks face the need to simultaneously support a massive number of wireless terminals accessing the network. Furthermore, in scenarios involving massive machine-type communication services, because only a small number of terminal devices are active and need to send and receive information within a given time interval, while other devices are temporarily in a dormant state to conserve energy, data requests from terminal devices in the network are typically sporadic.

[0003] These network characteristics mean that achieving the Internet of Things still faces some problems and challenges. On the one hand, in existing active address random access schemes, activating devices need to transmit their own unique pilot sequences to the base station beforehand. The base station uses these pilot sequences to obtain the activation status and corresponding channel information of the devices. To obtain accurate activation and channel information, devices need to send very long pilot sequences. Given the large number of devices and occasional data traffic, this leads to a waste of limited spectrum resources and excessive computational complexity. On the other hand, in current passive address random access schemes based on coupled structures, activating devices add redundancy to recombine segmented messages. Although the additional redundancy introduced is shorter than the pilot sequence length in active address random access schemes, it still leads to a reduction in the system's spectral efficiency.

[0004] Therefore, to address the above issues, it is necessary to design a novel passive address-free random access method that incorporates these characteristics. This invention studies a large-scale passive address-free random access technology based on a decoupled structure. All devices use the same codebook to send data information to the base station, eliminating the need to pre-send pilot sequences for device and channel detection. Furthermore, segmented messages do not require additional redundancy for information coupling to achieve message concatenation. Based on this, service latency and wireless resource consumption can be significantly reduced, data processing efficiency in the era of the Internet of Things can be improved, and it is expected to solve a series of challenges in enabling massive mobile devices to access wireless networks. Summary of the Invention

[0005] To address the problems of large connection capacity, high communication latency, and low spectrum efficiency when large-scale mobile devices access wireless networks, this invention proposes a large-scale passive address-free random access method based on a decoupled structure.

[0006] The specific technical solution adopted in this invention is as follows:

[0007] A large-scale passive address random access method based on a decoupled structure includes the following steps:

[0008] S1: In a wireless network, a base station with M antennas is pre-deployed, and K antennas are also present. t A potential single-antenna mobile device accesses the wireless network through the base station;

[0009] In a given time slot, only K a One device is active, among which K a <<K t And record the set of activated devices as

[0010] All activated devices will divide their respective b-bit information into equal segments of equal length, ensuring that each segment contains J bits of information, resulting in a total of L = b / J sub-blocks;

[0011] S2: The given time slot is divided into L sub-time slots without intervals. All activated devices are based on the same codebook known to the base station. They map the L sub-blocks to codewords in the codebook and send them to the base station in sequence through the L sub-time slots.

[0012] S3: In the l∈[1,L]th sub-slot, after the base station receives the data, it uses a matrix information geometric-assisted joint codeword detection and splicing method to estimate the estimated value of the activation codeword list of the lth sub-slot, and then splices the activation codeword of the lth sub-slot with the activation codewords from the same device in the previous l-1 sub-slots in order of reception time.

[0013] S4: Demap the concatenated codeword sequence to recover the original information sent by each activated device.

[0014] Based on the above technical solution, some of the steps can be implemented in the following preferred manner.

[0015] Preferably, in step S2, the method by which all activated devices map the L sub-blocks to codewords in the codebook based on the same codebook known to the base station is as follows:

[0016] S21: Given a sub-slot length of n0, set the codebook matrix. in Representing the complex field, each column c of the codebook matrix j Both represent a codeword, where j∈[1,2] J There are a total of 2 J Each codeword satisfies the following constraints. Where ||·||2 represents the 2-norm of the vector;

[0017] S22: In the l-th sub-slot, any activated device k maps the J bits of information of the l-th sub-block to be sent to an integer i. k,l ∈[1,2J ], and define the list of activation code words. in l∈[1,L];

[0018] S23: Activate device k and the i-th element of codebook matrix C k,l The column is the codeword transmitted by the activating device k in the l-th sub-time slot. The codeword is then sent to the base station.

[0019] As a preferred embodiment, in step S3, the joint codeword detection and splicing method with matrix information geometry assistance is as follows:

[0020] S31: Input codebook C, maximum value of iteration number t T iter Initialize the sub-slot index l = 1; input the received signal from the base station of the l-th sub-slot. in The channel vector of the activated device k follows a complex Gaussian distribution. in It is the large-scale fading coefficient that is greater than zero, I M It is an identity matrix of dimension M×M; (·) T Indicates matrix transpose; Let represent the large-scale fading diagonal matrix of the l-th sub-slot, and let diag(·) represent the matrix formed with the input vector as the diagonal elements, where It is the codeword activation vector of the l-th sub-slot, and satisfies And when i k,l =j when θ k,j =1, otherwise θ k,j =0, where j∈[1,2] J ]; W represents the small-scale fading matrix, whose elements follow a complex Gaussian distribution with zero mean and unit variance. l W represents the additive white Gaussian noise matrix. l Each element has a mean of zero and a noise variance of σ. 2 The complex Gaussian distribution of the received signal Y; l Each column Where Y l covariance matrix Ψ l Defined as It is an identity matrix of dimension n0×n0, (·) H Let E{·} denote the conjugate transpose of a matrix, and let E{·} denote the expected value, where m∈[1,M].

[0021] S32: Order X is a sparse codeword state matrix. l Non-zero row vector Follows a complex Gaussian distribution The covariance matrix of a non-zero row vector is defined as in It is a Hermitian positive definite matrix manifold; A > 0 indicates that matrix A is a positive definite matrix;

[0022] S33: Initialize the codeword activation vector estimate for the l-th sub-slot. Calculate the sample covariance matrix of the received signal.

[0023] S34: Calculate the approximate maximum likelihood initial estimate of the codeword activation vector. Initialize the number of iterations n = 1; initialize the covariance matrix. and

[0024] In the nth iteration, the coordinate index r∈[1,2] is randomly selected. J ], calculate intermediate variables Where c r It is the r-th column of the codebook matrix. yes Update the r-th element. Update covariance matrix Randomly select the next coordinate index r and let n = n + 1. Repeat the calculation of intermediate variables and update parameters until all coordinate indices have been traversed at least once.

[0025] S35: Determine the sub-slot index. If l = 1, proceed to step S36; otherwise, proceed to step S37.

[0026] S36: Let the codeword activation vector estimate of the l-th sub-slot be...

[0027] According to a hard thresholding method, the base station... The decision is made to obtain an estimate of the list of activation codewords transmitted in the current sub-slot. Let the estimated number of activated users be It equals the size of the estimated value of the activation codeword list, i.e.

[0028] make Each activation code is in its own category, totaling [number missing]. Different classes in

[0029] No. The covariance matrix corresponding to the activation codeword in the class Form a point on the Hermitian positive definite matrix manifold, and use it as the first... Geometric center of class in yes The jth p There are 10 elements, where ∩ represents the intersection of the two sets;

[0030] Let l = l + 1, then return to step S31 to continue execution;

[0031] S37: Compile a local estimator based on covariance Where ln(·) represents the natural logarithm, and tr(·) represents the trace of the matrix;

[0032] Calculate the sparsity estimator g(γ) l )=||γ l ||1, where ||·||1 represents the 1-norm of the vector;

[0033] Computation of the classification estimator The geodesic distance is defined as ·|| F represents the F-norm of the matrix, and Log(·) represents the logarithm of the square matrix;

[0034] Define the objective function p(γ) l ) is p(γ l )=f(γ l )+αg(γ l )+βφ(γ l ), where α and β are penalty factors;

[0035] The objective function value p(γ) is obtained using a proximal gradient iteration method. l The smallest solution is obtained.

[0036] According to the hard threshold method, the base station... The decision is made to obtain the final estimated value of the activation codeword list.

[0037] Calculate activation code The ultimate target index That is, determine the activation code. Belongs to the kind in yes The One element;

[0038] right Update # Geometric center of class in

[0039] Let l = l + 1. If l ≤ L, return to step S31 and execute; otherwise, execute S38.

[0040] S38: Concatenate the sequence numbers of L activation codewords belonging to the same class according to the order of their received codewords.

[0041] Preferably, in steps S36 and S37, the hard thresholding method is as follows:

[0042] Based on codeword activation vector estimation Using judgment criteria This indicates that the condition is met. All values ​​of q are set The elements are used to obtain the estimated value of the codeword activation list for the current sub-slot. in It is the decision threshold. yes The q-th element.

[0043] Preferably, in step S37, the proximal gradient iteration method is as follows:

[0044] a) Initialize the iteration count t = 1; initialize initialization Initialize error vector u t | t＝1 =0;

[0045] b): According to the hard thresholding method, the base station... The decision is made to obtain the estimated value of the activation codeword list for the t-th iteration.

[0046] Calculate the minimum geodesic distance Among them, the first Temporary target class index of codeword in the t-th iteration in yes The One element;

[0047] c): Computational estimator gradient vector in yes The j-th element, j∈[1,2] J ];

[0048] Define auxiliary variables in is the step size of the t-th iteration, α and β are the penalty factors, and sign(·) denotes the sign function;

[0049] d): Calculate the value of the (t+1)th iteration. in For j∈[1,2] J ],when otherwise in

[0050] e): Update Update error vector in yes The j-th element, j∈[1,2] J ];

[0051] f): Judgment Whether it is satisfied, among which It refers to convergence accuracy;

[0052] If this condition is not met, then in iteration number t = t + 1 and t ≤ T iter Repeat steps b)-f) if the condition is met or t>T. iter Output codeword activation vector estimation

[0053] Compared with the prior art, the present invention has the following advantages:

[0054] The large-scale passive address-free random access method based on a decoupled structure proposed in this invention solves the problems of large connection capacity, high communication latency, and low spectrum utilization caused by massive mobile devices accessing wireless networks. The matrix information-based geometrically assisted joint codeword detection and splicing method proposed in this invention has advantages such as low access latency and high spectrum efficiency. Attached Figure Description

[0055] Figure 1 This is a system diagram of a large-scale passive address random access method based on a decoupled structure;

[0056] Figure 2 The convergence performance of the method of the present invention is shown under different base station antenna numbers, that is, the mean square error (MSE) of the codeword activation vector estimate varies with the number of iterations (antenna numbers are 16, 32 and 64 respectively).

[0057] Figure 3 To compare the bit error rate (BER) performance of the method of the present invention under different base station antenna numbers and different device transmit signal-to-noise ratios (SNR) (antenna numbers are 32 and 64, and SNR ranges from -5dB to 20dB). Detailed Implementation

[0058] The present invention will be further described and illustrated below with reference to the accompanying drawings and specific embodiments. The technical features of each embodiment of the present invention can be combined accordingly, provided that there is no mutual conflict.

[0059] In this embodiment, a system block diagram of a large-scale passive address-free random access method based on a decoupling structure is shown below. Figure 1 As shown, the base station is equipped with M antennas. In any given time slot, K a Each activated mobile device with a single antenna accesses the wireless network through this base station. Based on a source-free random access protocol, all activated devices segment the information to be transmitted, then use the same codebook known to the base station to map the fragmented information into codewords, and then send them to the base station sequentially uplink according to sub-time slots. After receiving the codeword information superimposed on the wireless channel, the base station uses a matrix information geometry-assisted joint codeword detection and splicing method to recover the original information of the activated devices.

[0060] The specific technical solution adopted in this embodiment is as follows:

[0061] A large-scale passive address-free random access method based on a decoupled structure includes the following steps:

[0062] S1: In a wireless network, a base station with M antennas is pre-deployed, and K antennas are also present. t A potential single-antenna mobile device accesses the wireless network through the base station;

[0063] In a given time slot, only K a One device is active, among which K a <<K t And record the set of activated devices as

[0064] All activated devices will divide their respective b-bit information into equal segments of equal length, ensuring that each segment contains J bits of information, resulting in a total of L = b / J sub-blocks;

[0065] S2: The given time slot is divided into L sub-time slots without intervals. All activated devices are based on the same codebook known to the base station. They map the L sub-blocks to codewords in the codebook and send them to the base station in sequence through the L sub-time slots.

[0066] In this step, all activated devices map L sub-blocks to codewords in the codebook based on the same codebook known to the base station as follows:

[0067] S21: Given a sub-slot length of n0, set the codebook matrix. in Representing the complex field, each column c of the codebook matrix j Both represent a codeword, where j∈[1,2]J There are a total of 2 J Each codeword satisfies the following constraints. Where ||·||2 represents the 2-norm of the vector;

[0068] S22: In the l-th sub-slot, any activated device k maps the J bits of information of the l-th sub-block to be sent to an integer i. k,l ∈[1,2 J ], and define the list of activation code words. in l∈[1,L];

[0069] S23: Activate device k and the i-th element of codebook matrix C k,l The column is the codeword transmitted by the activating device k in the l-th sub-time slot. And send the codeword to the base station;

[0070] S3: In the l∈[1,L]th sub-slot, after the base station receives the data, it uses a matrix information geometric-assisted joint codeword detection and splicing method to estimate the estimated value of the activation codeword list of the lth sub-slot, and then splices the activation codeword of the lth sub-slot with the activation codewords from the same device in the previous l-1 sub-slots in order of reception time.

[0071] In this step, the matrix information-based geometrically assisted joint codeword detection and concatenation method is as follows:

[0072] S31: Input codebook C, maximum value of iteration number t T iter Initialize the sub-slot index l = 1; input the received signal from the base station of the l-th sub-slot. in The channel vector of the activated device k follows a complex Gaussian distribution. in It is the large-scale fading coefficient that is greater than zero, I M It is an identity matrix of dimension M×M; (·) T Indicates matrix transpose; Let represent the large-scale fading diagonal matrix of the l-th sub-slot, and let diag(·) represent the matrix formed with the input vector as the diagonal elements, where It is the codeword activation vector of the l-th sub-slot, and satisfies And when i k,l =j when θ k,j =1, otherwise θ k,j =0, where j∈[1,2] J ]; W represents the small-scale fading matrix, whose elements follow a complex Gaussian distribution with zero mean and unit variance. lW represents the additive white Gaussian noise matrix. l Each element has a mean of zero and a noise variance of σ. 2 The complex Gaussian distribution of the received signal Y; l Each column Where Y l covariance matrix Ψ l Defined as It is an identity matrix of dimension n0×n0, (·) H Let E{·} denote the conjugate transpose of a matrix, and let E{·} denote the expected value, where m∈[1,M].

[0073] S32: Order X is a sparse codeword state matrix. l Non-zero row vector Follows a complex Gaussian distribution The covariance matrix of a non-zero row vector is defined as in It is a Hermitian positive definite matrix manifold; A > 0 indicates that matrix A is a positive definite matrix;

[0074] S33: Initialize the codeword activation vector estimate for the l-th sub-slot. Calculate the sample covariance matrix of the received signal.

[0075] S34: Calculate the approximate maximum likelihood initial estimate of the codeword activation vector. Initialize the number of iterations n = 1; initialize the covariance matrix.

[0076] In the nth iteration, the coordinate index r∈[1,2] is randomly selected. J ], calculate intermediate variables Where c r It is the r-th column of the codebook matrix. yes Update the r-th element. Update covariance matrix Randomly select the next coordinate index r and let n = n + 1. Repeat the calculation of intermediate variables and update parameters until all coordinate indices have been traversed at least once.

[0077] S35: Determine the sub-slot index. If l = 1, proceed to step S36; otherwise, proceed to step S37.

[0078] S36: Let the codeword activation vector estimate of the l-th sub-slot be...

[0079] According to a hard thresholding method, the base station... The decision is made to obtain an estimate of the list of activation codewords transmitted in the current sub-slot. Let the estimated number of activated users be It equals the size of the estimated value of the activation codeword list, i.e.

[0080] make Each activation code is in its own category, totaling [number missing]. Different classes in

[0081] No. The covariance matrix corresponding to the activation codeword in the class Form a point on the Hermitian positive definite matrix manifold, and use it as the first... Geometric center of class in yes The jth p There are 10 elements, where ∩ represents the intersection of the two sets;

[0082] Let l = l + 1, then return to step S31 to continue execution;

[0083] S37: Compile a local estimator based on covariance Where ln(·) represents the natural logarithm, and tr(·) represents the trace of the matrix;

[0084] Calculate the sparsity estimator g(γ) l )=||γ l ||1, where ||·||1 represents the 1-norm of the vector;

[0085] Computation of the classification estimator The geodesic distance is defined as ||·|| F represents the F-norm of the matrix, and Log(·) represents the logarithm of the square matrix;

[0086] Define the objective function p(γ) l ) is p(γ l )=f(γ l )+αg(γ l )+βφ(γ l ), where α and β are penalty factors;

[0087] The objective function value p(γ) is obtained using a proximal gradient iteration method. l The smallest solution is obtained.

[0088] According to the hard threshold method, the base station... The decision is made to obtain the final estimated value of the activation codeword list.

[0089] Calculate activation code The ultimate target index That is, determine the activation code. Belongs to the kind in yes The One element;

[0090] right Update # Geometric center of class in

[0091] Let l = l + 1. If l ≤ L, return to step S31 and execute; otherwise, execute S38.

[0092] In steps S36 and S37, the hard thresholding method is as follows:

[0093] Based on codeword activation vector estimation Using judgment criteria This indicates that the condition is met. q∈[1,2 J All possible values ​​of q are set as follows: The elements are used to obtain the estimated value of the codeword activation list for the current sub-slot. in It is the decision threshold. yes The qth element;

[0094] In step S37, the proximal gradient iteration method is as follows:

[0095] a) Initialize the iteration count t = 1; initialize initialization Initialize error vector u t | t＝1 =0;

[0096] b): According to the hard thresholding method, the base station... The decision is made to obtain the estimated value of the activation codeword list for the t-th iteration.

[0097] Calculate the minimum geodesic distance Among them, the first Temporary target class index of codeword in the t-th iteration in yes The One element;

[0098] c): Computational estimator gradient vector in yes The j-th element, j∈[1,2] J ];

[0099] Define auxiliary variables in is the step size of the t-th iteration, α and β are the penalty factors, and sign(·) denotes the sign function;

[0100] d): Calculate the value of the (t+1)th iteration. in For j∈[1,2] J ],when hour otherwise in

[0101] e): Update Update error vector in yes The j-th element, j∈[1,2] J ];

[0102] f): Judgment Whether it is satisfied, among which It refers to convergence accuracy;

[0103] If this condition is not met, then in iteration number t = t + 1 and t ≤ T iter Repeat steps b)-f) in the following cases;

[0104] If the condition is already satisfied or t > T iter Output codeword activation vector estimation

[0105] S38: Concatenate the sequence numbers of L activation codewords belonging to the same class according to the order of their received time.

[0106] S4: Demap the concatenated codeword sequence to recover the original information sent by each activated device.

[0107] To further verify the effectiveness of the method of the present invention, the bit error rate performance of the proposed method was compared with that of the maximum likelihood estimation-K-means classification method, as detailed below:

[0108] As a comparative approach, the maximum likelihood estimation-K-means classification method refers to the present invention where steps S1-S31 and S38-S4 remain unchanged, while steps S32-S37 are changed to: for all sub-slots l∈[1,L], initialize the iteration count v=1; initialize the codeword activation vector estimate of the l-th sub-slot. Covariance Matrix In the v-th iteration, calculate the sample covariance matrix of the received signal. Randomly select coordinate exponent r∈[1,2] J ], calculate intermediate variables Where c r It is the r-th column of the codebook matrix. yes The r-th element; update Update covariance matrix Randomly select the next coordinate index r and repeatedly calculate the intermediate variable update parameters until all coordinate indices have been traversed at least once to obtain the codeword activation vector estimate for the l-th sub-slot. According to the hard thresholding method described in this invention, for Make a judgment to obtain an estimated value of the list of activation codewords. Calculate the total number of classes

[0109] Let the estimated value of the first list of activation codewords be... The codewords in the code are classified into different categories, and the class mean of each category is defined as follows: yes The One element;

[0110] Remove Estimates of the remaining list of activation codewords Each activation codeword in the array is compared sequentially with the class mean. When the minimum is reached, determine the first sub-slot in the l-th time slot. The code belongs to the first Class, and update the class mean of that class to Repeat this process until all activation codes have been categorized.

[0111] Computer simulations show that, Figure 2 As shown, in the large-scale passive address-free random access method based on decoupling structure proposed in this invention, the mean square error (MSE) of the codeword activation vector estimate gradually decreases and converges to a fixed value as iterations proceed. The convergence speed and convergence fixed point of the proposed method differ depending on the number of base station antennas. A larger number of antennas results in a faster convergence speed and a smaller MSE at convergence.

[0112] Figure 3This invention demonstrates that the proposed large-scale passive address-free random access method based on a decoupled structure reduces the bit error rate (BER) in recovering the original information as the device transmit signal-to-noise ratio (SNR) increases, and its performance improves with the increase in the number of antennas. Furthermore, the proposed method outperforms the comparative scheme, namely the maximum likelihood estimation-K-means classification method, under different antenna numbers or SNR values. Therefore, this invention provides an efficient access method for wireless networks with a large number of devices.

[0113] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the invention. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the invention. Therefore, all technical solutions obtained through equivalent substitution or transformation fall within the protection scope of the present invention.

Claims

1. A large-scale passive address random access method based on a decoupled structure, characterized in that, Includes the following steps: S1: In a wireless network, a base station with M antennas is pre-deployed, and simultaneously... A potential single-antenna mobile device accesses the wireless network through the base station; In a given time slot, only One device is in an active state, among which And record the set of activated devices as ; All activated devices will divide their respective b-bit information to be sent into equal segments of equal length, ensuring that each segment contains J bits of information, for a total of [number missing]. Each block; S2: The given time slot is divided into L sub-time slots without intervals. All activated devices are based on the same codebook known to the base station. They map the L sub-blocks to codewords in the codebook and send them to the base station in sequence through the L sub-time slots. S3: In the In each sub-time slot, after the base station receives the data, it uses a matrix information geometric-assisted joint codeword detection and splicing method to estimate the estimated value of the activation codeword list of the l-th sub-time slot, and then splices the activation codeword of the l-th sub-time slot with the activation codewords from the same device in the previous l-1 sub-time slots in order of reception time. S4: Demap the concatenated codeword sequence to recover the original information sent by each activated device; In step S3, the joint codeword detection and concatenation method assisted by matrix information geometry is as follows: S31: Input codebook The maximum value of the number of iterations t Initialize sub-slot index ; Enter the first Received signals at each time slot base station ;in The activation device is The channel vector follows a complex Gaussian distribution. ,in It is a large-scale fading coefficient that is greater than zero. It is a dimension of The identity matrix; Indicates matrix transpose; Indicates the first Large-scale fading diagonal matrix of sub-slots This represents a matrix formed by using the input vector as its diagonal elements, where It is the first The codeword activation vector of each sub-slot, and satisfying , and when hour ,otherwise ,in ; This represents the small-scale fading matrix, whose elements follow a complex Gaussian distribution with zero mean and unit variance. This represents the additive white Gaussian noise matrix. Each element follows a zero mean and a noise variance of . Complex Gaussian distribution; received signal Each column ,in covariance matrix Defined as , It is a dimension of The identity matrix, This represents the conjugate transpose of a matrix. This represents taking the mathematical expectation. ; S32: Order It is a sparse codeword state matrix. Non-zero row vector Follows a complex Gaussian distribution The covariance matrix of a non-zero row vector is defined as ,in ; It is a Hermitian positive definite matrix manifold; This indicates that matrix A is a positive definite matrix; S33: Initialize the codeword activation vector estimate for the l-th sub-slot. Calculate the sample covariance matrix of the received signal. ; S34: Calculate the approximate maximum likelihood initial estimate of the codeword activation vector. ; Initialize the number of iterations Initialize the covariance matrix and ; In the nth iteration, the coordinate index is randomly selected. Calculate intermediate variables ,in It is the r-th column of the codebook matrix. yes Update the r-th element. Update the covariance matrix Randomly select the next coordinate index r and let Repeatedly calculate intermediate variables and update parameters until all coordinate indices have been traversed at least once to obtain the result. ; S35: Determine the sub-slot index, if... If the above steps are not executed, proceed to step S36; otherwise, proceed to step S37. S36: Let the codeword activation vector estimate of the l-th sub-slot be... ; According to a hard thresholding method, the base station... The decision is made to obtain an estimate of the list of activation codewords transmitted in the current sub-slot. Let the estimated number of activated users be It equals the size of the estimated value of the activation codeword list, i.e. ; make Each activation code is in its own category, totaling [number missing]. Different classes ,in ; No. The covariance matrix corresponding to the activation codeword in the class Form a point on the Hermitian positive definite matrix manifold, and use it as the first... Geometric center of class ,in , yes The One element, Represents the intersection of two sets; make Return to step S31 and continue execution; S37: Compile a local estimator based on covariance ,in Represents the natural logarithm. Represents the trace of a matrix; Computational sparse estimator ,in The 1-norm of a vector; Computation of the classification estimator The geodesic distance is defined as , Denotes the F-norm of a matrix. Represents the logarithm of a square matrix; Define the objective function for ,in and It is a punishment factor; The objective function value is obtained using a proximal gradient iteration method. The minimum solution is obtained. ; According to the hard threshold method, the base station... The decision is made to obtain the final estimated value of the activation codeword list. ; Calculate activation code The ultimate target index That is, to determine the activation code. Belongs to the kind ,in , yes The One element; right Update # Geometric center of class ,in , ; make ,when If the condition is met, return to step S31 and execute; otherwise, execute S38. S38: Concatenate the sequence numbers of L activation codewords belonging to the same class according to the order of their received codewords.

2. The large-scale passive address random access method based on a decoupled structure as described in claim 1, characterized in that, In step S2, the method by which all activated devices map L sub-blocks to codewords in the codebook based on the same codebook known to the base station is as follows: S21: Given a sub-slot length of... Set the codebook matrix ,in Representing the complex field, each column of the codebook matrix Both represent a code character, among which There are a total of Each codeword satisfies the following constraints. ,in Denotes the 2-norm of a vector; S22: In the Activate the device arbitrarily within each sub-time slot. The first one to be sent The J-bit information of each sub-block is mapped to an integer. And define the list of activation codes. ,in ; S23: Activate device k to use the codebook matrix The The column is the codeword transmitted by the activating device k in the l-th sub-time slot. And then send the codeword to the base station.

3. The large-scale passive address random access method based on a decoupled structure as described in claim 1, characterized in that, In steps S36 and S37, the hard thresholding method is as follows: Based on codeword activation vector estimation Using judgment criteria , indicating that the condition is met All values ​​of q are set The elements are used to obtain the estimated value of the codeword activation list for the current sub-slot. ,in It is the decision threshold. yes The q-th element.

4. The large-scale passive address random access method based on a decoupled structure as described in claim 3, characterized in that, In step S37, the proximal gradient iteration method is as follows: a): Initialize the number of iterations ;initialization ;initialization Initialize the error vector ; b): According to the hard threshold method, the base station... The decision is made to obtain the estimated value of the activation codeword list for the t-th iteration. ; Calculate the minimum geodesic distance , of which Temporary target class index of codeword in the t-th iteration ,in , yes The One element; c): Computational estimator gradient vector ,in , , yes The One element, ; Define auxiliary variables ,in , It is the step size of the t-th iteration. and It is the aforementioned penalty factor. Represents a symbolic function; d): Calculate the first Next iteration value ,in ;for ,when hour ,otherwise ,in , ; e): Update Update the error vector ,in yes The j-th element, ; f): Judgment Whether it is satisfied, among which It refers to convergence accuracy; If not satisfied, in the number of iterations and Repeat steps b) - f) in the following cases. If the condition has been met or Output codeword activation vector estimation .

Images