A covariance-based non-orthogonal pilot-based active user detection method

Abstract

This invention discloses an active user detection method based on covariance-based nonorthogonal pilots for non-cellular massive MIMO systems, applicable to the communications field. A mathematical model for nonorthogonal random access based on non-cellular massive MIMO is established. By utilizing the block diagonal characteristics of the received signal covariance matrix, the active user detection problem is reconstructed into a maximum likelihood estimation problem. The coordinate descent algorithm is used to solve the reconstructed maximum likelihood estimation problem, and a parallel estimation algorithm is proposed to reduce computation time. This invention achieves a good trade-off between performance and complexity in active user detection, meeting the needs of scenarios with massive terminal random access and showing great application potential.

Description

Technical Field

[0001] This invention relates to the field of wireless communication technology, and more specifically to an active user detection method based on non-orthogonal pilots with covariance. Background Technology

[0002] Supporting massive machine-type communications (MMIs) has become a core focus of future mobile communication development. Unlike traditional communications, MMIs are characterized by a wide variety of IoT applications, sparse activity, large-scale concurrent connections, and short packet transmission. Improving the random access capacity and reliability of communication systems, especially under limited spectrum resources, has become a key research focus. Cellular-free massive MIMO systems, which can improve system reliability through cooperative processing of distributed access points, have become a promising new access network architecture. Compared to traditional multi-cell structures, cellular-free massive MIMO systems offer better macro-diversity gain and higher system capacity. Furthermore, this system exhibits good performance in non-orthogonal random access.

[0003] Due to the massive number of terminals accessing the network in mMTC scenarios, huge spectrum overhead is required. Non-orthogonal random access is a feasible access scheme, but non-orthogonal pilot signals generate strong interference, posing challenges to active user detection and data detection. Current compressed sensing-based methods are difficult to apply to large-scale non-orthogonal random access in terms of both performance and complexity, and therefore unsuitable for mMTC scenarios. Therefore, under the condition of limited spectrum resources, a non-orthogonal random access scheme based on covariance is urgently needed. Summary of the Invention

[0004] Technical issues

[0005] To address the aforementioned issues, this invention proposes an active user detection method based on non-orthogonal pilots with covariance, which is applicable to random access by a large number of terminals and achieves a good trade-off between performance and complexity in active user detection.

[0006] Technical solution

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] An active user detection method based on covariance-based nonorthogonal pilots includes:

[0009] S1. Establish a mathematical model for nonorthogonal random access based on noncellular massive MIMO;

[0010] S2. By utilizing the block diagonal property of the received signal covariance matrix, the active user detection problem is reconstructed into a maximum likelihood estimation problem;

[0011] S3. Based on the non-cellular macro diversity gain, the coordinate descent algorithm is used to solve the maximum likelihood estimation problem after S2 reconstruction;

[0012] S4. To reduce computation time, a parallel solution strategy is proposed;

[0013] Furthermore, the described step S1 specifically includes the following sub-steps:

[0014] S11. Configure a non-cellular massive MIMO nonorthogonal random access system with M access points, K single-antenna users, and a central computing unit. Each access point is equipped with N antennas. Information is transmitted losslessly between the access points and the central computing unit via a fiber optic backhaul link, and all access points and users are randomly distributed. Let K = {1, 2, ..., K} represent the set of all potential users. Let α represent the set of active users. Users within the system are randomly activated according to a certain activation probability ε. A Bernoulli random variable represents whether a user is active, i.e., α. k ∈{0,1} represents whether user k is active, i.e., α k =1 indicates that user k is active, α k =0 indicates that user k is inactive.

[0015] S12. Establish a non-orthogonal random access model: In the machine-type communication scenario under consideration, due to limited spectrum resources and the number of users being much greater than the number of pilot sequences K >> L p Therefore, the system pre-configures a non-orthogonal pilot sequence for each user, i.e. L p This indicates the length of the pilot sequence. The system's central computing unit knows the set of all pilot sequences, but not the pilot sequence for each user. Pilot transmissions from all users are synchronized.

[0016] S13. Establishing the Channel Model: This system uses a standard block fading channel. The channel model parameters described include: channel coherence time τ. c Let g be the channel gain, β be the large-scale fading coefficient, and h be the small-scale fading coefficient. The pilot length is less than the coherence time to ensure the channel gain remains constant during a single pilot transmission. Small-scale fading follows a Gaussian distribution with zero mean and unit variance. Large-scale fading is simulated using a macrocell propagation model. Then, the channel gain from the k-th user to the n-th antenna at the m-th access point can be expressed as: Where β mk h represents the large-scale fading coefficient from the k-th user to the m-th access point. mnk This represents the small-scale fading coefficient of the nth antenna from the kth user to the mth access point.

[0017] S14. Establish the input-output relationship of a cellular-free massive MIMO system: The received signal of the nth antenna at the mth access point can be expressed as...

[0018]

[0019] Where, ρ k This represents the transmit power of user pilot k. The matrix representing the set of all user pilots. Denotes the field of complex numbers, D α =diag{α}=diag{α1,α2,...,α K}, D ρ =diag{ρ}=diag{ρ1,ρ2,...,ρ K}, diag{·} denotes a diagonal matrix. It is the channel response vector from all K users to the m-th access point. This represents a white noise vector that follows a complex Gaussian distribution. Indicates dimension L p The unit matrix, σ 2 This represents the normalized background noise variance.

[0020] The received signal of the m-th access point can be represented as:

[0021]

[0022] in, It is the channel matrix of all K users at the m-th access point.

[0023] The set of received signals from all access points can be represented as

[0024]

[0025] in,

[0026] Furthermore, the described step S2 specifically includes the following sub-steps:

[0027] S21. Analyze the covariance characteristics of the signals received by all access points in the non-cellular system. From the signal model of the received signals at all access points in S14, it can be seen that each column of the received signal is independent and follows a distribution. Y(:,i) represents the i-th column of matrix Y. Indicates dimension L p The zero vector of M. Where ∑=blkdiag{∑1,∑2,...,∑ M},

[0028]

[0029] The reason why (a) is true is... This expresses the expectation. (D) λ =diag{λ}, λ=[α1ρ1, α2ρ2,...,α K ρ K ]. It is a diagonal matrix whose diagonal elements are the large-scale fading coefficients from all K users to the m-th access point.

[0030] S22. Utilizing the diagonal properties of the covariance matrices of all AP received signals and the summation property of the K rank-one elements from S21, the active user detection problem based on non-orthogonal pilots is reconstructed into a maximum likelihood estimation problem. The maximum likelihood function of Y, given λ, can be expressed as:

[0031]

[0032] Where, ∑ m As shown in S21, Tr(·) represents the trace operation.

[0033] Solving the maximum likelihood estimation problem described above can be done by maximizing the likelihood function p(Y|λ) or minimizing the negative log-likelihood function -log p(Y|λ).

[0034] Furthermore, solving the above maximum likelihood estimation problem by maximizing the likelihood function p(Y|λ) involves the following steps:

[0035] S23. Derive the equivalent log-likelihood function, which can be expressed as:

[0036]

[0037] S24. Due to The active user detection problem based on non-orthogonal pilots is then transformed into the following maximum likelihood estimation problem.

[0038]

[0039] in, This represents the empirical covariance of the received signal. This constraint must be satisfied because the user activity indicator and the user's pilot transmit power are non-negative.

[0040] Furthermore, the described step S3 specifically includes the following sub-steps:

[0041] S31. Define the cost function for the above maximum likelihood estimation problem as follows:

[0042]

[0043] in, This represents the m-th block of the cost function.

[0044] S32. Utilize the macro diversity gain of a cellless massive MIMO system to establish a cost function minimization update mechanism.

[0045] S33. For each access point, select the T users closest to that access point based on the large-scale fading coefficient, i.e.:

[0046] S34. Merge the user sets in S33, and find the corresponding service access point set for each user. k ,Right now By classifying in this way, we can obtain a set of indexes for access points that provide services to each user.

[0047] S35. Calculate set Θ k The size of the set determines the choice of solution strategy. When |Θ k When |≥1, the user is defined as a non-edge user, and the strongest primary access point in the set of service access points can be selected to serve this user. At this point, by differentiating the cost function and setting it equal to 0, the optimal step size for optimization and updating can be determined, i.e.,

[0048]

[0049] The superscript t indicates the t-th update.

[0050] When |Θ k When | = 0, the user is defined as an edge user, requiring multiple access points to work together to provide good communication performance. Therefore, selection should be based on communication quality service requirements. One access point serves user k. in This represents the index from which B maximum values ​​are returned from the set. At this point, user k is... The cost function of each access point is transformed into

[0051]

[0052] in, Differentiating it yields a polynomial function, and setting it equal to 0, the optimal step size can be obtained. in, This indicates the operation of taking the real part.

[0053] S36. To ensure the non-negativity of the estimated parameter λ, the optimal step size update mechanism is γ = max{d * , -λ k}

[0054] S37. For each user k, The update mechanism for the parameters to be estimated is λ. k ←λ k +γ.

[0055] S38. Since the cost function is non-convex, the maximum likelihood estimation problem described above is solved by iterating through S35-S37. This process continues until the cost function no longer decreases, at which point the optimal parameter estimate is obtained.

[0056] S39. To obtain activity detection information for each user, the estimated parameters in S38 are obtained. Then, it is compared with the threshold for the judgment, and finally the activity indicator for each user is obtained.

[0057]

[0058] in, This represents the decision threshold for the k-th user.

[0059] Furthermore, the described step S4 specifically includes the following sub-steps:

[0060] S41. Randomly arrange the user set Divide into G non-adjacent groups, that is

[0061] S42. For each user group Simultaneously execute the estimation process described in S34-S38. Continue until the cost function no longer decreases, obtaining the estimated parameters for all users. Then, a similar process is performed in S39 to obtain activity indicators for each user.

[0062] Furthermore, it also includes the following steps:

[0063] To avoid the optimization process from having too many active users concentrated in one location, which could negatively impact activity detection performance, a random arrangement mechanism for all users is considered. Furthermore, the transmission of all user pilot signals is strictly synchronized.

[0064] The present invention also provides an active user detection device based on covariance-based nonorthogonal pilots, including a processor and a storage medium;

[0065] The storage medium is used to store instructions;

[0066] The processor is configured to operate according to the instructions to execute the steps of the method.

[0067] Beneficial Effects: This invention establishes a mathematical model for nonorthogonal random access based on non-cellular massive MIMO. By utilizing the block diagonal characteristics of the received signal covariance matrix, the active user detection problem is reconstructed into a maximum likelihood estimation problem. The coordinate descent algorithm is used to solve the reconstructed maximum likelihood estimation problem. Based on user service quality requirements, all service users are adaptively divided into edge users and non-edge users. This fully utilizes the macro diversity gain of the non-cellular massive MIMO system to improve the accuracy of edge user activity detection and significantly reduce the overall algorithm complexity. To further reduce computation time, a parallel estimation algorithm is proposed. This invention achieves a good trade-off between performance and complexity in active user detection, and can meet the needs of massive terminal scheduling-free random access scenarios, showing great application potential. Attached Figure Description

[0068] Figure 1 A flowchart illustrating a specific embodiment of the present invention;

[0069] Figure 2 This is a schematic diagram of a non-cellular massive MIMO random access system according to a specific embodiment of the present invention;

[0070] Figure 3 This is a schematic diagram of an active user detection algorithm based on non-orthogonal pilots with covariance provided in an embodiment of the present invention;

[0071] Figure 4 This is a comparison chart showing the detection performance of the proposed detection method in a specific embodiment of the present invention with other existing solutions. Detailed Implementation

[0072] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described below are only for illustrating the present invention and do not limit the scope of the present invention.

[0073] This invention proposes an active user detection method based on covariance-based nonorthogonal pilots, the flowchart of which is shown below. Figure 1 As shown, the specific steps include the following:

[0074] S1. Establish a mathematical model for nonorthogonal random access based on non-cellular massive MIMO, specifically including the nonorthogonal random access model, channel model, and input-output relationship of non-cellular massive MIMO;

[0075] Specifically, this example establishes a mathematical model for nonorthogonal random access based on cellular-free massive MIMO according to the following steps:

[0076] S11. For example Figure 2As shown, a non-cellular massive MIMO non-orthogonal random access system is configured with M access points (APs), K single-antenna users (UEs), and a central computing unit. Each AP is equipped with N antennas, and information can be transmitted losslessly between the APs and the central computing unit via a fiber optic backhaul link. All APs and users are randomly distributed. Represents the set of all potential users. Let α represent the set of active users. Users within the system are randomly activated according to a certain activation probability ε. A Bernoulli random variable represents whether a user is active, i.e., α. k ∈{0,1} represents whether user k is active, i.e., α k =1 indicates that user k is active, α k =0 indicates that user k is inactive.

[0077] S12. Establish a non-orthogonal random access model: In the machine-type communication scenario under consideration, due to limited spectrum resources and the number of users being much greater than the number of pilot sequences K>>L p Therefore, the system pre-configures a non-orthogonal pilot sequence for each user, i.e. The system's central computing unit has a known set of all pilot sequences, but not the pilot sequence for each user. All users' pilot transmissions are synchronized.

[0078] S13. Establishing the Channel Model: This system uses a standard block fading channel. The channel model parameters described include: channel coherence time τ. c Let g be the channel gain, β be the large-scale fading coefficient, and h be the small-scale fading coefficient. The pilot length is less than the coherence time to ensure the channel gain remains constant during a single pilot transmission. Small-scale fading follows a Gaussian distribution with zero mean and unit variance. Large-scale fading is simulated using a macrocell propagation model. Then, the channel gain from the k-th user to the n-th antenna of the m-th AP can be expressed as: Where β mk h represents the large-scale fading coefficient from the k-th user to the m-th AP. mnk This represents the small-scale fading coefficient from the k-th user to the n-th antenna of the m-th AP.

[0079] S14. Establish the input-output relationship of a cellular-free massive MIMO system: The received signal of the nth antenna of the mth AP can be expressed as...

[0080]

[0081] Where, ρ k This represents the transmit power of user pilot k. D represents the set matrix of all user pilots. α=diag{α}=diag{α1, α2,...,α K}, D ρ =diag{ρ}=diag{ρ1, ρ2,...,ρ K}, deag{·} represents a diagonal matrix. It is the channel response vector from all K users to the m-th AP. σ represents the Gaussian white noise vector. 2 This represents the normalized background noise variance.

[0082] The received signal of the m-th AP can be expressed as:

[0083]

[0084] in, It is the channel matrix of all K users at the m-th AP.

[0085] The set of received signals from all APs can be represented as

[0086]

[0087] in,

[0088] S2. By utilizing the block diagonal property of the received signal covariance matrix, the active user detection problem is reconstructed into a maximum likelihood estimation problem;

[0089] Specifically, this example utilizes the block diagonal properties of the received signal covariance matrix to refactor the active user detection problem into a maximum likelihood estimation problem by following these steps:

[0090] S21. Analyze the covariance characteristics of the received signals from all access points (APs) in the non-cellular system. From the signal model of all AP received signals in S14, it can be seen that each column of the received signal is independent and follows a distribution. Y(:,i) represents the i-th column of matrix Y. Where ∑=blkdiag{∑1,∑2,...,∑ M},

[0091]

[0092] The reason why (a) is true is... This expresses the expectation. (D) λ =diag{λ}, λ=[α1ρ1,α2ρ2,...,α K ρ K ]. It is a diagonal matrix whose diagonal elements are the large-scale fading coefficients of all K users up to the m-th AP.

[0093] S22. Utilizing the diagonal properties of the covariance matrices of all AP received signals and the summation property of the K rank-one elements from S21, the active user detection problem based on non-orthogonal pilots is reconstructed into a maximum likelihood estimation problem. The maximum likelihood function of Y, given λ, can be expressed as:

[0094]

[0095] Where, ∑ m As shown in S21, the above maximum likelihood estimation problem can be solved by maximizing the likelihood function p(Y|λ) or minimizing the negative pair likelihood function -log p(Y|λ).

[0096] S23. Derive the equivalent log-likelihood function, which can be expressed as:

[0097]

[0098] S24. Due to The active user detection problem based on non-orthogonal pilots is then transformed into the following maximum likelihood estimation problem.

[0099]

[0100] in, This represents the empirical covariance of the received signal. Its constraint is that the user activity indicator symbol and the user pilot transmit power are non-negative.

[0101] S3. Based on the non-cellular macro diversity gain, the coordinate descent algorithm is used to solve the maximum likelihood estimation problem after S2 reconstruction;

[0102] Specifically, this example follows the steps below to solve the maximum likelihood estimation problem after S2 reconstruction using the coordinate descent algorithm based on the cellular-free macro diversity gain:

[0103] S31. Define the cost function for the above maximum likelihood estimation problem as follows:

[0104]

[0105] in, This represents the m-th block of the cost function.

[0106] S32. Utilize the macro diversity gain of a cellless massive MIMO system to establish a cost function minimization update mechanism.

[0107] S33. For each access point (AP), select the T users closest to that AP based on the large-scale fading coefficient, i.e.:

[0108] S34. Merge the user sets in S33, and find the corresponding service AP set for each user. k ,Right now By classifying in this way, we can obtain an index set of APs that provide services to each user.

[0109] S35. Calculate set Θ k The size of the set determines the choice of solution strategy. When |Θ k When |≥1, the user is defined as a non-marginal user, and the strongest primary AP in the set of APs being served can be selected to serve this user. At this point, by differentiating the cost function and setting it equal to 0, the optimal step size for optimization and updating can be determined, i.e.,

[0110]

[0111] The superscript t indicates the t-th update.

[0112] When |Θ k When | = 0, the user is defined as an edge user, requiring multiple APs to work together to provide good communication performance. Therefore, selection should be based on communication quality service requirements. k AP service users in This represents the index from which B maximum values ​​are returned from the set. At this point, user k is... The cost function of an AP is transformed into

[0113]

[0114] in, Differentiating it yields a polynomial function, and setting it equal to 0, the optimal step size can be obtained. in, This indicates the operation of taking the real part.

[0115] S36. To ensure the non-negativity of the estimated parameter λ, the optimal step size update mechanism is γ = max{d * , -λ k}

[0116] S37. For each user k, The update mechanism for the parameters to be estimated is λ. k ←λ k +γ.

[0117] S38. Since the cost function is non-convex, the maximum likelihood estimation problem described above is solved by iterating through S35-S37. This process continues until the cost function no longer decreases, at which point the optimal parameter estimate is obtained.

[0118] S39. To obtain activity detection information for each user, the estimated parameters in S38 are obtained. Then, it is compared with the threshold for the judgment, and finally the activity indicator for each user is obtained.

[0119]

[0120] in, This represents the decision threshold for the k-th user.

[0121] S4. To reduce computation time, a parallel solution strategy is proposed;

[0122] Specifically, this example follows the steps below to solve the maximum likelihood estimation problem after S2 reconstruction in parallel.

[0123] S41. Randomly arrange the user set Divide into G non-adjacent groups, that is

[0124] S42. For each user group Simultaneously execute the estimation process described in S34-S38. Continue until the cost function no longer decreases, obtaining the estimated parameters for all users. Then, a similar process is performed in S39 to obtain activity indicators for each user.

[0125] When applying the method of this invention, to avoid the excessive concentration of active users during the optimization process affecting the activity detection performance, all users are randomly arranged. The transmission of pilot signals for all users in the system is strictly synchronized. Furthermore, a dynamic threshold is set in this invention to achieve better active user detection performance. The active user detection method based on covariance-based non-orthogonal pilot signals obtained by this invention achieves a good trade-off between detection performance and complexity in non-cellular massive MIMO systems and is suitable for mMTC scenarios.

[0126] The technical means disclosed in this invention are not limited to those disclosed in the above embodiments, but also include technical solutions composed of any combination of the above technical features. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this invention, and these improvements and modifications are also considered within the scope of protection of this invention.

Claims

1. An active user detection method based on covariance-based nonorthogonal pilots, characterized in that, Includes the following steps: S1. Establish a mathematical model for nonorthogonal random access based on noncellular massive MIMO; S2. By utilizing the block diagonal property of the received signal covariance matrix, the active user detection problem is reconstructed into a maximum likelihood estimation problem; S3. Based on the macro diversity gain without cellular connectivity, the maximum likelihood estimation problem after reconstruction in S2 is solved using the coordinate descent algorithm to obtain the optimal parameter estimate; the activity indicator for each user is obtained based on the optimal parameter estimate. Step S1 specifically includes the following sub-steps: S11. Configure a non-cellular massive MIMO non-orthogonal random access system with M access points, K single-antenna users, and a central computing unit; each access point is equipped with N antennas, and information is transmitted losslessly between the access points and the central computing unit via a fiber optic backhaul link; all access points and users are randomly distributed. Represents the set of all potential users. This represents the set of active users; users within the system are randomly activated according to a certain activation probability ε, and the activity level of a user is represented by a Bernoulli random variable, i.e., α. k ∈{0,1} represents whether user k is active, i.e., α k =1 indicates that user k is active, α k =0 indicates that user k is inactive; S12. Establish a non-orthogonal random access model: The system pre-configures a non-orthogonal pilot sequence for each user, that is, the pilot sequence for the k-th user is... L p This indicates the length of the pilot sequence; the system's central computing unit has a known set of all pilot sequences, but does not know the pilot sequence for each user. All users' pilot transmissions are synchronized; S13. Establishing the Channel Model: The system uses a standard block fading channel, and the parameters of the described channel model include: channel coherence time τ. c Let g be the channel gain, β be the large-scale fading coefficient, and h be the small-scale fading coefficient; where the pilot length is less than the coherence time to ensure that the channel gain remains constant during a single pilot transmission; the small-scale fading follows a Gaussian distribution with a mean of 0 and a unit variance; the large-scale fading is simulated using a macrocell propagation model; then the channel gain from the k-th user to the n-th antenna of the m-th access point is expressed as: Where β mk h represents the large-scale fading coefficient from the k-th user to the m-th access point. mnk This represents the small-scale fading coefficient of the nth antenna from the kth user to the mth access point; S14. Establish the input-output relationship of a non-cellular massive MIMO nonorthogonal random access system: The received signal of the nth antenna at the mth access point is expressed as... Where, ρ k This represents the transmit power of user pilot k. The matrix representing the set of all user pilots. Denotes the field of complex numbers, D α =diag{α〉=diag{α1,α2,...,α K }, D ρ =diag{ρ}=diag{ρ1,ρ2,...,ρ K }, diag{·} denotes a diagonal matrix, It is the channel response vector from all K users to the m-th access point. This represents a white noise vector that follows a complex Gaussian distribution. Indicates dimension L p The unit matrix, σ 2 This represents the normalized background noise variance; The received signal of the m-th AP is represented as in, It is the channel matrix of all K users at the m-th access point. The set of received signals from all access points is represented as in, 2. The active user detection method based on covariance-based non-orthogonal pilots according to claim 1, characterized in that, Step S2 specifically includes the following sub-steps: S21. From the signal model of the received signals of all access points in S14, it can be seen that each column of the received signal is independent and follows a distribution. Y(:,i) represents the i-th column of matrix Y. Indicates dimension L p The zero vector of M; where ∑=blkdiag{∑1,∑2,...,∑ M }, The reason why (a) is true is... D represents the expectation; λ =diag{λ}, λ=[α1ρ1,α2ρ2,...,α K ρ K ]; It is a diagonal matrix, whose diagonal elements are the large-scale fading coefficients from all K users to the m-th access point; S22. Utilizing the diagonal property of the received signal covariance matrix of all access points in S21 and the summation property of K rank-one elements, the active user detection problem based on non-orthogonal pilots is reconstructed into a maximum likelihood estimation problem; the maximum likelihood function of Y given λ is expressed as: Where Tr(·) represents the trace operation; The maximum likelihood estimation problem is solved by maximizing the likelihood function p(Y|λ) or minimizing the negative log-likelihood function -log p(Y|λ).

3. The active user detection method based on covariance-based non-orthogonal pilots according to claim 2, characterized in that, The steps to solve the maximum likelihood estimation problem by maximizing the likelihood function p(Y|λ) include: S23. Derive the equivalent log-likelihood function of the likelihood function, expressed as: S24. Due to The active user detection problem based on non-orthogonal pilots is then transformed into the following maximum likelihood estimation problem. subject to λ≥0 K , in, Represents the empirical covariance of the received signal, 0 K This represents a zero vector of dimension K; since the user activity indicator and the user's pilot transmit power are non-negative, this constraint must be satisfied.

4. The active user detection method based on covariance-based non-orthogonal pilots according to claim 2, characterized in that, Step S3 specifically includes the following sub-steps: S31. Define the cost function for the maximum likelihood estimation problem as follows: in, This represents the m-th block of the cost function; S32. Utilize the macro diversity gain of a non-cellular massive MIMO nonorthogonal random access system to establish a cost function minimization update mechanism; S33. For each access point, select the T users closest to that access point based on the large-scale fading coefficient, i.e.: S34. Merge the user sets in S33, and find the corresponding service access point set for each user. k ,Right now The index set of access points that provide services to each user is obtained through classification; S35. Calculate set Θ k The size of the set determines the choice of different solution strategies; when |Θ k When |≥1, the corresponding user is defined as a non-edge user, and the strongest primary access point in the set of service access points can be selected to serve this user; at this time, by taking the derivative of the cost function and setting it equal to 0, the optimal step size for optimization and update can be found, that is, Wherein, the superscript t indicates the t-th update; When |Θ k When | = 0, the corresponding user is defined as an edge user. Such users require multiple access points to provide services in order to achieve good communication quality; selection is based on communication quality service requirements. Each access point serves edge user k. At this time, user k is The cost function of each access point is transformed into in, Differentiating it yields a polynomial function, and setting it equal to 0, the optimal step size is obtained. in, This indicates the operation of taking the real part; S36. To ensure the non-negativity of the estimated parameter λ, the optimal step size update mechanism is γ = max{d * , -λ k }; S37. For each user k, The update mechanism for the parameters to be estimated is λ. k ←λ k +γ; S38. Since the cost function is non-convex, the maximum likelihood estimation problem described above is solved by iterating through S35-S37. This process continues until the cost function no longer decreases, at which point the optimal parameter estimate is obtained. S39. To obtain activity detection information for each user, the estimated parameters in S38 are obtained. Then, it is compared with the threshold for the judgment, and finally the activity indicator for each user is obtained. in, This represents the decision threshold for the k-th user.

5. The active user detection method based on covariance-based non-orthogonal pilots according to claim 4, characterized in that, A parallel solution strategy is adopted to solve the reconstructed maximum likelihood estimation problem, which includes the following sub-steps: S41. Randomly arrange the user set Divide into G non-adjacent groups, that is S42. For each user group Simultaneously execute the estimation process described in S34-S38 until the cost function no longer decreases, and obtain the estimated parameters for all users. Then, a similar process is performed in S39 to obtain activity indicators for each user.

6. The active user detection method based on covariance-based non-orthogonal pilots according to any one of claims 1-5, characterized in that, It also includes the following steps: To avoid the optimization process from having too many active users concentrated in one location, which would affect the activity detection performance, a random arrangement mechanism for all users is adopted; the transmission of pilot signals for all users is strictly synchronized.

7. An active user detection device based on covariance-based non-orthogonal pilots, characterized in that, Including processor and storage media; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1 to 5.

Images