A method for grant-free random access in massive multiple-input multiple-output systems

Abstract

The application provides a kind of method for large-scale multiple-input multiple-output system license-free random access, comprising: establishing wireless communication system model;The receiving signal of base station is obtained, the part of larger target receiving power is selected, and the number of user equipment is estimated using minimum description length criterion;The preamble sequence and time offset of each user equipment sending signal are estimated using preamble detection algorithm;The channel estimation of each user equipment is carried out by base station using PI estimation value and TO estimation value, and the channel equalization of receiving signal is carried out;The estimation value of uplink service data of each user equipment is decoded by base station, and license-free random access is realized;The preamble detection is carried out based on maximum likelihood criterion in the application, and detection performance is excellent without setting detection threshold;Secondly, the complexity of ML preamble detection algorithm is reduced based on single-source approximation method in the application;In addition, the precision of TO estimation is improved by further using binary search auxiliary detection in the application.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wireless communication, and more particularly to a grant-free random access method for a large-scale multiple-input multiple-output system. BACKGROUND

[0002] With the rapid development of wireless networks, how to provide reliable connections for a large number of user equipment (UE) has become one of the key problems of future communication systems. Large-scale multiple-input multiple-output (MIMO) technology deploys a large number of antennas at the base station (BS), which is considered to be a promising solution to meet the demand for massive access. Since random access (RA) is a necessary process for UE to establish a communication link with BS, in a large-scale MIMO system, it is a key technical problem to quickly process the random access request of UE. In recent years, the research on RA has gradually attracted attention, and researchers have proposed many RA schemes for large-scale MIMO scenarios.

[0003] The long-term evolution (LTE) and new radio (NR) standards adopt a grant-based RA procedure, which is a four-step handshake process. In the grant-based RA process, the access UE first selects a preamble index (PI) and sends the corresponding preamble sequence to the base station. Only if the UE can receive the RA response sent by the BS, the UE can transmit uplink data. Therefore, when the data amount of the UE is small, especially when a large amount of small-scale data needs to be repeatedly sent, the grant-based RA procedure will generate more signaling overhead, resulting in a significant reduction in system efficiency.

[0004] One of the methods to reduce signaling overhead is to use a grant-free RA procedure. In such a procedure, the UE sends uplink data at the same time as sending the preamble sequence, and the BS can detect the data from the UE with the help of a large antenna array. Since the request-grant procedure is omitted, the signaling overhead of grant-free RA is significantly reduced compared with grant-based RA.

[0005] In the unlicensed RA procedure, Preamble Detection (PD) is a necessary step for the BS to identify the preamble sequence transmitted by the accessing UE, and is also the basis for the subsequent channel estimation and data detection operations. Most of the existing PD algorithms designed for unlicensed RA assume that the UE is synchronized. In an Orthogonal Frequency-Division Multiplexing (OFDM) system, the synchronization of the UE can be achieved by transmitting the preamble sequence on a single subcarrier, multiple consecutive symbols. In this case, the impact of Timing Offset (TO) can be ignored because the phase offset caused by TO is a constant for the same subcarrier and can be incorporated into the channel estimation module for transparent processing. However, if the preamble sequence is transmitted on multiple consecutive subcarriers, the TO of the UE cannot be ignored because it will cause a linearly varying phase shift in the frequency domain.

[0006] If the preamble sequence is transmitted on multiple consecutive subcarriers, even if different UEs choose the same preamble sequence, the BS still has the opportunity to distinguish these UEs by the difference of TO, which makes it possible to improve the RA performance by exploiting the TO diversity of UEs. Currently, some researchers have considered the different TO of UEs and proposed some PD algorithms for the design of grant-based RA. For example, the Frequency Domain Correlation (FDC) algorithm proposed in the literature “B. Liang, Z. He, K. Niu, B. Tian, and S. Sun, “The research on random access signal detection algorithm in LTE systems,” in Proc. IEEE Int. Symp. MAPE Wireless Commun., Chengdu, China, Oct. 2013, pp. 115-118.” determines PI and TO according to the strength and position of the correlation peak, respectively; further, this method can also be extended to an enhanced version for frequency-selective channels or high-mobility scenarios. In addition, the FDC algorithm ignores the preamble sequence collision between UEs, while the User Grouping (UG) based algorithm proposed in the literature “S. Mukherjee, A. K. Sinha, and S. K. Mohammed, “Timing advance estimation and beamforming of random access response in crowded TDD massive MIMO systems,” IEEE Trans. Commun., vol. 67, no. 6, pp. 4004-4019, Jun. 2019.” considers such collision, and through the grouping operation, if the difference of TO between multiple UEs that choose the same preamble sequence is large enough, these UEs can be detected respectively. Therefore, compared with the FDC algorithm, the UG algorithm can adapt to more scenarios of UE access.

[0007] However, the above algorithms considering TO (FDC algorithm and UG algorithm) need to set a detection threshold, and the value of the threshold is related to the system parameters such as noise power, preamble sequence length, target false alarm probability, etc. Generally, when the algorithm is applied to different scenes of various parameter combinations of the system, it is difficult to determine the optimal value of the detection threshold. In addition, the correlation operation of such algorithms is generally realized by Fast Fourier Transform (FFT), so the estimation accuracy of TO is limited by the length of FFT. The disadvantage of limited TO estimation accuracy will affect the channel estimation and data detection performance in grant-free RA, thereby reducing the UE access success rate. Therefore, a new type of PD scheme needs to be designed for grant-free RA to improve the access success rate. SUMMARY

[0008] To overcome the defects of the above prior art preamble detection algorithm for grant-free random access, which needs to set a detection threshold and has limited estimation accuracy, the present application provides a grant-free random access method for a large-scale multiple-input multiple-output system, which performs preamble detection based on the Maximum Likelihood (ML) criterion, without setting a detection threshold and with excellent detection performance. Secondly, the present application is based on the single-source approximation method, which further reduces the complexity of the ML preamble detection algorithm. In addition, the present application further utilizes binary search assisted detection to improve the accuracy of TO estimation.

[0009] To solve the above technical problems, the technical solutions of the present application are as follows:

[0010] A grant-free random access method for a large-scale multiple-input multiple-output system, comprising the following steps:

[0011] S1: Establish a wireless communication system model based on orthogonal frequency division multiplexing, wherein a plurality of user equipment respectively transmit signals to a base station through different types of PRACH resource blocks which are orthogonal in the frequency domain;

[0012] The transmitted signal includes a preamble sequence and encoded uplink data;

[0013] S2: Obtain the received signal of the base station, select the part with larger target received power in the received signal, and estimate the number of user equipment using the minimum description length criterion;

[0014] S3: Estimate the preamble sequence and time offset of the signal transmitted by each user equipment using a preset preamble detection algorithm, and obtain the PI estimate value and TO estimate value of each user equipment;

[0015] The preset preamble detection algorithm is specifically any one of ML preamble detection algorithm, AML preamble detection algorithm and DS-AML preamble detection algorithm.

[0016] S4: The base station estimates the channel for each user equipment using the PI estimate and the TO estimate, and performs channel equalization on the received signal to obtain an estimate of the uplink service data for the corresponding user equipment;

[0017] S5: The base station decodes the estimate of the uplink service data for each user equipment to complete the reception of the uplink service data and realize grant-free random access.

[0018] Preferably, in the step S1, the wireless communication system uses OFDM transmission signals, wherein part of the time-frequency resources are allocated as a physical random access channel; the transmission of the user equipment's transmission signal in the physical random access channel is divided into a preamble transmission phase and a data transmission phase, and the user equipment's transmission signal consists of two consecutive OFDM symbols in time domain;

[0019] A plurality of PRACH resource blocks are provided within the bandwidth of the wireless communication system; in each PRACH resource block, the bandwidth allocated in the preamble transmission phase and the data transmission phase is LΔf and wherein L is the number of symbols contained in the preamble sequence, is the number of data symbols, and Δf is the subcarrier spacing; the type of the PRACH resource block includes any one of the following types: Type 1: Type 2: and Type 3:

[0020] Preferably, in the wireless communication system model of the step S1, K user equipments access 1 base station, and K is unknown; each user equipment transmits signals through a single transmit antenna, and the base station receives signals using M receive antennas; the serial number of the user equipment is denoted as k, satisfying k = 1, …, K; the distance between the user equipment k and the base station is d k , and the maximum distance d max between the user equipment and the base station is known; the TO of the user equipment k is c = 3 × 10 8 m / s represents the speed of light, and the value range of the TO is [0, τ max ],

[0021] In the preamble transmission phase, each user equipment selects a preamble sequence with a length of L from a set of predefined preamble sequences {x q | q = 1, …, Q} and transmits it, wherein Q is the number of preamble sequences, x q ∈ C L×1 denotes the preamble sequence with serial number q, and C denotes the complex number set; the preamble sequence selected by the user equipment k is denoted as q k , and the transmitted preamble sequence is xqk ;

[0022] In the preamble transmission phase, the received signal of the mth antenna of the base station is denoted as y m , m = 1, …, M, and is specifically:

[0023]

[0024] wherein, denotes Hadamard product, γ k > 0 is the transmit power of the user equipment k at each subcarrier in the preamble transmission phase, h k,m is the channel coefficient between the user equipment k and the base station antenna m within the PRACH resource block, β k and v k,m represent large-scale and small-scale fading coefficients, respectively; a(τ k ) ∈ C L×1 denotes the phase offset caused by τ k in the preamble transmission phase; w m ∈ C L×1 is the Gaussian white noise vector in the preamble transmission phase;

[0025] In the data transmission phase, the user equipment k generates a data symbol vector through constellation mapping on the encoded uplink service data and transmits it, and the received signal of the base station in the data transmission phase is denoted as and is specifically:

[0026]

[0027] wherein, is the transmit power of the user equipment k at each subcarrier in the data transmission phase; denotes the phase offset caused by τ k in the data transmission phase; is the Gaussian white noise vector in the data transmission phase;

[0028] Convert y m and into matrix form, respectively, to obtain the wireless communication system model:

[0029]

[0030] wherein,

[0031]

[0032] wherein,

[0033] Preferably, in the step S2, the estimating the number of user equipments by using the minimum description length criterion comprises the following steps:

[0034] Selecting a part with larger target received power in the received signal of the base station, and calculating a sample covariance matrix R:

[0035]

[0036] Wherein, ρ and are the target received powers of the signals sent by the set user equipment in the preamble transmission stage and the data transmission stage to the base station;

[0037] Calculating eigenvalues {λ m | m = 1, …, M} of R, arranging them in the order of λ1≥…≥λ M , and the estimated value of the number of user equipments is calculated by the following formula:

[0038]

[0039] Wherein, μ is an intermediate variable, μ = 0, …, M-1; D(μ) is an intermediate function.

[0040] Preferably, in the step S3, the objective function of the ML preamble detection algorithm is represented as:

[0041]

[0042] Wherein, is the projection matrix of the matrix ; represents the square of the F-norm of the matrix; and are the trial values of PI and TO respectively; and are the PI estimate and TO estimate output by the ML preamble detection algorithm respectively;

[0043] Densely sampling on the continuous interval [0, τ max ], and obtaining a discrete set Taking the discrete set as the new value range of TO; by exhaustively searching all possible combinations of and , the objective function of the ML preamble detection algorithm is solved to obtain the PI estimate and the TO estimate

[0044] Preferably, in the step S3, the ML preamble detection algorithm is improved based on the single-source approximation method to obtain an AML preamble detection algorithm;

[0045] The objective function of the AML leader detection algorithm is expressed as:

[0046]

[0047] in, is the intermediate matrix, y m The conjugate vector of ; is the experimental value of normalized TO, which is recorded as ε=τNΔf, where N≥L is the fast Fourier transform length, and ε max is the maximum value of ε; For is the phase shift of the variable, Indicates rounding up;

[0048] According to the discrete set Determine the experimental value of normalized TO The value range of is solved by fast Fourier transform All values ​​of , output local maxima, and the PI corresponding to each local maximum is used as the PI estimate Using the normalized TO corresponding to each local maximum Further calculation of TO estimate

[0049] Preferably, in step S3, the accuracy of TO estimation in the AML leader detection algorithm is further improved based on a binary search method to obtain a DS-AML leader detection algorithm;

[0050] The PI estimate output by the DS-AML leader detection algorithm

[0051] In the DS-AML preamble detection algorithm, based on the estimated number of user devices and user equipment κ selected The base station performs a binary search operation on each user equipment for the normalized TO, and the number of iterations of the binary search is I; wherein, for The κth element in ;

[0052] The estimated value of the normalized TO of the user equipment κ at the i-th iteration is recorded as Its initial value is given by the AML leader detection algorithm, that is, for The κth element in ; in the binary search process, according to The objective function of the AML leader detection algorithm is searched in the order of exist The peak value near the edge of the y-axis is continuously updated during the iteration process. - ,ψ +}, used to determine Update direction; {ψ0,ψ - ,ψ + The initial value of} is set to:

[0053]

[0054] During the binary search process, the search step size of each iteration becomes the size of the previous iteration. According to the i-th iteration ψ - and ψ + The size relationship, The update method is:

[0055]

[0056] Completed in one iteration After the update of {ψ - ,ψ +}, to determine the update direction of normalized TO at the next iteration: If ψ + >ψ - , then ψ - ←ψ0,ψ + remains unchanged; otherwise ψ + ←ψ0,ψ - remains unchanged; at the same time, ψ0 is updated to:

[0057]

[0058] After I iterations, the estimated value of the normalized TO of the user equipment κ is The estimated value of TO is

[0059] The normalized TO estimate output by the DS-AML leader detection algorithm is Further calculation of TO estimate

[0060] Preferably, in step S4, an equivalent channel matrix G of each user equipment in the preamble transmission phase and an equivalent channel matrix G of each user equipment in the data transmission phase are constructed.

[0061]

[0062] and satisfy

[0063] The base station uses the PI estimation value and TO estimates An estimate of the equivalent channel matrix G of each user equipment in the preamble transmission phase is made, in particular:

[0064]

[0065] wherein, is an estimate of G; (·) + denotes the pseudo-inverse operation;

[0066] An estimate of the equivalent channel matrix G of each user equipment in the data transmission phase is further calculated An estimate of the equivalent channel matrix G of each user equipment in the data transmission phase is further calculated The channel estimation is completed.

[0067] Preferably, the transmission power of the user equipment k is calculated as:

[0068]

[0069] wherein P max is the maximum total transmission power.

[0070] Preferably, in the step S4, the received signal is channel equalized to obtain an estimate of the uplink service data of the corresponding user equipment is expressed as:

[0071]

[0072] wherein, denotes the zero-forcing equalizer, and * denotes the matrix conjugate.

[0073] Compared with the prior art, the technical scheme of the present application has the beneficial effects that:

[0074] The present application provides a kind of for large-scale multiple-input multiple-output system's license-free random access method, first establishes the wireless communication system model based on orthogonal frequency division multiplexing, then obtains the received signal of base station, selects the part of target received power in received signal greater, uses minimum description length criterion to estimate the number of user equipment;Using preset preamble detection algorithm estimates the preamble sequence and time offset of each user equipment signal, obtains the PI estimate value and TO estimate value of each user equipment;After that, base station uses PI estimate value and TO estimate value to each user equipment channel estimation is carried out, and the received signal is channel equalized, obtains the estimate of the uplink service data of the corresponding user equipment;Finally, base station will each user equipment uplink service data estimate is decoded, completes the reception of uplink service data, realizes license-free random access;

[0075] Unlike existing threshold-based algorithms, the ML preamble detection algorithm designed based on the ML criterion of the application does not need to set a detection threshold, and can jointly detect the PI and TO of multiple UEs, wherein the number of UEs is calculated by the Minimum Description Length (MDL) criterion; the algorithm has excellent detection performance, and its complexity increases rapidly with the increase of the number of UEs, and is suitable for application scenarios that need to obtain absolute optimal performance under the condition of allowing computing power, and can also serve as a theoretical upper limit reference for the PD scheme in the unlicensed RA problem;

[0076] Secondly, based on the similarity between the Direction-of-Arrival (DOA) estimation problem and the PD problem, the application borrows the approximate ML estimation method of DOA to reduce the complexity of the ML preamble detection algorithm; specifically, the application extends the single-source approximate method for estimating continuous DOA to the PD problem with both discrete and continuous variables, thereby obtaining a detection algorithm based on the Approximated ML (AML) criterion, which can significantly reduce the computational complexity of detection while maintaining relatively optimal performance;

[0077] In addition, to solve the problem that the TO estimation accuracy is limited by the FFT length, the application further improves the aforementioned AML preamble detection algorithm by using dichotomous search, which is called Dichotomous Search aided AML (DS-AML) detection algorithm; when the FFT length is small, the DS-AML algorithm can provide greater performance gain compared to the AML preamble detection algorithm, while still maintaining similar computational complexity. BRIEF DESCRIPTION OF DRAWINGS

[0078] Figure 1 A flowchart of an unlicensed random access method for a large-scale multiple-input multiple-output system provided by the application.

[0079] Figure 2 A PRACH resource block structure diagram provided by the application.

[0080] Figure 3 An unlicensed RA flowchart provided by the application.

[0081] Figure 4 A convergence performance diagram of the DS-AML preamble detection algorithm provided by Example 4.

[0082] Figure 5 A RA success rate performance diagram of different detection algorithms provided by Example 4.

[0083] Figure 6Fig. 4 shows the RA success rate performance of different detection algorithms provided in Example 4 under Q=K=2. DETAILED DESCRIPTION

[0084] The accompanying drawings are only intended to illustrate the present application, and should not be construed as limiting the present application;

[0085] In order to better illustrate the present application, some components in the drawings may be omitted, enlarged or reduced, and do not represent the actual size of the product;

[0086] It is understandable for those skilled in the art that some well-known structures and their descriptions in the drawings may be omitted.

[0087] The technical solutions of the present application will be further described below in combination with the drawings and examples.

[0088] As shown in Figure 1 The present application provides a grant-free random access method for a large-scale multiple-input multiple-output system, comprising the following steps:

[0089] S1: Establish a wireless communication system model based on orthogonal frequency division multiplexing, wherein a plurality of user equipment respectively transmit signals to a base station through different types of PRACH resource blocks which are orthogonal in the frequency domain;

[0090] The transmitted signals include a preamble sequence and encoded uplink service data;

[0091] S2: Obtain the received signals of the base station, select the part with larger target received power in the received signals, and estimate the number of user equipment using the minimum description length criterion;

[0092] S3: Estimate the preamble sequence and time offset of the signal transmitted by each user equipment using a preset preamble detection algorithm, and obtain the PI estimate value and TO estimate value of each user equipment;

[0093] The preset preamble detection algorithm is specifically any one of ML preamble detection algorithm, AML preamble detection algorithm and DS-AML preamble detection algorithm;

[0094] S4: The base station performs channel estimation on each user equipment using the PI estimate value and TO estimate value, and performs channel equalization on the received signals to obtain the estimate value of the uplink service data of the corresponding user equipment;

[0095] S5: The base station decodes the estimate value of the uplink service data of each user equipment to complete the reception of the uplink service data, and realizes grant-free random access.

[0096] As shown in Figure 2As shown, the wireless communication system considered in the present invention adopts OFDM transmission, wherein part of the time-frequency resources are allocated as a physical random access channel (PRACH); Figure 2 As shown in FIG, the signal transmission of the UE in the PRACH is divided into two phases: the Preamble Transmission Phase (PTP) and the Data Transmission Phase (DTP). The transmission signal constitutes two continuous OFDM symbols in the time domain.

[0097] In addition to the regular service data blocks, multiple PRACH resource blocks can be set within the system bandwidth. In each PRACH resource block, the bandwidth allocated to PTP and DTP is LΔf and LΔf respectively. Where L is the number of symbols contained in the leading sequence, is the number of data symbols, Δf is the subcarrier spacing; Figure 2 In the PRACH structure scheme shown, the PRACH resource blocks can be of different types. For each PRACH resource block, L and There can be different value combinations, namely type 1: Type 2: and type 3: And they are all applicable to the authorization-free RA method proposed in the present invention;

[0098] According to different application scenarios, using different types of frequency domain orthogonal PRACH resource blocks in the same frame can have higher flexibility; for example, it can support UEs with different uplink data sizes to select PRACH resource blocks of appropriate sizes for access, thereby avoiding the waste of spectrum resources caused by uniformly using larger PRACH resource blocks to be compatible with UEs with different uplink data sizes; it is pointed out here that the unlicensed RA method proposed in the present invention is applicable to a single PRACH resource block of any type. Figure 2 In the case where there are multiple PRACH resource blocks within the system bandwidth shown, the method can be performed independently on each of the PRACH resource blocks;

[0099] Assume that there are K UEs evenly distributed in the cell accessing the BS, whose serial numbers are k=1,…,K; the distance between UE k and the BS is d k , whose upper bound is d max Since each UE is located at a different position in the cell, the signals sent by each UE simultaneously will reach the BS with different TOs. Under the premise of completing downlink synchronization, the TO of the UE is equal to the round trip time (RTD) between the UE and the BS. Let the TO of UE k be where c = 3 x 10 8 m / s represents the speed of light; the maximum TO of a UE is where d max is the maximum distance between a UE and a BS; for simplicity of description, the range of TO of a UE is denoted as T = [0, τ max ]; in addition, it is assumed that the maximum distance d max of a UE is known by the BS.

[0100] A. The grant-free RA procedure:

[0101] As shown in FIG. 1, Figure 3 a typical grant-free RA operation procedure is shown; Figure 3

[0102] In the PTP procedure, each UE selects a sequence of length L from a pre-defined preamble sequence set {x q | q = 1, …, Q} and transmits, where Q is the number of preamble sequences, x q ∈ C L×1 denotes the preamble sequence with index q, and C denotes the complex set; the set of preamble indices is denoted as Q = {q = 1, …, Q}; each preamble sequence satisfies the following constant modulus constraint:

[0103] |[x q ] l | 2 = 1, l = 1, …, L (1)

[0104] Let q k denote the PI selected by UE k, then the preamble sequence transmitted by UE k is x It is assumed that each UE transmits the preamble sequence through a single transmit antenna, and the BS uses M receive antennas to receive the preamble sequence, then the received signal of the mth (m = 1, …, M) antenna of the BS in the PTP can be expressed as:

[0105]

[0106] where: denotes the Hadamard product, γ k > 0 is the transmit power of UE k on each subcarrier, h k,m ∈ C L×1 is a vector composed of channel coefficients between UE k and the mth antenna of the BS on L subcarriers, and a(τ) ∈ C L×1 denotes the phase offset caused by τ ∈ T, and its expression is:

[0107] a(τ) = [1, e -j2πΔfτ ,…, e -j2π(L-1)Δfτ ] T (3)​

[0108] w m ∈C L×1 is a Gaussian white noise vector; the noise power of each subcarrier is v = N0Af, where N0denotes the noise power spectral density;

[0109] In the DTP procedure, UE k generates data symbol vector from channel-coded uplink data through constellation mapping and transmits, with the data symbols satisfying the power normalization condition:

[0110]

[0111] where is the transmit power of UE k per subcarrier in DTP, are the channel vector, phase offset vector, and Gaussian white noise vector of DTP, respectively; similar to equation (3), the expression of the phase offset vector of DTP is:

[0112]

[0113] The vectors composed of the Pis and TOs of all UEs are denoted as q = [q1, …, q K ] T ∈Q K×1 and r = [r1, …, r K ] T ∈T K×1 respectively; based on the received signals {y m |m = 1, …, M} of PTP, the BS obtains the PI and TO estimates of UEs after performing PD operation, which are denoted as and where is the estimate of the number of UEs, and are the kth elements of and respectively; it is noted that, since the BS does not know the actual number order of UEs, the estimates of (q k , r k ) are usually not, so a different subscript k is used to avoid misleading;

[0114] After obtaining the PI estimate and the TO estimate , the channel estimation module of the BS estimates the channel coefficients using and passes them to the channel equalization module to obtain An estimated value of the uplink service data symbol is obtained; finally, the BS obtains the uplink service data information of the UE through channel decoding;

[0115] Details of channel estimation and channel equalization will be introduced below.

[0116] B. Channel model:

[0117] The present application assumes that the channel coefficients remain unchanged within a PRACH resource block, which is reasonable in some typical communication scenarios; for example, the coherence bandwidth of the urban macro cell TDL-B channel defined by 3GPP is about 2.25 MHz; the sea surface channel can be considered as a non-frequency selective fading channel within a bandwidth of 20 MHz; in the LTE and NR standards, the typical bandwidth of PRACH is 1.08 MHz, and the subcarrier spacing is 1.25 kHz, under this configuration, it can be considered that in the urban macro cell and sea surface scenarios, the channel of the PRACH resource block is unchanged in the frequency domain; at the same time, in the typical NR frequency band of 3.4 GHz, in the urban scenario, the typical moving speed of pedestrians is 1.5 m / s, so the coherence time of the channel is In the sea surface scenario, the typical speed of the ship is 20 knots (about 10 m / s), so the coherence time of the channel is Since the typical cyclic prefix (CP) duration of the PRACH configuration set in the LTE and NR standards is 103us, the duration of two consecutive OFDM symbols containing CP is which is much smaller than the channel coherence time in the above two typical application scenarios; therefore, the present application assumes that the channel coefficients remain approximately unchanged within the PRACH resource block duration;

[0118] Under the assumption that the channel remains unchanged, the signal models shown in equations (2) and (4) can be rewritten as:

[0119]

[0120] where h k,m is a scalar, representing the channel coefficient between UE k and BS antenna m within the PRACH resource block, and its value is:

[0121]

[0122] where β k and v k,m represent large-scale and small-scale fading coefficients, respectively; without loss of generality, it is assumed that β k is known at the UE k end and unknown at the BS end; in the actual system, UE k can estimate β k by measuring the downlink reference signal power.

[0123] The PTP signal model shown in equation (6) can be further expressed in matrix form as:

[0124]

[0125] wherein:

[0126]

[0127] Similarly, the DTP signal model shown in equation (7) is expressed as:

[0128]

[0129] wherein:

[0130]

[0131] Based on the channel coefficient model of equation (8) and the received signal models of equations (9) and (11), the following further introduces the design of power control, channel estimation, channel equalization and other modules of the RA scheme;

[0132] 1) Power control;

[0133] A power control method is adopted which allows each UE to have similar RA performance; specifically, the target received power of each UE at the BS for the signals sent by PTP and DTP is set to ρ and respectively, where ρ and The size of ρ and is determined according to the required signal-to-noise ratio (SNR) at the BS and is broadcast by the BS to each UE in the cell; the per-subcarrier transmit power of UE k is

[0134] i) Total transmit power constraint condition: wherein Lγ k and represent the total transmit power of all subcarriers of PTP and DTP respectively, P max is the maximum total transmit power;

[0135] ii) Transmit power ratio constraint condition: such that the equivalent channel matrices of PTP and DTP are:

[0136]

[0137] satisfy Based on this condition, after estimating the estimate of the equivalent channel matrix G of PTP, the BS can directly obtain the estimate of the equivalent channel matrix of DTP, i.e.

[0138] Based on the above two constraints, the transmission power calculation method of UE k is:

[0139]

[0140] 2) Channel estimation and channel equalization:

[0141] According to Figure 3 , the BS detects the uplink data of the UE by using the estimated value of PI and the estimated value of TO ; according to formula (9) and formula (13), the least square estimation of the equivalent channel matrix of PTP is:

[0142]

[0143] where (·) + represents the pseudo-inverse operation; according to the aforementioned conditions and formula (11), the estimated value of the uplink data symbol can be calculated by the following formula:

[0144]

[0145] where represents the zero-forcing equalizer, and “ * ” represents the conjugate;

[0146] Finally, is sent to the channel decoder to complete the uplink data decoding;

[0147] C. Preamble detection algorithm:

[0148] Since the performance of channel estimation and channel equalization operations directly depends on the quality of the estimated value of PI and the estimated value of TO , the performance of preamble detection plays a very important role in the grant-free RA process; the main goal of the present application is to design a new type of preamble algorithm to obtain accurate estimates of PI and TO;

[0149] According to the different purposes and performances of algorithm design, the PD algorithm proposed in the present application can be specifically divided into several sub-algorithms, including:

[0150] 1) ML preamble detection algorithm: the design purpose of this algorithm is to obtain the theoretically optimal performance; it can be used as the performance upper bound of other similar algorithms, or can be used in scenarios with sufficient computing power support and requiring absolute performance advantage;

[0151] 2) AML preamble detection algorithm: the design purpose of this algorithm is to reduce the computational complexity of the ML preamble detection algorithm; it improves the computational efficiency by using a single-source approximation method, while still maintaining good detection performance;

[0152] 3) DS-AML preamble detection algorithm: the design purpose of this algorithm is to further improve the TO estimation accuracy of the AML preamble detection algorithm; it is based on the binary search algorithm, and increases the search range of TO without increasing the FFT length, so that more accurate TO estimation can be obtained;

[0153] To distinguish the estimated values of PI and TO calculated by different algorithms The outputs of the three algorithms will be denoted as and the κth element of will be denoted as

[0154] Since the number of UEs is required for the three algorithms, the UE number estimation process will be introduced first, and then the design of the three algorithms will be introduced respectively in combination with the embodiments;

[0155] The traditional threshold-based PD algorithm does not need to estimate the number of UEs in advance, while the PD algorithm based on the ML criterion needs to use the prior information of the number of UEs; specifically, the MDL criterion is used to estimate the number of UEs in the present application; the MDL criterion is an effective method for estimating the number of target sources from a signal containing noise, and has the advantages of not needing to set a threshold and small estimation error; in the present application, first, the part with larger target received power in the received signal of the base station is selected, and the sample covariance matrix R is calculated:

[0156]

[0157] Then, the eigenvalues {λ m |m=1,…,M} of R are calculated, and arranged in the order of λ1≥…≥λ M The estimated value of the number of UEs can be calculated by the following formula:

[0158]

[0159] Wherein, the definition of D(μ) (μ=0,…,M-1) is as follows:

[0160]

[0161] Based on formula (18), the estimated value of the number of UEs can be calculated as follows: for subsequent preamble detection algorithms.

[0162] Embodiment 1

[0163] In this embodiment, the ML preamble detection algorithm is used to estimate the preamble sequence and time offset of the signal sent by each user equipment, and obtain the PI estimation value and TO estimation value of each user equipment

[0164] In the implementation process, the principle of ML detection is to select the parameter value that maximizes the probability density function as the estimate of the target parameter; substituting equation (13) into equation (9) can obtain:

[0165]

[0166] Since W is a matrix composed of Gaussian white noise with power v, the probability density function of Y is:

[0167]

[0168] In the actual system, the noise power v at the BS end may be known or unknown, so the following two cases are discussed;

[0169] 1) Case 1: The noise power v at the BS end is known:

[0170] In this case, according to equation (21), the ML estimate of the unknown parameters {q, τ, G} is:

[0171]

[0172] where respectively represent the trial values of the unknown parameters {q, τ, G}; under the condition that , the solution of equation (22) with respect to is:

[0173]

[0174] Substituting into equation (22) obtains:

[0175]

[0176] where:

[0177] P M = M (M H M) -1 M H (25)

[0178] denotes the projection matrix of matrix M, substituting into equation (25) can obtain In addition, the fifth equation of equation (24) uses the following two properties:

[0179]

[0180] 2) Case 2: The noise power v at the BS end is unknown:

[0181] In this case, the ML estimate of the unknown parameters {q, τ, G, v} is:

[0182]

[0183] where denotes the trial value of v; under the given condition, the solution of equation (28) with respect to is:

[0184]

[0185] Substitute and equation (21) into equation (28), and use equation (29) to obtain:

[0186]

[0187] By comparing equation (22) and (30), it can be found that the ML estimation of {q, τ} has the same expression under the conditions of known and unknown noise power v at the BS end; therefore, according to equation (24), the ML preamble detection algorithm under the conditions of known and unknown noise power can be uniformly expressed as:

[0188]

[0189] The ML preamble detection algorithm shown in equation (31) has no closed-form solution, and therefore needs to be solved by exhaustive search of all possible combinations of and Note that, since T = [0, τ max ] is a continuous interval, all values in cannot be exhaustively searched; therefore, in the algorithm implementation, a set with a limited number of elements needs to be used instead of T, and at this time the search space of τ becomes In order to accurately estimate τ, the set should be obtained by dense sampling on the interval T;

[0190] Finally, by exhaustively searching all possible combinations of and , equation (31) is solved, and the PI estimation value and the TO estimation value are obtained.

[0191] Embodiment 2

[0192] In this embodiment, the AML preamble detection algorithm is used to estimate the preamble sequence and time offset of the signal sent by each user equipment, and the PI estimation value and TO estimation value of each user equipment are obtained

[0193] In the implementation process, it can be seen from Example 1 that, in order to solve formula (31), it is necessary to substitute the values of the possible combinations of the set into the objective function The values of the objective function are calculated, wherein |·| represents the number of elements in the set; since the calculation complexity grows exponentially with the estimated number of UEs , when the value of is large, directly solving formula (31) is difficult to implement in an actual system; therefore, this embodiment further proposes a suboptimal algorithm version of the AML preamble detection algorithm which is easier to implement in an actual scenario on the basis of the ML preamble detection algorithm;

[0194] The approximation principle of the AML preamble detection algorithm is as follows:

[0195] The signal model shown in formula (20) has high similarity with the signal model of the DOA estimation problem; more specifically, the unknown parameter combination (q k ,τ k ) of the UE k can be analogized as the DOA of the kth signal source, and the M elements of the kth row of the equivalent channel matrix G can be analogized as the M fast shots of the signal emitted by the kth signal source; and similar to the PD problem based on the ML criterion, the DOA estimation based on the ML criterion also has the problem of excessively high calculation complexity, so different suboptimal methods for this problem have been proposed; inspired by the similarity between the PD and the DOA estimation problem, some suboptimal ML DOA estimation methods can be used to solve the PD problem considered in this embodiment;

[0196] Since the single-source approximation method of the ML DOA has high calculation efficiency, this embodiment adopts this method; the basic principle of the single-source approximation method is to search for multiple local maximum values of the ML objective function under the single-signal-source assumption, and the positions of these local maximum values are regarded as the DOAs of multiple signal sources; it is noted that in the DOA problem, the DOA of each signal source is a continuous variable; while in the PD problem, the PI q k of the UE k is a discrete variable, because Q = {1,..., Q} is a discrete set; therefore, the single-source approximation method of the DOA estimation needs to be improved and adapted to the PD problem faced by this embodiment;

[0197] Based on the principle of the single-source approximation method, this embodiment first calculates the ML objective function when a single UE, i.e., UE 1, accesses; in this case, the objective vector combination becomes a scalar combination , and the matrix combination also correspondingly becomes a vector combination Therefore, referring to the definition of formula (25), formula (31) can be converted into: ​​

[0198]

[0199] Where: Tr(·) represents the trace of the matrix, and the sixth equation uses and The modulus of the elements is 1. According to formula (32), based on the principle of single-source approximation, the objective function is defined as:

[0200]

[0201] Equation (33) can be viewed as the received signal Y and the preamble sequence affected by the phase offset The correlation value between Will be in and A larger value appears at

[0202] Note that a significant difference between the DOA estimation problem and the PD problem is that in the former, DOA is a continuous variable, but the latter includes the discrete variable PI; from Equation (33), we can see that PI Only used to indicate the leading sequence The serial number does not participate in any mathematical operations, so A better way to treat it is to treat it as a sequence number The independent variable is As shown in formula (33), the single variable function instead of As a function notation, to emphasize is a single variable function; based on this processing method, an approximate solution to the ML leader detection algorithm shown in formula (31) is to find the function middle The positions of the largest local maxima are recorded as The corresponding TO is recorded as This algorithm is named AML leader detection algorithm, and formula (33) is called AML objective function;

[0203] Similar to the ML leader detection algorithm in Equation (31), when using the AML leader detection algorithm to solve, it is necessary to use a discrete set Instead of the continuous interval T; in general, the AML leader detection algorithm needs to calculate the objective function The number of times is However, the special form of the phase offset vector a(·) can be used to reduce the amount of calculation with the help of FFT;

[0204] Specifically, for a general TO variable τ∈T, the normalized TO is defined as:

[0205] ε = τNΔf (34)

[0206] where N ≥ L is the FFT length; similarly, for the τ-related variables already used in Embodiment 1, their normalized versions can be defined as follows, respectively:

[0207] Maximum normalized TO for a UE within a cell: ε max = τ max NΔf;

[0208] Range of normalized TO: E = [0, ε max ];

[0209] Normalized TO for UE k: ε k = τ k NΔf, k = 1, …, K;

[0210] Vector of normalized TO for K UEs: ε = [ε1, …, ε K ] T ∈ E K×1 ;

[0211] Trial values of normalized TO in AML algorithm:

[0212] Estimated value of ε using AML algorithm:

[0213] According to equation (3) and equation (34), the phase offset vector with normalized TO ε as variable is defined as:

[0214]

[0215] Using equation (35), the AML objective function of equation (33) can be rewritten as:

[0216]

[0217] where is defined as:

[0218]

[0219] From equation (35) and equation (36), it can be observed that, When is an integer, the value of can be obtained by computing the FFT of

[0220] According to equation (36), finding the local maximum of is equivalent to finding the local maximum of ; for FFT-based implementation, it is specified that The conditions that should be met when it is a local maximum are:

[0221]

[0222] in, The value range is Compared with formula (36), formula (38) The upper bound of the range of values ​​is from ε max Expanded to The reason is that FFT can only calculate When it is an integer The value of ε max is not necessarily an integer; in addition, to make all Whether it is a local maximum can be determined by formula (38), and the function Add the following definition:

[0223] 1) When ε max If is not an integer, then at this time There is no definition, so according to formula (36), the definition is supplemented:

[0224] 2) Due to and Undefined, cannot be judged using formula (38) and Is it a local maximum? Therefore, we need to define

[0225] In summary, the AML objective function of formula (36) is expanded to:

[0226]

[0227] Based on the above discussion, the AML leader detection algorithm proposed in this embodiment is summarized in Algorithm 1;

[0228]

[0229] In the AML leader detection algorithm, according to the discrete set Determine the experimental value of normalized TO The value range of is solved by fast Fourier transform All values ​​of , output local maxima, and the PI corresponding to each local maximum is used as the PI estimate Use the normalization corresponding to each local maximum Further calculation of TO estimate

[0230] Example 3

[0231] In the embodiment, the preamble sequence and time offset of the signal transmitted by each user equipment are estimated by using the DS-AML preamble detection algorithm, and the PI estimation value and TO estimation value of each user equipment are obtained

[0232] In the specific implementation process, in the AML preamble algorithm of embodiment 2, the trial value of the normalized TO is an integer, and the value range is The estimation resolution is 1; according to the definition of the normalized TO in formula (34), the original TO estimation value of the AML algorithm has a resolution of which is limited by the FFT length N, and using a larger FFT length N helps to obtain a more accurate TO estimation, and also increases the calculation complexity; referring to formula (15) and formula (16), the channel estimation and channel equalization both need to use the estimation value of the TO Therefore, accurate TO estimation is very important for data detection in the unlicensed RA; therefore, in order to further improve the TO estimation accuracy of the AML preamble detection algorithm, the DS-AML preamble detection algorithm is designed based on the idea of binary search in the embodiment; it is pointed out here that the DS-AML algorithm does not affect the estimation of PI, so the PI estimation value provided is the same as that of the AML algorithm, that is

[0233] The following specifically introduces the use of the DS-AML algorithm to obtain a more accurate TO estimation value;

[0234] Binary search is a low-complexity estimation method that can be used to solve the frequency estimation problem in the PD process; by substituting formula (6) into formula (37), the following can be obtained:

[0235]

[0236] wherein represents a set of UEs that select the preamble sequence , and

[0237]

[0238] represents the corresponding additive interference and noise when the preamble sequence is detected; it can be known from the definition of the phase rotation vector in formula (35) that the obtained from formula (40) has similarity with the signal model in which a plurality of complex exponential sequences with different frequencies are superimposed, wherein can be analogous to the frequency of different complex exponential sequences; therefore, the binary search method designed for frequency estimation can be used to estimate the normalized TO of the UE in the scenario considered in the present application.

[0239] In this embodiment, based on the estimated number of UEs and UE selected The BS performs a binary search operation on the TO of each UE one by one. The binary search is an iterative operation process under the condition of a specified number of iterations I. The estimated value of the normalized TO of UEκ at the i-th (i=1,…,I) iteration is recorded as Its initial value is given by the AML algorithm, that is, In the binary search process, according to The function given by formula (39) is searched in the order of exist During the iteration process, it is necessary to continuously update the three variables {ψ0,ψ - ,ψ +}, used to determine Update direction; {ψ0,ψ - ,ψ + The initial value of} is set to:

[0240]

[0241] During the binary search process, the search step size of each iteration becomes the size of the previous iteration. Because the normalized TO estimation initial value resolution provided by the AML leader detection algorithm is 1, the search step size of the i-th iteration of the binary search is 2 -i ; According to the i-th iteration ψ - and ψ + The size relationship, The update method is:

[0242]

[0243] Finish After the update of {ψ - ,ψ +} to determine the update direction of normalized TO in the next iteration; the update rule is: if ψ + >ψ - , then ψ - ←ψ0,ψ + remains unchanged; otherwise ψ + ←ψ0,ψ - remains unchanged; finally, ψ0 is updated to:

[0244]

[0245] Based on the above discussion, the DS-AML leader detection algorithm proposed in this embodiment is summarized in Algorithm 2;

[0246]

[0247]

[0248] Example 4

[0249] This embodiment provides a simulation experiment of a specific scenario to test the performance of the three preamble detection algorithms provided in Embodiments 1 to 3.

[0250] In the specific implementation process, this embodiment considers a marine communication system with a cell radius of R and a center frequency of f c The BS is located on an island in the center of the cell, and the UEs are randomly distributed on the sea level. Unless otherwise stated, the simulation parameters are given in Table 1.

[0251] Table 1 Simulation parameters of this embodiment

[0252]

[0253] In this embodiment, the large-scale fading coefficient of the sea channel is modeled as:

[0254]

[0255] in Indicates the carrier wavelength, A BS and A UE Represent the altitude of BS and UE respectively; small-scale fading coefficient v k,m Modeled as a complex Gaussian random variable with zero mean and unit variance; according to the NR standard, the discrete Fourier transform (DFT) of the ZC sequence is selected as the leading sequence of the PTP:

[0256]

[0257] where c q ∈C L×1 Represents a ZC sequence, whose lth element is F∈C L×L represents the DFT matrix, whose (l1,l2)th element is In the simulation, the RA process of a UE is considered successful only if the UE's uplink data packet is correctly decoded;

[0258] Since the DS-AML leader detection algorithm proposed in Example 3 is an iterative algorithm, its convergence performance is first tested below; Figure 4The curve of the RA success rate of the DS-AML algorithm is drawn, and it can be seen that the RA success rate increases with the increase of I when I≤4, and almost does not change with the increase of I when I>4; therefore, in the subsequent simulation, the iteration number of the DS-AML algorithm is set to I=4;

[0259] As shown in Figure 5 , Figure 5 The RA success rate performance curves of a plurality of preamble detection algorithms under different UE numbers K and cell radii R are given, including the existing FDC algorithm, the UG algorithm, and the AML and DS-AML algorithms; it can be seen that as the FFT length N increases, all algorithms can provide more accurate TO estimation, so the RA success rates of different algorithms are improved; in addition, under the same FFT length N, the RA success rates of the AML and DS-AML algorithms proposed in embodiments 2 and 3 are higher than those of the FDC and UG algorithms; in particular, the performance of the DS-AML algorithm when N=1024 is better than that of the FDC, UG, and AML algorithms when N=2048, which shows that the bisection search operation can significantly improve the TO estimation accuracy, so that the FFT complexity of the system can be reduced under the same performance target requirement;

[0260] In addition, since the ML preamble detection algorithm has an exponential complexity, it is difficult to implement in actual scenarios, and a specific set of parameters will be selected for testing to verify the superiority of the ML algorithm and prove that the AML and DS-AML algorithms proposed in embodiments 2 and 3 are two better approximate methods of the ML algorithm; specifically, Q=K=2 is selected, so that the ML algorithm can be implemented on a computer; in order to ensure fairness, the search space of TO in the ML algorithm is consistent with the search space of TO in the FDC, UG, and AML algorithms and the initial search space of TO in the DS-AML algorithm, that is, This setting makes the search space of the ML preamble detection algorithm change with the FFT length N, even though the ML algorithm does not include the FFT operation;

[0261] As shown in Figure 6 , Figure 6The RA success rate performance of different detection algorithms under Q=K=2 is drawn; when N=1024, the performance of the AML algorithm is close to that of the ML algorithm, and the DS-AML algorithm is supported by binary search, and a more refined search space is supported, so that the performance of the DS-AML algorithm is better than that of the ML algorithm; this is because the FFT length is relatively small at this time, which limits the TO estimation accuracy of the ML algorithm, and the DS-AML algorithm can overcome the influence of small-size FFT through binary search, so that the performance of the DS-AML algorithm is better than that of the ML algorithm; when N=2048, the FFT length is large enough at this time, so the ML algorithm has the optimal performance, and the AML and DS-AML algorithms have the performance close to that of the ML algorithm; Figure 6 It is verified that the ML algorithm has excellent performance when the TO search scale is fine enough, that is, N is increased, and the AML and DS-AML algorithms have the performance close to that of the ML algorithm.

[0262] The same or similar reference signs correspond to the same or similar components;

[0263] The terms describing the positional relationship in the drawings are only used for exemplary illustration, and cannot be understood as a limitation on the patent;

[0264] Obviously, the above embodiments of the present application are only examples for clearly illustrating the present application, and are not a limitation on the embodiments of the present application. Based on the above description, other different forms of changes or variations can be made by those skilled in the art. Here, all the embodiments need not and cannot be exhausted. Any modification, equivalent replacement and improvement made within the spirit and principle of the present application should be included in the protection scope of the claims of the present application.

Claims

1. A method for grant-free random access in a large-scale multiple-input multiple-output system, characterized in that: The following steps are involved: S1: Establishing a wireless communication system model based on orthogonal frequency division multiplexing, in which a plurality of user equipments transmit signals to a base station through different types of PRACH resource blocks that are orthogonal in the frequency domain; The transmission signal includes a preamble sequence and encoded uplink service data; S2: Obtain a received signal from the base station, select a portion of the received signal with a larger target received power, and estimate the number of user equipments using a minimum description length criterion; S3: Using a preset preamble detection algorithm, estimate the preamble sequence and time offset of the signal transmitted by each user equipment, and obtain a PI estimation value and a TO estimation value of each user equipment; where PI is the preamble sequence number and TO is the time offset; The preset preamble detection algorithm is specifically: any one of the ML preamble detection algorithm, the AML preamble detection algorithm and the DS-AML preamble detection algorithm; S4: The base station performs channel estimation for each user equipment using the PI estimation value and the TO estimation value, and performs channel equalization on the received signal to obtain an estimation value of the uplink service data of the corresponding user equipment; S5: The base station decodes the estimated value of the uplink service data of each user equipment, completes the reception of the uplink service data, and implements unlicensed random access.

2. The method for grant-free random access in a large-scale multiple-input multiple-output system according to claim 1, wherein: In step S1, the wireless communication system uses OFDM to transmit signals, wherein part of the time-frequency resources are allocated as a physical random access channel; the transmission of the user equipment's transmit signal in the physical random access channel is divided into a preamble transmission phase and a data transmission phase, and the user equipment's transmit signal constitutes two consecutive OFDM symbols in the time domain; Several PRACH resource blocks are set within the bandwidth of the wireless communication system; in each PRACH resource block, the bandwidth allocated in the preamble transmission phase and the data transmission phase are respectively and ,in is the number of symbols contained in the leading sequence, is the number of data symbols, is the subcarrier spacing; the type of the PRACH resource block includes any of the following: Type 1: Type 2: and type 3: .

3. The method for grant-free random access in a large-scale multiple-input multiple-output system according to claim 2, wherein: In the wireless communication system model of step S1, Each user equipment accesses one base station. Unknown; each user device sends signals through a single transmit antenna, and the base station uses The root receiving antenna receives the signal; the serial number of the user equipment is recorded as ,satisfy User equipment The distance between the base station and , and the maximum distance between the user equipment and the base station is known User equipment TO is , Represents the speed of light, the value range of TO is , ; During the preamble transmission phase, each user equipment receives a preamble sequence from a predefined set of preamble sequences. Select a length of The leading sequence is sent, where is the number of leading sequences, Indicates the serial number is The leading sequence of Represents a plural set; user equipment The selected leader sequence is denoted as , the leading sequence sent is ; In the preamble transmission phase, the base station The received signal of the root antenna is recorded as , , specifically: in," " indicates Hadamard. It is a user device The transmit power of each subcarrier during the preamble transmission phase, For user equipment in a PRACH resource block and base station antennas The channel coefficients between and represent the large-scale and small-scale fading coefficients respectively; Indicates that the preamble transmission phase is The phase shift caused by is the Gaussian white noise vector in the preamble transmission phase; During the data transmission phase, the user equipment Generate data symbol vectors by constellation mapping the encoded uplink service data And send, the base station receiving signal in the data transmission phase is recorded as , specifically: in, It is a user device The transmit power per subcarrier during the data transmission phase; Indicates that the data transmission phase is The phase shift caused by is the Gaussian white noise vector in the data transmission phase; Will and Convert them into matrix forms respectively to obtain the wireless communication system model: in, ; in, .

4. The method for grant-free random access in a large-scale multiple-input multiple-output system according to claim 3, wherein: In step S2, estimating the number of user equipments using the minimum description length criterion includes the following steps: Select the part of the base station's received signal with a larger target received power and calculate the sample covariance matrix : in, and The target received power of the signal sent by the user equipment during the preamble transmission phase and the data transmission phase reaching the base station; calculate The eigenvalue of ,according to The estimated number of user devices is arranged in the order of Calculated by the following formula: in, is an intermediate variable, ; is an intermediate function.

5. The method for grant-free random access in a large-scale multiple-input multiple-output system according to claim 4, wherein: In step S3, the objective function of the ML leader detection algorithm is expressed as: in, is a matrix The projection matrix; Indicates finding the square of the F norm of the matrix; and are the test values ​​of PI and TO respectively; and They represent the PI estimation value and TO estimation value output by the ML leader detection algorithm respectively; In the continuous interval Dense sampling on the upper layer to obtain a discrete set , the discrete set As the new value range of TO; and Exhaustively search all possible value combinations to solve the objective function of the ML leader detection algorithm and obtain the PI estimate and TO estimates .

6. The method for grant-free random access in a large-scale multiple-input multiple-output system according to claim 5, wherein: In step S3, the ML leader detection algorithm is improved based on the single-source approximation method to obtain the AML leader detection algorithm; The objective function of the AML leader detection algorithm is expressed as: in, is the intermediate matrix, , for The conjugate vector of ; is the experimental value of normalized TO, which is recorded as , is the fast Fourier transform length, for The maximum value of For is the phase shift of the variable, ; Indicates rounding up; According to the discrete set Determine the experimental value of normalized TO The value range of is solved by fast Fourier transform All values ​​of , output local maxima, and the PI corresponding to each local maximum is used as the PI estimate , using the normalized TO corresponding to each local maximum Further calculation of TO estimate .

7. The method for grant-free random access in a large-scale multiple-input multiple-output system according to claim 6, wherein: In step S3, the accuracy of TO estimation in the AML leader detection algorithm is further improved based on the binary search method to obtain the DS-AML leader detection algorithm; The PI estimate output by the DS-AML leader detection algorithm ; In the DS-AML preamble detection algorithm, based on the estimated number of user devices and user devices Selected PI , the base station performs a binary search operation on each user equipment for the normalized TO, and the number of iterations of the binary search is ;in, for The elements; User equipment The normalization of TO in the The estimated value at the iteration is recorded as , whose initial value is given by the AML leader detection algorithm, that is, , for The elements; in the binary search process, according to The objective function of the AML leader detection algorithm is searched in the order of exist The peak value near the edge of the grid is obtained, and the three variables are continuously updated during the iteration process. , used to determine Update direction; The initial value of is set to: During the binary search process, the search step size of each iteration becomes the size of the previous iteration. According to the At the iteration and The size relationship, The update method is: Completed in one iteration After the update, further updates are required , to determine the update direction of normalized TO at the next iteration: If ,but , Remain unchanged; otherwise , remain unchanged; at the same time Updated to: go through After iterations, the user device The estimated value of normalized TO is , the estimated value of TO is ; The normalized TO estimate output by the DS-AML leader detection algorithm is , further calculate the TO estimate .

8. The method for grant-free random access in a large-scale multiple-input multiple-output (MIMO) system according to any one of claims 5 to 7, wherein: In step S4, an equivalent channel matrix is ​​constructed for each user equipment in the preamble transmission phase. and the equivalent channel matrix during the data transmission phase : and satisfy , ; The base station uses the PI estimation value and TO estimates Estimate the equivalent channel matrix of each user equipment during the preamble transmission phase , specifically: in, for estimated value of; represents the pseudo-inverse operation; Further calculate the equivalent channel matrix in the data transmission stage Estimated value of : , completing channel estimation.

9. The method for grant-free random access in a large-scale multiple-input multiple-output system according to claim 8, wherein: User Equipment The transmit power is calculated as follows: in, is the maximum total transmit power.

10. The method for grant-free random access in a large-scale multiple-input multiple-output system according to claim 9, wherein: In step S4, channel equalization is performed on the received signal to obtain an estimated value of the uplink service data of the corresponding user equipment. , expressed as: in, represents the zero-forcing equalizer, ” denotes matrix conjugation.

Images