A generalized superposed pilot optimization method for iot urllc service

Abstract

The application discloses a generalized superposed pilot optimization method for an Internet of Things URLLC service, constructs a transmission frame structure of a generalized superposed pilot, estimates a channel based on the transmission frame structure, adopts a detector to perform data detection based on a channel estimation result, and obtains a closed-form expression of a lower bound of a reachable rate, constructs a weighted sum rate maximization mathematical model of joint optimization of pilot power and data power according to the closed-form expression, and finds a local optimal solution of the problem by using an iterative algorithm. The method effectively reduces total energy consumption and improves short packet transmission performance under the premise of meeting a reachable rate of URLLC transmission.

Description

Technical Field

[0001] This invention belongs to the field of Internet of Things (IoT) and relates to a generalized superimposed pilot optimization method for IoT URLLC services. Background Technology

[0002] Ultra-low latency, high-reliability communication (URLLC) is a typical IoT application scenario, such as autonomous vehicles, virtual reality, and the haptic internet. Typical key performance indicators for URLLC include latency (air interface latency less than 1ms) and reliability (bit error rate less than 10^6). -5 This will be achieved in two parts, which will increase performance by an order of magnitude in 6G. However, due to the conflicting and difficult-to-achieve objectives, related research is still in its early stages.

[0003] Low latency implies a finite block length or a finite number of channels. In other words, the size of the data packets or the length of the codewords are very short, unlike traditional communication methods. Therefore, the classic Shannon theorem no longer applies in the case of short packet transmission. Consequently, some have proposed introducing URLLC reachability rate to characterize this.

[0004] Many existing URLLC works with short packet transmission rely on the assumption of achieving perfect channel state information (CSI) through negligible pilot overhead. However, this assumption is impractical. Due to the finite block length, the impact of pilot overhead becomes more significant when the packet length is shortened. Currently, there are three pilot transmission schemes in URLLC research: conventional pilot (RP), superimposed pilot (SP), and generalized superimposed pilot (GSP). In the RP scheme, pilots and data are transmitted separately, resulting in low resource efficiency; in the SP scheme, pilots and data occupy the same transport block, leading to significant mutual interference; the GSP scheme shortens the data sequence, utilizes partially orthogonal sequences to construct the precoding matrix, and reduces mutual interference by optimizing the data length.

[0005] In view of this, it is indeed necessary to design a generalized superimposed pilot optimization method for IoT URLLC services to solve the above problems. Summary of the Invention

[0006] The technical problem to be solved by this invention is to overcome the shortcomings of the prior art and provide a short packet transmission method in a multi-user uplink mMIMO system based on URLLC transmission in a large-scale wireless network environment. Under the condition of short packet transmission in a multi-user uplink mMIMO system with given delay and decoding error probability, a pilot scheme is proposed to obtain channel state information and eliminate MI. On this basis, a weighted sum rate maximization problem under given delay and reliability objectives is proposed, a resource allocation strategy is designed, and a local optimal solution of the problem is found through an iterative algorithm to achieve joint optimization of variables such as data length, pilot power, and data power under minimum rate and energy constraints.

[0007] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0008] In the notation, lowercase bold letters represent vectors (e.g., x), and uppercase bold letters represent matrices (e.g., X). H ,‖x‖, tr{x} and R e {x} denote the conjugate transpose (vector), norm, expected value, trace, and real part of the complex matrix (vector), respectively, denoted by X. -1 The inverse matrix of X is represented by... To represent a circularly symmetric complex Gaussian distribution, use... The set of complex numbers.

[0009] In the system model, it is assumed that a single cell system consists of one M-antenna base station and K single-antenna users, with the user set being K = {1, 2, ..., K}. These short data packets are transmitted simultaneously by the users using the same BHz bandwidth. To implement URLLC, it is stipulated that the decoding error probability ε for each transport block does not exceed the channel usage T.

[0010] In the channel model, because the data packet length is short, the data packet transmission time is negligible compared to the channel coherence time. It is constant within the same transport block, but varies in different transport blocks. Channel fading follows a Rayleigh distribution. Where α i I represents the large-scale fading coefficient between the i-th user and the base station. M It is the covariance matrix, representing the small-scale fading between the user and the base station.

[0011] The core idea of ​​the GSP scheme is to select a portion of column vectors from an orthogonal matrix as user pilots, and construct a precoding matrix with the remaining columns. This precoding matrix is ​​adapted to the shortened user data sequence. By feeding the data sequence into the precoding, it is ensured that the pilots and user data are orthogonal, thereby eliminating MI.

[0012] Step 1: constructing a transmission frame structure of generalized superimposed pilots;

[0013] Step 2: performing channel estimation based on the transmission frame structure;

[0014] Step 3: performing data detection by using a detector based on the channel estimation result, to obtain a closed-form expression of the lower bound of achievable rate;

[0015] Step 4: constructing a weighted sum rate maximization mathematical model for jointly optimizing pilot power and data power according to the closed-form expression;

[0016] Step 5: finding a local optimal solution of the model by using an iterative algorithm.

[0017] As a further improvement of the present invention, the frame length of the transmission frame in step 1 is T, K (K<T) columns are selected from a T×T orthogonal matrix as the pilot sequences of users, τ columns are selected from the remaining T-K columns to construct a precoding matrix W for a user and applied to user data, and the following conclusion is obtained according to the relationship between the precoding matrices of different users:

[0018]

[0019] wherein Φ ij represents a permutation matrix, when i=j, the permutation matrix is an identity matrix and has full rank, I τ is an identity matrix.

[0020] As a further improvement of the present invention, the channel estimation in step 2 adopts the MMSE criterion, and the calculation formula is as follows:

[0021]

[0022]

[0023]

[0024] wherein y k is a Gaussian signal, ρ k is the normalized pilot transmit power of the k-th user, T is the number of channel uses, h k is the channel between the k-th user and the base station, is the solution of MMSE channel estimation, n k is noise and follows a complex Gaussian distribution, λ k is the weighting coefficient of linear estimation that minimizes the mean square error between the channel estimation value and the true value, is the expectation of matrix , is the expectation of matrix , is yk The conjugate transpose of the matrix. It is a matrix The inverse matrix.

[0025] As a further improvement of the present invention, the detectors in step 3 include, but are not limited to, a maximum ratio merging detector and a zero-forcing detector;

[0026] When using the aforementioned maximum ratio combined MRC detector, the closed-form expression for the achievable rate lower bound is:

[0027]

[0028] in,

[0029]

[0030]

[0031]

[0032] in, In the case of an MRC detector, the achievable rate lower bound LB in the closed form is given by τ = TK, where τ is the user data length. It is the effective signal-to-interference-plus-noise ratio (SINR) of the k-th user under the MRC detector condition; It is about A specific functional form, where ε is the decoding error probability, and Q... -1 (·) is the inverse function of the Gaussian Q(·) function. It is about The channel dispersion function; M is the number of antennas at the base station, η k It is the normalized transmit power of the k-th user, α k It is a large-scale fading between the k-th user and the base station, r ik Describing the permutation matrix Φ ik The rank of the precoding matrix W of the i-th user is... i and the precoding matrix W of the k-th user k There is r ik There are three identical columns, K = {1, 2, ..., K}, where K is the set of users and K is the number of users per antenna.

[0033] When using the zero-forcing ZF detector, the closed-form expression for the achievable rate lower bound is:

[0034]

[0035] in,

[0036]

[0037]

[0038]

[0039] in, The achievable rate lower bound (LB) in closed form under the ZF detector reflects the impact of SINR, the number of transmitted data symbols, the amount of channel usage, and the decoding error probability on the achievable rate. It is the effective signal-to-interference-plus-noise ratio (SINR) of the i-th user under the ZF detector condition; It is about A specific function form, It is about The channel dispersion function;

[0040] As a further improvement of the present invention, the weighted sum rate maximization mathematical model in step 4 is as follows:

[0041]

[0042]

[0043]

[0044] 1≤τ≤TK,τ∈N+

[0045] in, μ k R is the weight of the k-th user. k It is a closed-form expression for the reachable rate of the k-th user, when using a maximum ratio merging detector. When using a zero-forcing detector R min E and E are the minimum achievable rate and minimum energy consumption of the optimization model, respectively.

[0046] As a further improvement of the present invention, the iterative algorithm in step 5 specifically includes the following steps:

[0047] Step 51: Transform the mathematical model P1 into a geometric programming (GP) problem;

[0048] Step 52: Transform the GP problem into a convex form through logarithmic transformation, and then propose an iterative algorithm for solving local optima;

[0049] Step 53: Initialize the iteration count n=1, error probability ε, and feasible power allocation of the iterative algorithm.

[0050] Step 54: Execute the iterative algorithm based on the initial value of the iterative algorithm to find the optimal solution to the convex problem.

[0051] Compared with the prior art, the present invention, employing the above technical solution, has the following beneficial effects:

[0052] Transmitting data and pilot signals via the GSP scheme can reduce data length and thus eliminate latency, while also improving SINR and eliminating noise caused by mutual interference between pilot signals and data. Under the premise of meeting URLLC reachability, the optimal data length is determined, and the overhead of uplink data transmission and pilot signal transmission is reduced, thus achieving effective management of resource allocation. Attached Figure Description

[0053] Figure 1 This is a pilot frame structure diagram of the GSP scheme;

[0054] Figure 2 This is a flowchart of a short packet transmission method in a multi-user uplink mMIMO system based on URLLC transmission in a large-scale wireless network environment.

[0055] Figure 3 This is a graph showing the number of antennas and the weighted sum rate under three different schemes. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0057] It should also be noted that, in order to avoid obscuring the present invention with unnecessary details, only the structures and / or processing steps closely related to the present invention are shown in the accompanying drawings, while other details that are not closely related to the present invention are omitted.

[0058] Consider uplink short packet transmission in a single-cell system. The system consists of a base station and users. The base station (BS) contains M antennas, and all users are single-antenna users. The user set is K = {1, 2, ..., K}.

[0059] Before performing joint optimization, the specific implementation method of the GSP scheme will be explained first. The framework structure diagram of the GSP scheme in this example is as follows. Figure 1As shown in the figure, where T represents the number of channel uses, τ represents the data length, and each transmission block does not exceed T channel uses. Assuming K<T, the base station selects K columns from the T×T orthogonal matrix as users' pilots. For the i-th user, τ columns need to be selected from the remaining T-K columns of the T×T orthogonal matrix as the user's precoding matrix to precode the user's short packet data. According to the form of an orthogonal matrix, data in different rows (columns) are orthogonal, which ensures that the precoding matrix is orthogonal to the pilot, thus eliminating MI. When τ=T-K, each user can only have one precoding matrix; while when Y>T-K, the pilot and data are not orthogonal, MI increases rapidly, leading to severe deterioration of SINR.

[0060] As Figure 2 shown, it is a flowchart of a short packet transmission method in a multi-user uplink mMIMO system based on URLLC transmission in a large-scale wireless network environment provided by an embodiment of the present invention, the method includes the following steps:

[0061] Step 1: Implement a GSP scheme and allocate specified power gains to data and pilots. For the i-th user, the uplink transmission power satisfies the following constraint:

[0062]

[0063] wherein, the transmitted signal represents a superimposed signal of pilot data and precoded data occupying T channel uses, which can be expressed as represents a zero-mean data vector and satisfies

[0064] Step 2: Channel estimation. At the base station end, despread the received data, remove interference caused by user data, and estimate channel characteristics according to the received signal. It includes several sub-steps:

[0065] Step a: Multiply the received signal Y at the base station end by to generate:

[0066]

[0067] wherein Y can be expressed as represents an additive white Gaussian noise matrix, and elements in the matrix follow a complex Gaussian distribution with a mean of 0 and a variance of 1, y k is a Gaussian signal;

[0068] Step b: Use an MMSE estimator to obtain the corresponding channel estimation value:

[0069]

[0070] through matrix operation, we can obtain

[0071]

[0072] Step 3: Data Detection. This includes several sub-steps:

[0073] Step a: The channel estimation matrix is The receiver matrices of the MRC and ZF detectors can then be written as follows:

[0074]

[0075] Step b: Use a receiver matrix to process the received signal to obtain... in Represents the estimated value of the user dataset;

[0076] Step c: Restore the data of the kth user to

[0077]

[0078] Where ψ k This represents the k-th column of Ψ. This represents the estimated value of the k-th user's data;

[0079] Step d: Using the Use-Forget (UatF) technique, approximate the recovered data with the average value of the effective channel gain, and rewrite it as...

[0080]

[0081] Wherein, the combined vector and effective noise are defined as follows:

[0082]

[0083]

[0084] in

[0085] Step e: The effective SINR of the k-th user can be represented as:

[0086]

[0087] Step 4: Establish a planning model based on the weighted sum rate maximization problem, including joint optimization under MRC and ZF conditions. First, solve for the optimal data length. Then, transform the mathematical model into a form that facilitates problem solving. Finally, design an iterative algorithm to find the local optimum of the optimal power allocation. Finally, perform convergence analysis of the algorithm. The specific implementation is as follows:

[0088] Step a: Establish a weighted sum rate maximization problem that jointly optimizes pilot power and data power:

[0089]

[0090]

[0091]

[0092] 1≤τ≤TK,τ∈N+

[0093] Step b: Use Replace R in P1 k According to positive value With the lemma of monotonically increasing, the optimization problem in P1 can be rewritten as:

[0094]

[0095]

[0096]

[0097] Step c: Solve the above problem using an algorithm based on logarithmic functions and continuous convex approximation. This algorithm first introduces auxiliary variables. P2 can be transformed into the following equivalent problem:

[0098]

[0099]

[0100]

[0101]

[0102] in and

[0103] Step d: Continue to simplify the objective function, scale it using the inequality theorem, and apply the inequality formula.

[0104] P(y)≤σln(y)+θ

[0105] as well as

[0106]

[0107] Where y represents the independent variable of the inequality function, and here y = v k , Obtain the LB of the objective function in P3.

[0108] Step e: The variable ρ in the nth iteration k η kand v k Represented as and The lower bound of the objective function in the (n+1)th iteration is obtained as follows:

[0109]

[0110] in, and It is The replaced parameters, and LB strictly satisfies them. Therefore, a new optimization problem is obtained:

[0111]

[0112]

[0113]

[0114]

[0115] in, and constraints Deleted.

[0116] Step f: P4 is transformed into the following GP problem:

[0117]

[0118]

[0119]

[0120]

[0121]

[0122] Step g: To trigger the iterative process, an initial solution needs to be provided for the algorithm, which can be obtained by solving the following optimization problem:

[0123]

[0124]

[0125]

[0126] Here, v is an auxiliary variable that determines the availability of the initial feasible solution. Only when v ≥ 1 can the solution of P6 be used to initialize the problem of P5.

[0127] The specific steps of the ZF joint optimization method are as follows:

[0128] Step a: Transform the P2 problem into a GP problem. The SINR constraint in the ZF case is not a standard GP constraint because the denominator of the SINR expression is not a positive term. Further processing is needed, employing a successive convex approximation method to construct a monomial function to approximate the positive term in each iteration. The sub-steps are as follows:

[0129] (1) Using the following lemma: In the l-th iteration, the positive term is expressed as T(x) = ∑ i t i (x), to obtain

[0130]

[0131] in, And x l-1 It is the optimal solution obtained in the (l-1)th iteration, and the boundary strictly satisfies x = x l-1 In the denominator of SINR The denominator of this expression is a positive term, therefore it can be used for... The denominator is approximated using a lemma and a monomial function, as shown in the following expression:

[0132]

[0133] And have

[0134]

[0135] in, It is the solution for the pilot power of i users in the l-th iteration;

[0136] (2) Using an approximation, γ in the ZF case k The denominator of the expression is transformed into a monomial, and then the SINR constraint in the (l+1)th iteration is rewritten as follows:

[0137]

[0138] (3) Apply the same transformation to P2 as in the MRC case, and express the GP problem in the ZF case as follows:

[0139]

[0140]

[0141]

[0142]

[0143] in,

[0144] Step b: Initialize the pilot power and data power for the 0th iteration, and solve the P7 iterative algorithm.

[0145] Step 5: Perform performance testing of the optimization scheme from a simulation implementation perspective. Consider a rectangular simulated area where base stations are located at the center, and user locations are randomly generated and follow a uniform distribution. Without loss of generality, the path loss model can be expressed as PL k =35.3 + 37.6log 10 d k (dB), noise power spectral density is -174dBm / Hz, decoding error probability is 10. -9 It includes the following two sub-steps:

[0146] Step a: As Figure 3 This verifies that the optimal data length in MRC and ZF cases is τ = TK, and that when the data length τ > TK, the data rate drops sharply, and the image features are consistent with the theoretical analysis.

[0147] Step b: Under the same settings, simulations were performed on the three schemes RP, SP, and GSP respectively. The simulation results were compared and analyzed to determine the impact of the base station antenna on the weighted sum rate.

[0148] like Figure 3 As shown, the system performance of all three schemes improves with the increase in the number of base station antennas. The weighted sum rate of the GSP scheme is significantly higher than the other two schemes under the same conditions, demonstrating the superiority of the GSP scheme. Although the SP scheme outperforms the GSP scheme when the number of antennas is very large, considering the cost of antenna deployment, having too many antennas is impractical in real-world scenarios.

[0149] In summary, transmitting data and pilot signals via the GSP scheme can reduce data length and thus eliminate latency, while also improving SINR and eliminating noise caused by mutual interference between pilot signals and data. Under the premise of satisfying URLLC reachability, the optimal data length is determined, and the overhead of uplink data transmission and pilot signal transmission is reduced, thus achieving effective management of resource allocation.

[0150] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A generalized superimposed pilot optimization method for Internet of Things (IoT) URLLC services, characterized in that, Includes the following steps: Step 1: Construct the transmission frame structure of generalized superimposed pilots; Step 2: Estimate the channel based on the transmission frame structure; Step 3: Based on the channel estimation results, use a detector to perform data detection and obtain a closed-form expression for the lower bound of the achievable rate; Step 4: Construct a weighted sum rate maximization mathematical model for jointly optimizing pilot power and data power based on the closed-form expression; Step 5: Use an iterative algorithm to find the local optimum of the model; The detectors in step 3 include, but are not limited to, the maximum ratio merging detector and the zero-forcing detector; When using the aforementioned maximum ratio combined MRC detector, the closed-form expression for the achievable rate lower bound is: ， in, ， ， ， in, This is the lower bound of the achievable rate (LB) in the closed-form case of an MRC detector. , It is the length of user data. In the case of an MRC detector, the first k Effective signal-to-interference-plus-noise ratio (SINR) for each user; It is about A specific function form, It is the probability of decoding error. It is a Gaussian function inverse function, It is about The channel dispersion function; It refers to the number of antennas at the base station. It is the first Normalized transmit power for individual users It is the first Large-scale fading between individual users and base stations Describing the permutation matrix The rank, that is, the first Precoding matrix of each user and the Precoding matrix of each user have The same column, , It is a collection of users. This refers to the number of users with a single antenna. When using the zero-forcing ZF detector, the closed-form expression for the achievable rate lower bound is: ， in, ， ， ， in, The achievable rate lower bound (LB) in closed form under the ZF detector reflects the impact of SINR, the number of transmitted data symbols, the amount of channel usage, and the decoding error probability on the achievable rate. It is the effective signal-to-interference-plus-noise ratio (SINR) of the i-th user under the ZF detector condition; It is about A specific function form, It is about The channel dispersion function; The weighted sum rate maximization mathematical model in step 4 is as follows: in, , , It is the first The weight of each user It is the first A closed-form expression for the reachable rate for each user, when using a maximum ratio merging detector. When a zero-forcing detector is used, , and These are the minimum achievable rate and minimum energy consumption of the optimization model, respectively.

2. The generalized superimposed pilot optimization method for IoT URLLC services according to claim 1, characterized in that, The frame length of the transmission frame in step 1 is ,from Selecting from orthogonal matrices ( The remaining columns serve as the user's pilot sequences. Select from column Constructing the user's precoding matrix And it applies to user data; for the first Each user selects a precoding matrix. To satisfy the pilot frequency orthogonality, that is ,in yes The conjugate transpose of the matrix. Indicates the first Based on the orthogonal pilot vectors of individual users and the relationship between the precoding matrices of different users, the following conclusions are drawn: ， in Denotes the permutation matrix, when When the permutation matrix is ​​an identity matrix and has full rank, It is a unit array.

3. The generalized superimposed pilot optimization method for IoT URLLC services according to claim 1, characterized in that, The channel estimation in step 2 uses the MMSE criterion, and the calculation formula is as follows: in It is a Gaussian signal. It is the first Pilot normalized transmit power for each user, It is the channel usage. It is the first Channel between individual users and base stations It is the solution for MMSE channel estimation. It is noise and follows a complex Gaussian distribution. These are the weighting coefficients of the linear estimate that minimize the mean square error between the channel estimate and the true value. It is a matrix Expectations It is a matrix Expectations yes The conjugate transpose of the matrix. It is a matrix The inverse matrix.

4. The generalized superimposed pilot optimization method for IoT URLLC services according to claim 1, characterized in that, The iterative algorithm in step 5 specifically includes the following steps: Step 51: Transform the mathematical model P1 into a geometric programming (GP) problem; Step 52: Transform the GP problem into a convex form through logarithmic transformation, and then propose an iterative algorithm for solving local optima; Step 53: Initialize the number of iterations of the iterative algorithm. Error probability and feasible power allocation ; Step 54: Execute the iterative algorithm based on the initial value of the iterative algorithm to find the optimal solution to the convex problem.

Images