A method for detecting active users and transmission delay in asynchronous large-scale systems

Abstract

This invention belongs to the field of information and communication technology, and relates to a method for detecting active users and transmission delays in asynchronous large-scale systems. Addressing the impracticality of allocating different pilot sequences to such a large number of users in current random access protocols, and the inevitable collisions due to the limited number of preamble sequences, this invention proposes a joint active user and transmission delay detection scheme. This scheme iteratively detects active users and estimates their transmission delays, identifying conflicting users through different transmission delays. Due to the sporadic user activity patterns, the active user detection problem can be formulated as a compressed sensing problem, which is solved using an improved Turbo-CS algorithm. Furthermore, for the transmission delay estimation of active users, a delay calibration method based on expectation maximization is proposed. Simulation results verify the superior performance of the proposed algorithm in terms of missed detection probability and the number of identifiable active users.

Description

Technical Field

[0001] This invention belongs to the field of information and communication technology, and relates to a method for detecting active users and transmission delays in asynchronous large-scale systems. Background Technology

[0002] In recent years, the Internet of Things (IoT) has experienced explosive growth across various sectors, such as healthcare and transportation. It is predicted that by 2025, over 75.4 billion devices will be connected to the internet globally, and this number will continue to grow over the next decade. In large-scale random access scenarios, base stations need to serve a large number of devices, but their activity patterns are sporadic. Most devices remain dormant to conserve energy, and only a small number are activated and send data to the base station when triggered by random external events. In this situation, the base station needs to dynamically identify active users in order to allocate resources for subsequent data transmission.

[0003] However, achieving massive access in current wireless networks is not easy. The commonly used contention-based random access protocol is slotted ALOHA. When a device needs to establish a connection with a base station, it randomly selects and sends a preamble sequence from an orthogonal preamble sequence pool to the base station. Since the number of orthogonal preamble sequences in the pool is limited, in large-scale random access scenarios, the probability of multiple active devices selecting the same sequence is high, leading to access conflicts and failures. In the paper "Coded slotted ALOHA: A graph-based method for uncoordinated multiple access," IEEE Trans. Inf. Theory, vol. 61, no. 12, pp. 6815–6832, 2015, the authors proposed a coded slotted ALOHA scheme. In this scheme, each transmitted data packet is divided into several segments, each encoded with a deletion code. At the receiving end, continuous interference cancellation and deletion decoding are performed to recover information from collisions. In the literature O. Del Rio Herrero and R. De Gaudenzi, “High efficiency satellite multipleaccess scheme for machine-to-machine communications,” IEEE Trans. Aerosp. Electron. Syst., vol. 48, no. 4, pp. 2961–2989, 2012, users at the transmitting end use direct sequence spread spectrum technology to solve the problem of time slot synchronization required at the transmitting and receiving ends. More recently, in the literature Y. Polyanskiy, “A perspective on massive random-access,” in IEEE International Symposium on Information Theory (ISIT), pp. 2523–2527, 2017, a new random access mode called Passive Random Access (URA) was introduced. In URA, all users share the same codebook for transmitted messages, which greatly improves the resource utilization of the system. The task of the receiving end is to identify and recover the transmitted messages. Summary of the Invention

[0004] This invention proposes a receiver for asynchronous large-scale random access systems. The main advantage of this receiver is that, without changing the existing system resources, it designs a reliable joint detection scheme for active users at the receiving end and transmission delay to alleviate the collision conflict problem of large-scale access users and avoid the problem of excessive access time caused by repeated retransmissions due to multiple collisions. It has high practical value.

[0005] This invention proposes a novel joint detection scheme for active users and transmission delay in asynchronous large-scale random access systems, aiming to achieve joint estimation and detection of active users and transmission delay at the base station.

[0006] Consider a narrowband communication system, the system model is as follows: Figure 1 As shown, the base station is set to have N r There are N receiving antennas serving K single-antenna users. q(t) is a pulse shaping filter, m(t) is a matched filter, and M is the oversampling factor. Define N... preamble N represents the number of available leader sequences, each with a length of N. pilot During the uplink access process, each active user randomly selects from N... preamble If a preamble is selected from a set of preamble sequences, and the number of active users exceeds the number of available preambles, then the preamble signal sent by the k-th user is:

[0007]

[0008] Among them, T s Let δ(·) be the symbol duration, and x be the Dirac function. k,n This is the nth element of the preamble sequence for the kth user. (The nth element at the base station...) r The signal received by each receiving antenna is

[0009]

[0010] Where, α k The active user indicator function is α, which indicates when a user is active. k It is 1 if it is true, otherwise it is 0. For user k to the nth base station r The channel impulse response of each receiving antenna is assumed to have flat fading. k Let z(t) be the transmission delay from user k to the base station, z(t) be the convolution of q(t) and m(t), and w(t) be the ambient noise. After sampling, the results were obtained.

[0011]

[0012] in, To round down, Let z(t) be in discrete form and only consider j∈[-3M,3M], x k (t) is obtained after sampling

[0013]

[0014] Considering the sliding window technique, the signal received by the u-th sliding window is: Where t u Let y be the starting time of the u-th sliding window, and L be the number of symbols contained in the sliding window. nr,u Can be written as

[0015]

[0016] Where, α k,u For active user indicator function, τ k,u For transmission delay, x k,u (t k,u ) for t k,u The delayed leader sequence, Z is the Topulitz matrix, and its structure is as follows:

[0017]

[0018] Since z(t) is symmetric about t=0, then Z is a symmetric matrix. It is a Topelitz matrix structure consisting of m(t) such as Z. With a mean of 0 and a variance of Gaussian noise.

[0019] Consider all N r One receiving antenna, to obtain

[0020]

[0021] in, Since the joint detection method used for each sliding window is the same, the sliding window number u will be ignored below.

[0022] The present invention employs a sliding listening window-assisted method for joint detection of active users and transmission delay, comprising the following steps:

[0023] S1. At the receiving end, perform correlation peak detection on the received signal and all preamble sequences in the preamble sequence pool to preliminarily estimate the number of active users included in the sliding window. and its delay Where N Init This represents the initial number of active users.

[0024] S2, using the Turbo-CS-MMV algorithm based on Solving for G, where Turbo-CS-MMV consists of module A and module B, where module A is the LMMSE estimator and module B is the MMSE denoising unit, as shown below. Figure 2 As shown. Module A calculates the posterior mean of G. and variance for

[0025]

[0026]

[0027] in, for The j-th column, for The (j,j)th element.

[0028] S3. Solve for the extrinsic information about G, including its mean. and variance for

[0029]

[0030]

[0031] in, for The j-th column, for The (j,j)th element.

[0032] S4, Order

[0033] S5. Module B calculates the posterior mean of G. and variance for

[0034]

[0035]

[0036] Where ⊙ represents the Hadamarda accumulation. for The j-th column, for The (j,j)th element, p B,i Calculated as

[0037]

[0038] S6. Solve for the extrinsic information about G, and its mean. and variance for

[0039]

[0040]

[0041] in, for The j-th column, for The (j,j)th element.

[0042] S7, Order

[0043] S8. If the maximum number of iterations is exceeded, the process ends and outputs the result. and Otherwise, switch to S2.

[0044] S9. Update the sparsity ρ using the expectation-maximization algorithm, and calculate it as follows:

[0045]

[0046] S10. If the maximum number of iterations is exceeded, the process ends; otherwise, proceed to S2.

[0047] S11. Update the delay using the expectation-maximization algorithm. Update sequentially starting from the first element and continuing until the last element, where the i-th element... exist Search within the range, with a step size of Make

[0048]

[0049] in, for The j-th column, It is a diagonal matrix with diagonal elements V G The j-th column, ε is a small positive real number to reduce search complexity.

[0050] S12. If the maximum number of iterations is exceeded, the process ends; otherwise, proceed to S2.

[0051] The beneficial effects of this invention are as follows: Addressing the impracticality of allocating different pilot sequences to such a large number of users in current random access protocols, and the inevitable collisions due to the limited number of preamble sequences, this invention proposes a joint active user and transmission delay detection scheme. This scheme iteratively detects active users and estimates their transmission delays, identifying conflicting users through different transmission delays. Due to the sporadic user activity patterns, the active user detection problem can be formulated as a compressed sensing problem, which is solved using an improved Turbo-CS algorithm. Furthermore, for the transmission delay estimation of active users, a delay calibration method based on expectation maximization is proposed. Simulation results verify the superior performance of the proposed algorithm in terms of missed detection probability and the number of identifiable active users. Attached Figure Description

[0052] Figure 1 System model.

[0053] Figure 2 Turbo-CS-MMV algorithm model.

[0054] Figure 3 : Simulation results of missed detection probability.

[0055] Figure 4 Simulation results of normalized mean square error (NMSE) for time delay estimation.

[0056] Figure 5 Simulation results of the number of users successfully connected for the first time.

[0057] Figure 6 Simulation results of the expectation-maximization algorithm's delay update. Detailed Implementation

[0058] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples.

[0059] Example

[0060] In this specific embodiment, the path coefficient β k Calculated as

[0061] β k = -128.1 - 36.7log 10 (d k )

[0062] Considering the ZC sequence as the leading sequence, the nth element in the sequence is x. k,n =exp(-jπvn(n+1) / N pilot ), n = 0, 1, ..., N pilot -1, the specific parameter settings for the method are shown in the table below:

[0063] Table 1 Main Simulation Parameters

[0064]

[0065]

[0066] Based on the above parameter settings, the specific steps of this simulation are as follows:

[0067] S1. At the receiving end, perform correlation peak detection on the received signal and all pilot sequences in the pilot sequence pool to preliminarily estimate the number of active users included in the sliding window. and its delay Where N Init This represents the initial number of active users.

[0068] S2. Solve for the posterior mean of G in module A. and variance for

[0069]

[0070]

[0071] in, for The j-th column, for The (j,j)th element.

[0072] S3. Solve for the extrinsic information about G, including its mean. and variance for

[0073]

[0074]

[0075] in, for The j-th column, for The (j,j)th element.

[0076] S4, Order

[0077] S5. Calculate the posterior mean of G in module B. and variance for

[0078]

[0079]

[0080] Where ⊙ represents the Hadamarda accumulation. for The j-th column, for The (j,j)th element, p B,i Calculated as

[0081]

[0082] S6. Solve for the extrinsic information about G, and its mean. and variance for

[0083]

[0084]

[0085] in, for The j-th column, for The (j,j)th element.

[0086] S7, Order

[0087] S8. If the maximum number of iterations is exceeded, the process ends and outputs the result. and Otherwise, switch to S2.

[0088] S9. Update the sparsity ρ using the expectation-maximization algorithm, and calculate it as follows:

[0089]

[0090] S10. If the maximum number of iterations is exceeded, the process ends; otherwise, proceed to S2.

[0091] S11. Update the delay using the expectation-maximization algorithm. Update sequentially starting from the first element and continuing until the last element, where the i-th element... exist Search within the range, with a step size of Make

[0092]

[0093] in, for The j-th column, It is a diagonal matrix with diagonal elements V G The j-th column, ε is a small positive real number to reduce search complexity.

[0094] S12. If the maximum number of iterations is exceeded, the process ends; otherwise, proceed to S2.

[0095] Figure 3 , Figure 4 and Figure 5 The performance comparison between the proposed algorithm and traditional random access algorithms is presented. In traditional random access algorithms, correlation peak detection is performed only on the received signal and the local preamble to identify active users and determine the transmission delay. It can be seen that oversampling can effectively improve system performance for both algorithms. Compared with traditional random access algorithms, the proposed algorithm utilizes the estimated time delay as prior information for Turbo-CS, resulting in a significant performance improvement. For example, when SNR = 15dB and M = 1 (or M = 2), Figure 5 The number of identifiable active users increased by 61% (or 46%) compared to traditional algorithms. Furthermore, Figure 6 The relationship between the false negative probability, the normalized mean square error of the time delay τ, and the number of identified active users and the number of iterations of the Expectation-Maximization (EM) algorithm is shown. The decreasing curve indicates that the estimation accuracy of the time delay gradually improves with increasing iterations, validating the effectiveness of the proposed EM-based time delay calibration method. Furthermore, the improved time delay estimation accuracy further contributes to continuously reducing the false negative probability and increasing the number of identifiable active users. The proposed algorithm converges after 2 or 3 EM learning iterations.

Claims

1. A method for detecting active users and transmission delay for asynchronous large-scale system, the system comprising a base station with N r receive antennas, K single-antenna users, a pulse-shaping filter q(t), a matched filter m(t), an oversampling module with an oversampling factor M, the uplink access procedure of a user being to select a preamble from N preamble preambles, send the preamble through a transmit antenna after passing through the pulse-shaping filter q(t), and obtain a received signal after passing through the matched filter m(t) at the base station end after receiving the signal through the antenna, and then send the received signal into the oversampling module to obtain a sampled signal; defining the length of each preamble sequence as N pilot , and the number of active users being greater than the number of available preambles, the preamble signal sent by the kth user being: wherein T s is the symbol duration, δ(·) is the Dirac function, x k,n is the nth element of the preamble sequence of the kth user; the signal received by the nth r receive antenna at the base station end is: where α k is an active user indicator function, which is 1 when the user is active and 0 otherwise, k is the channel impulse response from user k to the n r th receive antenna of the base station and assuming flat fading, τ k is the transmission delay from user k to the base station, z(t) is the convolution of q(t) and m(t), and w(t) is the ambient noise; After sampling, we get:​ wherein is rounded down, is the discrete form of z(t) and only j ∈ [-3M, 3M] is considered; The detection method is characterized in that the detection method comprises the following steps: S1, using a sliding detection window technology, defining sampling samples in a u-th sliding window as: where t u is the starting time of the u-th sliding window, L is the number of symbols contained in the sliding window, and y nr,u is written as: wherein a k,u is an active user indicator function within the window, τ k,u is a transmission delay, x k,u (τ k,u ) is a preamble sequence delayed by τ k,u , and Z is a symmetric Toeplitz matrix, wherein is a Toeplitz matrix structure as Z, is a Gaussian noise with mean 0 and variance is a Gaussian noise with mean 0 and variance Consider all N r receive antennas, we get Y u = ZX u (τ u )G u + FW u wherein, , The goal is to estimate τ u and G u from Y u where G u is a row-sparse matrix with non-zero rows being active users; Since the joint detection method is consistent for each sliding window, the sliding window index u is ignored; at the receiving end, the received signal is correlated with all the preamble sequences in the preamble sequence pool to detect the correlation peaks, and a threshold is set to preliminarily estimate the potential active users contained in the sliding window and the time delay where N Init is the number of potential active users; S2, solving G based on the Turbo-CS-MMV algorithm solving G, including module A and module B, wherein module A is an LMMSE estimator and module B is an MMSE denoiser; solving the prior mean and variance of G based on module A and variance of G based on module B wherein is the jth column of is the (j,j)th element of S3, solve for the extrinsic information about G, whose mean and variance are: wherein is the jth column of is the (j,j)th element of S4, let S5. Compute posterior mean about G using module B and variance are: where is the Hadamard product, is the jth column of is the (j, j)th element of B,i is calculated as: S6, solve the extrinsic information about G, whose mean and variance are: wherein is the jth column of is the (j,j)th element of S7, let S8, if the maximum iteration round is exceeded, ending, otherwise, turning to S2; S9, updating the sparsity p by using an expectation maximization algorithm: S10, if the maximum number of iterations is exceeded, then end and output the posterior mean for G and variance else go to S2; S11, updating the time delay by using the expectation maximization algorithm updating in order from the first element to the last, wherein the i-th element In searching in the range of such that: wherein, is the jth column of is the jth column of is a diagonal matrix with diagonal elements V G is the jth column of, and ε is a small positive real number to reduce the search complexity. S12, if the maximum number of iteration is exceeded, then end and output the estimated and transmission delay where The non-zero rows in are the detected active users, otherwise go to S2.

Images