Joint Optimization Method for Multi-Hop Backhaul Link Selection and Power Allocation in 5G Elastic Coverage System

Abstract

The present invention discloses a joint optimization method for multi-backhaul link selection and power allocation in a 5G elastic coverage system, which solves the problems of backhaul link selection and backhaul power allocation in a multi-backhaul link system. The method classifies user services into three typical 5G services, assumes that the transmission queue of a small base station has three sub-queues corresponding to three backhaul methods, analyzes the transmission delay of data packets on the sub-queues of the small base station through queuing theory, models the optimization objective of maximizing the delay-tolerant elasticity value, and decomposes the optimization problem into a backhaul link selection sub-problem and a backhaul power allocation sub-problem, and finally obtains a joint optimization algorithm for multi-backhaul link and power allocation. This method reduces the average delay of transmitting service data packets by network base stations and effectively improves the transmission rate of the network.

Description

Technical Field

[0001] The present invention relates to the technical field of communication networks, and mainly relates to a joint optimization method for multi-backhaul link selection and power allocation in a 5G elastic coverage system. Background Art

[0002] With the rapid development of 5G mobile communication technology, ubiquitous high-bandwidth networks bring diverse and high-quality services to people. However, a large number of high-traffic services such as video live broadcasts in urban and rural hotspots still pose new challenges to the network, requiring the network to have large-traffic access and backhaul capabilities for on-demand coverage. Since 5G access and backhaul integrated base stations have the characteristics of flexibility, high efficiency, economy, etc., they provide an important solution for large-capacity on-demand coverage requirements.

[0003] In recent years, domestic and foreign scholars have conducted in-depth research on the joint optimization technology of access and backhaul for single backhaul links. Existing research works are all about the joint optimization of resource allocation for the access link of small base stations and a single backhaul link in heterogeneous ultra-dense networks. In many hotspot scenarios, in order to improve the reliability, flexibility, and applicability of communication, base stations need to have the ability of multi-backhaul links. However, existing research lacks in-depth research on the optimization problem of small base stations with multi-backhaul links in application scenarios that meet the requirements of large-capacity multi-services. Summary of the Invention

[0004] Object of the Invention: To solve the problems of backhaul link selection and power allocation of small base stations with multi-backhaul links in application scenarios that meet the requirements of large-capacity multi-services, the present invention studies a 5G elastic network coverage model with multi-backhaul links, introduces a matching degree factor, and under the condition of meeting the differential service requirements of various user services in typical application scenarios, proposes a joint optimization method for multi-backhaul link selection and power allocation in a 5G elastic coverage system. While selecting a suitable backhaul link, the delay tolerance elasticity value is maximized, which can reduce the average transmission delay of network service data packets and meet the differential service requirements of various user services.

[0005] Technical Solution: To achieve the above object, the technical solution adopted by the present invention is as follows:

[0006] A joint optimization method for multi-backhaul link selection and power allocation in a 5G elastic coverage system, the method comprising the following steps:

[0007] Step S1, calculating the average arrival rate of each service data packet at the current moment through a user service model, calculating the total backhaul link rate of three backhaul channels at the current moment through a backhaul channel model, and calculating the average queuing delay of each service data packet when entering the small base station queue through a small base station queuing model;

[0008] Step S2: Calculate the matching degree factor, obtain the effective backhaul rate of the service data packet on different backhaul links, and establish a model for the backhaul link allocation problem with the goal of maximizing the delay-tolerant elasticity value;

[0009] Step S3: Use the backhaul link allocation strategy to solve the backhaul link allocation problem;

[0010] Step S4: Use the backhaul power allocation strategy to solve the backhaul power allocation problem;

[0011] Step S5: Combine the results of Step S3 and Step S4 to obtain the multi-hop backhaul link selection result and power allocation result of the 5G elastic coverage system.

[0012] Furthermore, this method considers a two-tier heterogeneous network composed of 5G macro base stations (MBSs) and small base stations (SBSs); among them, the SBSs use three backhaul methods, namely satellite backhaul, wireless Mesh backhaul, and millimeter-wave (MMW) backhaul, to backhaul user service data to the MBS or the core network; all users are randomly and uniformly distributed within the network coverage area, and a user can only be served by one small base station or one macro base station at the same time;

[0013] The network coverage area includes K small base station users (SUEs) served by SBSs and L macro base station users (MUEs) served by MBSs. The data packets of MUEs are directly transmitted to the MBSs, the data packets of SUEs are backhauled to the MBSs through the SBSs, and the MBSs are connected to the core network through optical fibers.

[0014] Furthermore, the user services are divided into three categories: ultra-reliable and low-latency communication (uRLLC) services, enhanced mobile broadband (eMBB) services, and massive machine type communication (mMTC) services. The establishment method of each type of user service model is as follows:

[0015] (1) Use a two-layer model to model the uRLLC services. The session layer is used to describe the characteristics of the uRLLC service requests initiated by users, where the characteristics of the uRLLC service requests include the service arrival time interval and the session length; the data packet layer is used to describe the characteristics of the data packets included in each uRLLC service, where the characteristics of the data packets include the data packet arrival interval and the data packet size;

[0016] The uRLLC service arrival model in the session layer is an ON / OFF model. Let the durations of the ON / OFF states be T on and T off , then T on and T off both follow an exponential distribution, and their means are t on and t off , and t on >t off , so T onand T off The probability densities are as follows:

[0017]

[0018]

[0019] When the uRLLC service source is in the ON state, data packets are sent at fixed time intervals T p The size of each data packet sent is φ p follows a truncated Pareto distribution, and its probability density function is expressed as follows:

[0020]

[0021] where l sg represents the minimum value of the data packet group size and is the scale parameter of the Pareto distribution; h sg represents the maximum value of the group size; 1 / α sg is the shape parameter;

[0022] The average size φ of the uRLLC service data packets uRLLC is:

[0023]

[0024] The number of uRLLC users in the network coverage area follows a Poisson distribution with a mean of θ uRLLC Then the average packet arrival rate λ1 of the uRLLC service is expressed as:

[0025]

[0026] (2) The FTP3 model is used to model the eMBB service. The FTP3 model defines that the arrival of eMBB data packets follows a Poisson distribution with a mean of D eMBB and the size of the data packets is a fixed value φ eMBB , and the eMBB users are uniformly distributed in the network coverage area with a mean of θ eMBB Thus, the average packet arrival rate of the eMBB service is λ2 = θ eMBB ×D eMBB ;

[0027] (3) The number of mMTC users in the network coverage area follows a uniform distribution with a mean of θ mMTC , and the mMTC users upload data packets of a fixed size φ mMTC per second. The data packet traffic of the mMTC service within one second is randomly distributed. Thus, the average packet arrival rate of the mMTC service is λ3 = θ mMTC .

[0028] Furthermore, an access channel model is established. It is assumed that the access and backhaul of the SBS use different frequencies for transmission, orthogonal frequency division multiplexing is adopted among base station cells, the wireless channel is a Rayleigh channel, and there is no interference among the user UEs in the whole system;

[0029] There are three cases for the uplink access transmission of users: If the UE is within the coverage of the MBS but not within the coverage of the SBS, the user directly accesses the MBS; If the UE is within the coverage of the SBS but the SBS is congested, the user directly accesses the MBS; When the UE is within the coverage of both the MBS and the SBS and the SBS is not congested, the user accesses the core network through the SBS;

[0030] According to the Shannon formula, the theoretical transmission rate of the user's wireless access to the base station is:

[0031]

[0032] where B UE represents the bandwidth evenly allocated by the base station to its underlying users, and γ k (t) represents the signal-to-noise ratio between the UE k and the base station, which can be specifically expressed as:

[0033]

[0034] where P UE represents the transmission power of the UE, h k (t) represents the channel gain from the UE to the base station at time slot t, and it follows an exponential distribution with a mean of l, and N0 is the white noise power spectral density.

[0035] 5. Furthermore, a backhaul channel model is established. The backhaul channel includes a millimeter-wave backhaul channel, a wireless Mesh backhaul channel, and a satellite backhaul channel. The total rate of the backhaul link for each backhaul channel is calculated as follows:

[0036] (1) Define the probability that the system millimeter-wave backhaul link is LOS transmission as p LOS (d):

[0037]

[0038] where α0 is the environmental occlusion factor, and d represents the distance from the receiving end to the transmitting end of the millimeter-wave link;

[0039] Define the probability that the system millimeter-wave backhaul link is NLOS transmission as p NLOS (d), then:

[0040] p NLOS (d) = 1 - p LOS (d)

[0041] Define the transmission power of the SBS as P t , then the received power of the MBS is:

[0042] P r = P t ·G·L(d) -1

[0043] where G represents the antenna gain coefficient; L(d) -1 represents the large-scale channel gain, which is a path loss model including shadow fading, and is denoted by PL db (d) represents its dB form:

[0044]

[0045] where α represents the path loss value for determining the reference distance from the transmitter to the receiver; β represents the path loss exponent; represents the shadow fading loss parameter, which is a zero-mean normal Gaussian random variable with variance ;

[0046] Define the transmission power of the SBS using millimeter wave for backhaul as P t W , after passing through the transmission path, the received power reaching the MBS is:

[0047] P r W = P t W ·G·[L LOS (d) -1 ·p LOS (d)+L NLOS (d) -1 ·(1 - p LOS (d))]

[0048] The signal-to-noise ratio of the millimeter wave link in time slot t is defined as:

[0049]

[0050] where N0 represents the white noise power spectral density, and B W represents the millimeter wave backhaul link allocated bandwidth;

[0051] Furthermore, it can be obtained that the total rate of the SBS millimeter wave backhaul link in time slot t is:

[0052]

[0053] (2) Assume that the satellite-ground link uses the Ka band, and the channel fading of the SBS satellite backhaul link is set to follow large-scale fading and shadow Rice fading. Then the channel correlation coefficient of the satellite link is:

[0054]

[0055] where g s represents a complex Gaussian variable of Rayleigh fading, and β s (d) follows a lognormal distribution, and α s represents the path loss coefficient of the satellite link;

[0056] The signal-to-noise ratio of the SBS satellite backhaul link in the t-th time slot is:

[0057]

[0058] where P t S represents the transmission power of the SBS to the satellite; G S represents the antenna gain; represents the satellite link channel gain; represents the noise variance of additive white Gaussian noise, B S represents the transmission bandwidth of the satellite link;

[0059] It can be seen from the above formula that the total rate of the SBS satellite backhaul link in the t-th time slot is:

[0060]

[0061] (3) The SBS backhauls data to the MBS through the wireless Mesh backhaul link. Assume that there are n Mesh nodes M1, M2,..., M n in the process of this wireless Mesh backhaul link, and the node set M = {M1, M2,..., M n};

[0062] The signal-to-interference-plus-noise ratio state parameter SINR a of the connection link between any two nodes M b and M a,b in the wireless Mesh network in the t-th time slot is:

[0063]

[0064]

[0065] where, represents the transmission power of the SBS at M a ; G a,b represents Ma The antenna gain between M b ; h a,b (t) represents the M in time slot t a The antenna gain between M b ; B M represents the transmission channel bandwidth of the SBS wireless Mesh backhaul link; G x,x+1 represents the node M x The antenna gain between M x and the next adjacent node M x+1 ; |h x,x+1 | represents the M in time slot t x The antenna gain between M x+1 ; γ is the orthogonality factor; γ0 represents the given threshold of the Mesh link signal-to-interference-plus-noise ratio parameter;

[0066] The set SINR of the signal-to-interference-plus-noise ratio status parameters of the connected link between any two nodes in the SBS wireless Mesh link in time slot t M (t) is expressed as:

[0067] SINR M (t) = {SINR 0,1 (t), SINR 1,2 (t),..., SINR n-1,n (t)}

[0068] The total rate of the SBS wireless Mesh backhaul link in time slot t is obtained from the minimum value in SINR M (t):

[0069]

[0070] Furthermore, assume that there are Q W , Q M , Q S Three backhaul queues at the SBS for caching the user data packets received from the SUE, and the queue set Q = {Q W , Q M , Q S}; where Q W , Q M , Q S respectively represent that the data packets queuing in the queues are backhauled to the MBS through millimeter wave, wireless Mesh network, and satellite methods;

[0071] ​Suppose a packet group consisting of N' packets of different service types arrives at the SBS. The SBS distributes the packets back to the sub-queues and transmits them sequentially in the sub-queues. At the same time, assume that the packets of different services are independent of each other, and the arrival process of a batch of packets follows a Poisson distribution;

[0072] According to the set queue rules, the small cell queuing model of queue Q belongs to a birth-death process, where represents that the successive arrival times of packets follow a negative exponential distribution with parameter λ, D represents the service time, N represents the maximum length of the queue, and ∞ represents an infinite packet source;

[0073] Use p i,n to represent the probability that the length of queue Q i is n, n ∈ {0, 1,..., N}. According to the analysis method of queuing theory, list the steady-state equations of the probabilities of each state of the queue for the queue state:

[0074]

[0075] where i ∈ {W, M, S}, 0 ≤ n ≤ N; for queue Q i the utilization rate ρ i is the ratio of the arrival rate λ i to the service rate μ i ;

[0076] From the above equation, we get:

[0077]

[0078] where p i,0 is the idle probability of queue Q i ;

[0079] According to the regularity of the queue state, we have:

[0080]

[0081] Furthermore, the idle probability p i of queue Q i,0 is obtained as:

[0082]

[0083] When the queue length is greater than N, it is considered that the queue is blocked. Therefore, the blocking rate of queue Q i is expressed as B 1,i = p i,N , and the non-blocking rate of queue Q i is B 0,i = 1 - B 1,i = 1 - pi,N ;

[0084] Queue Q i The average queue length of is obtained by adding two parts. One part is the queuing length L increased during the process of data packets queuing and waiting in the queue i,w , and the other part is the waiting length L caused during the transmission of data packets in the SBS sub-queue i,s . Then, the average queue length after the queue reaches the equilibrium state is:

[0085]

[0086] The effective arrival rate of data packets is the average number of users who can enter the system queue per unit time:

[0087] λ i,e = λ i (1 - p i,N ) = μ i (1 - p i,0 )

[0088] According to Little's formula, the average queuing delay of data packets is:

[0089]

[0090] Then, the average queuing delay of data packets is:

[0091]

[0092] Furthermore, define the set as the set of data packets to be transmitted in time slot t, where a m,x (t) represents the m-th data packet of service type x, and the effective backhaul rate of a m,x (t) on the backhaul link i is:

[0093]

[0094] Among them, represents the normalized matching degree factor, MF i,x represents the matching degree factor of a m,x (t) to queue i in time slot t, represents the k-th relevant backhaul network parameter of a m,x (t) in queue i, is 's weight, represents the total rate of the SBS backhaul link i in time slot t, a m,x(t) represents the m - th data packet of service type x in the set of data packets to be transmitted, where x ∈ {uRLLC, eMBB, mMTC} and m ∈ {1, 2,..., M}.

[0095] Furthermore, a backhaul link allocation problem with the goal of maximizing the delay - tolerant elasticity value is established:

[0096]

[0097]

[0098]

[0099]

[0100]

[0101]

[0102]

[0103]

[0104] Among them, is the delay - tolerant elasticity value, T(a m,x (t)) represents the delay threshold of a m,x (t), α is the sensitivity factor, χ x represents the service gain of a m,x (t), is the average size of the data packet, φ MMV 、φ mesh 、φ S represent the data packet sizes of the millimeter - wave backhaul link, the mesh backhaul link, and the satellite backhaul link respectively; is the value of the m - th row and i - th column of the backhaul link selection matrix , represents that a m,x (t) is assigned to the backhaul link i, represents that a m,x (t) is not assigned to the backhaul link i; d i represents the maximum tolerable delay for the backhaul link i to transmit a data packet; D max 、D min represent the maximum and minimum average queuing delays of the data packet in the queue respectively, P m,i represents the transmission power of a m,x (t) on the backhaul link i, P maxDenote the maximum value of the total instantaneous transmission power of SBS, φ(a m,x (t)) represents the magnitude of a m,x (t), denote the average queuing delay of a m,x (t) accessing SBS,

[0105] denote the average queuing delay of a m,x (t) in queue Q W ; denote the average queuing delay of a m,x (t) in queue Q M ; denote the average queuing delay of a m,x (t) in queue Q S ; δ i ∈{0, 1} is the congestion flag corresponding to queue Q i , δ i =0 indicates that the queue is congested, and δ i =1 indicates that the queue is not congested; τ i is the set threshold;

[0106] The above backhaul link allocation problem is decomposed into two sub - problems: backhaul link selection and backhaul power allocation by using the two - step method. Solve the two sub - problems separately to obtain the backhaul link selection matrix and P m,i .

[0107] Furthermore, the backhaul link allocation problem is simplified to a backhaul link selection sub - problem:

[0108]

[0109]

[0110]

[0111]

[0112]

[0113] The solution process of the backhaul link selection sub - problem is as follows:

[0114] Step 1, initialize D min , D max , d uRLLC , d eMBB , d mMTC , λ, λ = λ1 + λ2 + λ3;

[0115] Step 2, initialize the packet transmission power:

[0116] Step 3, initialize the range of the allowable length of each queue and initialize λ W , λ M , λ S is 1 / 3 of the total arrival rate λ;

[0117] Step 4, calculate the delay tolerance elasticity value of each data packet and queue and fill it into the delay tolerance elasticity matrix; among them, the element value of the m-th row and the i-th column of the delay tolerance elasticity matrix is

[0118] Step 5, transform the delay tolerance elasticity matrix into a square matrix to simplify the backhaul link selection sub-problem into a one-to-one assignment problem;

[0119] Step 6, take maximizing the delay elasticity tolerance value as the goal, execute the Hungarian algorithm to solve the one-to-one assignment problem, obtain the backhaul link selection matrix, and update λ W , λ M , λ S ;

[0120] Step 7, calculate the average queuing delay of each queue according to the backhaul link selection matrix;

[0121] Step 8, if D i ≤d i is satisfied, then execute Step 10, otherwise execute Step 9;

[0122] Step 9, reassign the data packet with the smallest delay tolerance elasticity value in the queue that does not satisfy D i ≤d i to other queues that satisfy D i ≤d i to obtain a new backhaul link selection matrix and jump to Step 4;

[0123] Step 10, output the latest backhaul link allocation matrix and the maximum delay elasticity tolerance value.

[0124] Furthermore, simplify the backhaul link allocation problem into a backhaul power allocation sub-problem:

[0125]

[0126]

[0127]

[0128]

[0129] The Lagrange dual method is used to transform the backhaul power allocation sub-problem into a concave problem, construct the Lagrangian function to handle the non-linear constraint conditions, and use the gradient descent method to solve the optimal backhaul power allocation value that satisfies the KKT conditions.

[0130] Beneficial effects:

[0131] The present invention proposes a joint optimization method for multi-backhaul link selection and power allocation in a 5G elastic coverage system. This method assumes that the small cell transmission queue has three sub-queues corresponding to three backhaul modes. The transmission delay of data packets on the small cell sub-queues is analyzed through queuing theory. A model is established with the optimization goal of maximizing the delay tolerance elasticity value, and the optimization problem is decomposed into a backhaul link selection sub-problem and a backhaul power allocation sub-problem. Finally, a joint optimization algorithm for multi-backhaul links and power allocation is obtained. Compared with traditional algorithms, the present invention can reduce the average delay of service data packets and effectively improve the transmission rate of the network. Description of the drawings

[0132] Figure 1 is a multi-backhaul access and backhaul integrated network scenario;

[0133] Figure 2 is a uRLLC service model;

[0134] Figure 3 is an M / D / 1 / ∞ queuing model. Detailed implementation manners

[0135] In order to make the objectives, technical solutions and advantages of the present invention clearer, the following further details the present invention in conjunction with the accompanying drawings and examples. It should be understood that these examples are only used to illustrate the present invention and not to limit the scope of the present invention. After reading the present invention, various equivalent forms of modification by those skilled in the art fall within the scope defined by the appended claims of this application.

[0136] Based on Figure 1 the multi-backhaul access and backhaul integrated network scenario shown, the present invention proposes a joint optimization method for multi-backhaul links and power allocation in a 5G elastic coverage system, and the method includes the following steps:

[0137] Step S1: Calculate the average arrival rate of each service data packet at the current moment through the user service model, calculate the total backhaul link rate of the three backhaul channels at the current moment through the backhaul channel model, and calculate the average queuing delay when the data packet enters the small cell queue through the small cell queuing model;

[0138] Step S2: Calculate the matching degree factor, obtain the effective backhaul rate of the data packet on different backhaul links, and establish a model of the backhaul link allocation problem with the goal of maximizing the delay tolerance elasticity value;

[0139] Step S3: Use the backhaul link allocation strategy to solve the backhaul link allocation problem;

[0140] Step S4: Use the backhaul power allocation strategy to solve the backhaul power allocation problem;

[0141] Step S5: Combine the results of Step S3 and Step S4 to obtain the multi-backhaul link selection result and power allocation result of the 5G elastic coverage system.

[0142] To illustrate the effectiveness of the method proposed in the present invention, the present invention gives an example. The example scenario is Figure 1 Take the multi-backhaul access and backhaul integrated network scenario shown as an example. Consider a two-tier heterogeneous network composed of 5G macro-cell base stations (MBS) and small-cell base stations (SBS). The MBS supports a wide range of mobile user coverage, with a coverage range of about hundreds of meters to thousands of meters. In the scenario, SBSs are deployed to meet the explosive network traffic in some areas within the coverage range of the macro base station. The coverage range of the SBS is about dozens of meters to hundreds of meters. Assume that the SBSs use three backhaul methods, namely satellite backhaul, wireless Mesh backhaul, and millimeter wave (MMW) backhaul, to transmit user service data back to the MBS or the core network.

[0143] Assume that the area includes K small-cell users (SUE) served by small-cell base stations and L macro-cell users (MUE) served by macro-cell base stations. The data packets of the MUEs are directly transmitted to the MBS, and the SUEs transmit the data packets back to the MBS through the SBS. The MBS is connected to the core network through optical fiber. Assume that all single-antenna users are randomly and uniformly distributed within the coverage range of the network, and a user can only be served by one small-cell base station or macro-cell base station at the same time.

[0144] Assume that the data transmission system of the SBS in the scenario is a time-division multiplexing system, and consider the transmission of uplink user services in discrete time slots τ = {1, 2,..., T}. Multiple data packets arrive at the SBS successively within the t time slot. Since the t time slot is very short, the data packets arriving within this time slot are considered to arrive in the same batch, and the wireless channel within one time slot remains relatively stable. Within the t time slot, assume that M data packets arrive at the SBS and queue up for transmission. Define the set as the set of data packets to be transmitted in the t time slot, where a m,x (t) represents the m-th data packet of service type x.

[0145] In one embodiment, user services are divided into three major categories, namely ultra-reliable and low-latency communication (uRLLC) services, enhanced mobile broadband (eMBB) services, and massive machine type communication (mMTC) services. By establishing a user service model for each category, the average arrival rate of each service data packet at the current moment is calculated.

[0146] The present invention uses a two-layer model to model uRLLC services. The two-layer model is as Figure 2 shown. The session layer is used to describe the characteristics of user-initiated uRLLC service requests, where T s represents the service arrival time interval, and T l represents the session length; the packet layer mainly describes the characteristics of the data packets included in each uRLLC service, where T p represents the packet arrival interval, and φ p represents the packet size.

[0147] The uRLLC service arrival model in the session layer is an ON / OFF model. Let the durations of the ON / OFF states be T on and T off , then both T on and T off follow an exponential distribution, and their means are t on and t off , respectively, and t on > t off . Therefore, the probability densities of T on and T off are respectively:

[0148]

[0149]

[0150] When the uRLLC service source is in the ON state, data packets are sent at a fixed time interval T p , and the size φ p of each sent data packet follows a truncated Pareto distribution, and its probability density function is expressed as follows:

[0151]

[0152] where l sg is called the scale parameter of the Pareto distribution, representing the minimum value of the data packet group size; h sg represents the maximum value of the group size; 1 / α sg is called the shape parameter.

[0153] The average size φ uRLLC of uRLLC service data packets is:

[0154]

[0155] Assume that the number of uRLLC users in the cell follows a Poisson distribution with a mean of θ uRLLC , then the average packet arrival rate λ1 of the uRLLC service is expressed as:

[0156]

[0157] In the present invention, the eMBB service in the scenario is modeled using the File Transfer Protocol (FTP) Model 3 recommended by 3GPP [3GPP13-36872] (referred to as the FTP3 model). The FTP3 model defines that the arrival of eMBB data packets follows a Poisson distribution with a mean of D, and the size of the data packets is a fixed value φ eMBB , and the eMBB users are uniformly distributed in the cell with a mean of θ eMBB . From the above assumptions, the average packet arrival rate of the eMBB service is λ2 = θ eMBB ×D.

[0158] The number of mMTC users in the base station cell follows a uniform distribution, assume the mean is θ mMTC , and the mMTC users upload data packets of a fixed size φ mMTC per second. The data packet traffic of the mMTC service within one second is randomly distributed. Therefore, the average packet arrival rate of the mMTC service is λ3 = θ mMTC .

[0159] Combining the data packet sizes of the three service types, the average size of the data packets sent by users in this scenario and the total arrival rate λ of the user data packets are:

[0160]

[0161] λ = λ1 + λ2 + λ3 (7) where φ MMV , φ mesh , φ S represent the data packet sizes of the millimeter-wave backhaul link, the mesh backhaul link, and the satellite backhaul link respectively; λ is the sum of the average arrival rates of the data packets of the three services.

[0162] In one embodiment, an access channel model is established. It is assumed that the access and backhaul of the SBS use different frequency transmissions, orthogonal frequency division multiplexing is adopted among base station cells, the wireless channel is a Rayleigh channel, and there is no interference among the user UEs (including MUE and SUE) in the entire system. There are three cases for the uplink access transmission of users: If the UE is within the coverage of the macro base station but not within the coverage of the small base station, the user directly accesses the MBS; If the UE is within the coverage of the small base station but the SBS is congested, the user directly accesses the MBS; When the UE is within the coverage of both the macro base station and the small base station and the SBS is not congested, the user accesses the core network through the SBS.

[0163] According to the Shannon formula, the theoretical transmission rate of the user's wireless access to the base station is:

[0164]

[0165] where B UE represents the bandwidth evenly allocated by the base station to its subordinate users, and γ k (t) represents the signal-to-noise ratio between UE k and the base station, which can be specifically expressed as:

[0166]

[0167] where P UE represents the transmission power of the UE, h k (t) represents the channel gain from the UE to the base station at time slot t, and follows an exponential distribution with a mean of l. N0 is the white noise power spectral density.

[0168] In one embodiment, a backhaul channel model is established. The backhaul channel includes a millimeter-wave backhaul channel, a wireless Mesh backhaul channel, and a satellite backhaul channel.

[0169] In one embodiment, it is defined that the probability of the system millimeter-wave backhaul link being LOS transmission is p LOS , and according to the simplified sphere model:

[0170]

[0171] where α0 is the environmental occlusion factor, which is set according to the occlusion degree of the transmission environment; d represents the distance from the receiving end to the transmitting end of the millimeter-wave link.

[0172] Let the probability of the system millimeter-wave backhaul link being NLOS transmission be p NLOS , then:

[0173] p NLOS (d) = 1 - p LOS (d) (11)

[0174] Let the transmission power of the SBS be P t , then the received power of the MBS is:

[0175] P r = P t ·G·L(d) -1 (12)

[0176] where G represents the antenna gain coefficient; L(d) -1 represents the large-scale channel gain, which is a path loss model including shadow fading, and is denoted by PL db (d) represents its dB form:

[0177]

[0178] where α represents the path loss value for determining the reference distance from the transmitter to the receiver; β represents the path loss exponent; represents the shadow fading loss parameter, which is a zero-mean normal Gaussian random variable with variance and is in dB; d represents the link distance between the SBS and the MBS, in meters. The α, β, and shadow fading loss exponents for LOS and NLOS links are different.

[0179] Assume that the transmission power of the SBS using millimeter wave for backhaul is P t W , after passing through the transmission path, the received power reaching the MBS is:

[0180] P r W = P t W ·G·[L LOS (d) -1 ·p LOS (d)+L NLOS (d) -1 .(1 - p LOS (d))] (14)

[0181] In millimeter wave transmission, beamforming technology makes the transmitted signal highly directional, and the signal is mainly interfered by white noise. The signal-to-noise ratio of the millimeter wave link in the t time slot is defined as:

[0182]

[0183] where N0 represents the white noise power spectral density, and B W represents the allocated bandwidth of the millimeter wave backhaul link.

[0184] Based on the above formula and the Shannon formula, the total rate of the small cell millimeter wave backhaul link in the t time slot is:

[0185]

[0186] In one embodiment, assuming that the satellite-ground link uses the Ka band, the channel fading of the SBS satellite backhaul link is set to follow large-scale fading and shadow Rice fading. Therefore, the channel correlation coefficient of the satellite link is:

[0187]

[0188] where g s represents a complex Gaussian variable of Rayleigh fading, β s follows a log-normal distribution, and α s represents the path loss coefficient of the satellite link. Therefore, the signal-to-noise ratio of the SBS satellite backhaul link at time slot t is:

[0189]

[0190] where P t S represents the transmission power from the small base station to the satellite. Obviously, P t S is less than the maximum transmission power of the small base station; G S represents the antenna gain; represents the satellite link channel gain; represents the noise variance of additive white Gaussian noise, and the calculation formula is where B S represents the transmission bandwidth of the satellite link.

[0191] From the above formula, it can be seen that at time slot t, the total rate of the SBS satellite backhaul link is:

[0192]

[0193] In one embodiment, the SBS backhauls data to the MBS through the wireless Mesh backhaul link. Assuming that the wireless Mesh backhaul link passes through n Mesh nodes, the set M = {M1, M2,..., M n} represents all the Mesh nodes in the link. The SBS can obtain the signal-to-interference-plus-noise ratio state parameter SINR a of the connected link between any two nodes M b and M a,b (t) in the wireless Mesh network at time t, M a , M b ∈M. In particular, M0 represents the MBS node.

[0194]

[0195]

[0196] Among them, represents the transmission power of SBS at the wireless Mesh node a; G a,b represents the antenna gain; h a,b (t) represents the channel correlation coefficient between nodes a and b in time slot t; N0 represents the power spectral density of Gaussian white noise; B M represents the transmission channel bandwidth of the SBS wireless Mesh backhaul link; γ (0 ≤ γ ≤ 1) is the orthogonality factor, indicating the degree to which the signal is interfered by other adjacent Mesh nodes; b + 1 represents the next adjacent Mesh node of node b on the link, and similarly, b - 1 represents the previous adjacent Mesh node of node b on the link; the expression in the denominator indicates that only the one-hop adjacent Mesh nodes of the receiving node b are considered when calculating the link signal-to-noise ratio; γ0 represents the given threshold of the Mesh link signal-to-interference-plus-noise ratio parameter.

[0197] In time slot t, for all pairs of nodes M a and M b in the SBS wireless Mesh link, the signal-to-interference-plus-noise ratio status parameter SINR M (t) can be expressed as:

[0198] SINR M (t) = {SINR 0,1 (t), SINR 1,2 (t),..., SINR n-1,n (t)} (21)

[0199] Based on the obtained signal-to-interference-plus-noise ratio SINR M (t) of the entire wireless Mesh link, SBS calculates the minimum From this, the maximum transmission rate of the small cell SBS wireless Mesh backhaul link in time slot t is obtained:

[0200]

[0201] In one embodiment, a queue model is established. It is assumed that there are Q W , Q M , Q S three backhaul queues at the SBS for buffering the user data packets received from the SUEs. The queue set Q = {Q W , Q M , Q S}. The SBS selects a suitable queue in Q to queue and wait for backhaul for the user data packets received from the SUEs. Q W , QM , Q S respectively represent that the data packets queuing in the queue are backhauled to the MBS via millimeter wave, wireless Mesh network, and satellite methods respectively, and the maximum length of each queue is N.

[0202] Suppose a packet group consisting of N' data packets of different service types arrives at the SBS. The SBS distributes the data packets to the backhaul sub-queues and transmits them sequentially in the queue. At the same time, assume that the data packets of different services are independent of each other, and the arrival process of a batch of data packets follows a Poisson distribution.

[0203] According to the set queue rules, the queuing model of the queue in Q belongs to a birth-death process, as Figure 3 shown, where represents that the successive arrival times of data packets follow a negative exponential distribution with parameter λ, D represents the service time, N represents the maximum length of the queue, and ∞ represents an infinite data packet source.

[0204] Use p i,n , n ∈ {0, 1,..., N} to represent the probability that the queue length of queue Q i is n. For the convenience of expression, define the utilization rate of this queue as the ratio of the arrival rate to the service rate, denoted as According to the analysis method of queuing theory, the steady-state equations of the probabilities of each state of the system can be listed for the queue state:

[0205]

[0206] It can be seen from equation (24) that when n = 1, 2,..., N, the probability that the queue length is n is

[0207]

[0208] where i ∈ {W, M, S}, 0 ≤ n ≤ N. According to the regularity of the queue state, there is:

[0209]

[0210] At this time, according to the sum of the probabilities of the queue state, there should be The idle probability p i of queue Q i,0 can be obtained as

[0211]

[0212] When the queue length is greater than the maximum length N, it is considered that the queue is blocked. Therefore, the blocking rate of queue Q i is expressed as B 1,i = p i,N, correspondingly, queue Q i The non-blocking rate is B 0,i =1-B 1,i =1-p i,N .

[0213] Queue Q i The average queue length is obtained by adding two parts. One is the queue length L that increases during the waiting process of the data packet in the queue. i,w , and secondly, the waiting queue length L caused by the data packet being transmitted in the SBS queue i,s . The average length of the system after reaching equilibrium can be calculated:

[0214]

[0215] The average number of users that can enter the system queue per unit time, that is, the effective arrival rate of data packets is:

[0216] λ i,e =λ i (1-p i,N )=μ i (1-p i,0 ) (28)

[0217] According to Little's formula, we can enter Q i The average stay time of a single user in D i , that is, the average queuing delay of the data packet is:

[0218]

[0219] Substituting the average length of the system after it reaches equilibrium into equation (30), we can obtain i The average queuing delay equation is:

[0220]

[0221] Among them, for each queue, the queue service rate {μ W ,μ M ,μ S} are all fixed values.

[0222] Assume that in time slot t, the arrival rates of the three queues are λ W ,λ M ,λ S , the arrival rate satisfies λ=λ W +λ M +λ S .

[0223] The backhaul rate of the backhaul link is used to represent the service rate of the corresponding subqueue. According to the transmission rates of the three backhaul methods at the SBS, it can be seen that in time slot t, QW (t) The feedback rate of the queue is Q M (t) The feedback rate of the queue is Q S (t) The feedback rate of the queue is Then the service rate of the SBS queue system can be expressed as:

[0224]

[0225] Then the total service rate of the SBS queue Q in time slot t is

[0226] μ = μ M + μ M + μ S (32)

[0227] Substituting the arrival rate and service rate of the queue into formula (31), the average queuing delay of the data packet in each sub-queue at time t can be obtained. Further, the average queuing delay of data packet a m,x (t) accessing the SBS is

[0228]

[0229] where, δ i ∈ {0, 1}, i ∈ {W, M, S} is the congestion flag of the corresponding feedback queue. When δ i = 0, it means the queue is congested. When δ i = 1, it means the queue is not congested. denotes m,x (t)'s queuing delay in queue Q W (t)'s queuing delay in queue Q denotes m,x (t)'s queuing delay in queue Q M (t)'s queuing delay in queue Q denotes m,x (t)'s queuing delay in queue Q S (obtained from Equation 30).

[0230] In one embodiment, the calculation process of the matching degree factor (MF) can be divided into the following 5 steps:

[0231] (1) Numericalize the feedback network parameters: For the feedback network parameters that already have numerical values, such as feedback bandwidth, jitter, etc., there is no need to numericalize them. For the feedback network parameters that represent a relative degree, such as the quality of the decoder, the quality of the link security, according to the rule that the larger the value, the better the network parameter, convert them into certain numerical values.

[0232] (2) Define the upper and lower bounds of the feedback parameters and standardize the feedback parameters. Obtain Denote the k-th relevant backhaul network parameter of user data packets with service type x in queue i.

[0233] (3) Use the grey relational analysis method to calculate the grey relational coefficients and standardize them, and then obtain the matching degrees of data packets of three service types to three backhaul sub-queues respectively.

[0234] (4) Weight the backhaul network parameters: By weighting each network parameter to represent the relative importance degree of this network parameter for the transmission of data packets in the backhaul link, define the weight of the k-th relevant backhaul network parameter as Then there is The unnormalized value of the matching degree factor MF after weighting is:

[0235]

[0236] (5) Normalize the matching degree factor: Normalize the MF value in (4) to obtain the matching degree factor as:

[0237]

[0238] Among them, max{MF i,x} represents the maximum value of the matching degree factors of all types of data packets to all queues i in the SBS transmission system at time t.

[0239] Weight MF to the maximum transmission rate R B (t) of the SBS backhaul link to obtain the effective backhaul rate of data packet a m,x (t) to backhaul link i as:

[0240]

[0241] In one embodiment, use the Z-type utility function to show the relationship between the data packet transmission delay and the delay threshold, and call this Z-type function the delay tolerance elasticity value, and the expression is as follows:

[0242]

[0243] Among them, T(a m,x (t)) represents the delay threshold of data packet a m,x (t), represents the transmission delay of data packet a m,x (t) during the backhaul process, α is the sensitivity factor, χ x represents the service gain of data packet a m,x (t), is the average size of the data packet.

[0244] In one embodiment, a backhaul link allocation problem with the goal of maximizing the delay-tolerant elasticity value is established, and the model is as follows:

[0245]

[0246]

[0247]

[0248]

[0249]

[0250]

[0251]

[0252]

[0253] Among them, is the value of the m-th row and the i-th column of the backhaul link selection matrix , represents whether a m,x (t) is allocated to the backhaul link i, represents that a m,x (t) is allocated to the backhaul link i, represents that a m,x (t) is not allocated to the backhaul link i; D max , D min respectively represent the maximum and minimum average queuing delays of the data packet in the queue, P m,i represents the transmission power of a m,x (t) on the backhaul link i, P max represents the maximum value of the total instantaneous transmission power of the SBS, represents the average backhaul rate of a m,x (t), τ iis the set threshold. Constraints (39a) and (39b) ensure that each data packet can only be assigned to one backhaul queue within a time slot, and allow the transmission system to actively discard the data packet in the event of congestion in the backhaul queue to improve the overall backhaul network quality. Constraint (39c) limits the maximum delay of the sub-queue. Constraint (39d) restricts the number of user packets that the SBS assigns to each sub-queue, preventing most data packets from always being assigned to sub-queues with good backhaul quality while only a small fraction are assigned to sub-queues with poor backhaul quality, resulting in the situation where good sub-queues are frequently congested and poor sub-queues have no packets to transmit. Constraint (39e) indicates that the sum of the transmission powers of all data packets on the SBS sub-link does not exceed the maximum value of the total instantaneous transmission power. Constraint (39f) means that the transmission power of the data packet is a non-negative value. Constraint (39g) limits the deviation of the backhaul rate of data packets of the same service type from the average effective backhaul rate of this type within 2τ i range, aiming to limit the difference in backhaul rates between data packets of the same service type.

[0254] In one embodiment, the classical two-step method is adopted to decompose the optimization problem into two sub-problems: backhaul link selection and backhaul power allocation, and solve them sequentially and P m,i .

[0255] In one embodiment, according to different services, the transmission power of the data packet is initialized as:

[0256]

[0257] where χ x represents the service gain of the data packet. It can be seen that the initial power satisfies P m,i > 0 and Therefore, it obviously satisfies constraint (39d) and constraint (39e) in the optimization problem (38). The original optimization problem is simplified to the backhaul link selection sub-problem as:

[0258]

[0259]

[0260]

[0261]

[0262]

[0263] After determining the initial power P m,iAfter that, the sub-problem (41) and the constraints do not contain non-linear variables. Therefore, the original NP-hard problem is transformed into an integer linear programming problem. The sub-problem (41) is simplified and expressed as a 0-1 assignment problem. In the 0-1 assignment problem, the time for employees to handle tasks can be represented by an efficiency matrix. In this problem, the efficiency of the SBS queue in processing data packets will be represented by a delay-tolerant elasticity matrix of size M×3. Generally, the number of data packets is much larger than the number of SBS queues (M >> 3). Obviously, for this general case, the fronthaul link allocation problem belongs to a one-to-many assignment problem, simply referred to as a 1-n problem. Next, the fronthaul link selection problem belonging to the 1-n problem will be solved.

[0264] The constraint condition (42c) of this problem belongs to continuous variables. If we want to solve this optimization problem, it should be discretized into the range of queue lengths allowed for sub-queues. The range of the number of data packets in each sub-queue should be:

[0265]

[0266] where, represents rounding n down, represents rounding n up.

[0267] First, since the matrix of the 1-n assignment problem is required to be a square matrix, the delay-tolerant elasticity matrix is first expanded into a square matrix. The delay-tolerant elasticity matrix is expanded into a square matrix according to the following steps: First, the sub-queues of the original delay-tolerant elasticity matrix are copied and expanded into the columns of the delay-tolerant elasticity matrix until the range of the sub-queue size given by equation (43) is reached. The delay-tolerant elasticity values corresponding to each row and column of the copied and filled part remain unchanged, and a new delay-tolerant elasticity matrix is obtained; then, if the new matrix is still not a square matrix, the rows or columns of the matrix are filled with zeros until it becomes a square matrix; finally, the original delay-tolerant elasticity matrix is transformed into a square matrix, and the original 1-n assignment problem is reduced to a one-to-one assignment problem.

[0268] The fronthaul link selection strategy is used to solve the fronthaul link selection problem, and the specific process is shown in Algorithm 1.

[0269]

[0270] In one embodiment, the fronthaul link allocation problem is simplified to a fronthaul link selection sub-problem, described as:

[0271]

[0272]

[0273]

[0274]

[0275] The sub - problem is transformed into a concave problem by using the Lagrange dual method. A Lagrangian function is constructed to handle the non - linear constraint conditions. The Lagrangian function is defined as:

[0276]

[0277] where κ, ψ, and π are the Lagrange multiplier vectors corresponding to the constraint conditions of the sub - problem respectively.

[0278] The Lagrange dual objective function can be expressed as:

[0279]

[0280] The conjugate function is used to represent the dual problem, and the conjugate function can be expressed as:

[0281]

[0282] where

[0283]

[0284]

[0285] The specific form of the Lagrange dual function can be obtained from the Lagrange dual objective function and the conjugate function as follows:

[0286]

[0287] Observing the above formula, it can be seen that f * (y) is related to P m,n , while the other terms in the formula are not related to P m,n . Therefore, taking the derivative of f * (y) and setting it equal to 0 can obtain the optimal solution. The formula is expressed as follows:

[0288]

[0289] where represents 's first - order derivative. According to the above - mentioned content, for each link's can be uniformly expressed as:

[0290]

[0291] where Taking the derivative of the above formula can obtain

[0292]

[0293] Using the KKT conditions, the solution that satisfies the following conditions is the optimal solution of the Lagrangian dual problem:

[0294]

[0295] P m,y ≥0κ m ≥0ψ m ≥0π≥0;

[0296]

[0297] In this section, the gradient descent method is used to solve for the optimal backhaul power allocation value P that satisfies the KKT conditions m,i , and the specific process is as follows: First, the gradient projection method is used to obtain the backhaul power allocation P m,i , and then, based on the obtained P m,i the parameters are updated. If the difference between the current power value and the power value of the previous iteration is less than the iteration accuracy Δ or the number of iterations K is exceeded, the iteration stops and the final solution P m,i is obtained and the power allocation result P m,i is output. Otherwise, the iteration continues.

[0298] The power value iteration formula during the iteration process is as follows:

[0299]

[0300] where, [x] + indicates x≥0, γ represents the step size, the superscript n represents the current iteration process, n - 1 represents the previous iteration process, and n = 1, 2,..., K.

[0301] After each iteration to obtain the power P m,i , the parameters κ m , ψ m , π are updated, and the parameter update formulas are as shown below

[0302]

[0303]

[0304]

[0305] This process continues to iterate and update until the accuracy judgment formula is satisfied and the iteration stops, and the result is output.

[0306] The present invention adopts a backhaul power allocation strategy to solve the backhaul power allocation problem, and the specific process is as shown in Algorithm 2.

[0307]

[0308]

[0309] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A joint optimization method for multi-backhaul link selection and power allocation in a 5G elastic coverage system, characterized in that, The method includes the following steps: Step S1: Calculate the average arrival rate of each service data packet at the current moment through the user service model, calculate the total backhaul link rate of the three backhaul channels at the current moment through the backhaul channel model, and calculate the average queuing delay of each service data packet when it enters the small cell queue through the small cell queuing model; Step S2: Calculate the matching degree factor, obtain the effective backhaul rate of the service data packet on different backhaul links, and establish a model for the backhaul link allocation problem with the goal of maximizing the delay tolerance elasticity value; Step S3: Use the backhaul link allocation strategy to solve the backhaul link allocation problem; Step S4: Use the backhaul power allocation strategy to solve the backhaul power allocation problem; Step S5: Combine the results of Step S3 and Step S4 to obtain the multi-hop link selection result and power allocation result of the 5G elastic coverage system.

2. The method according to claim 1, wherein This method considers a two-tier heterogeneous network composed of a 5G macro base station MBS and small base stations SBS; among them, the SBS uses three backhaul methods, namely satellite backhaul, wireless Mesh backhaul, and millimeter wave MMW backhaul, to backhaul the user service data to the MBS or the core network; all users are randomly and evenly distributed within the network coverage area, and a user can only be served by one small base station or macro base station at the same time; The network coverage area includes K small cell users SUE served by the SBS, and L macro cell users MUE served by the macro base station MBS. The data packets of the MUE are directly transmitted to the MBS, the data packets of the SUE are backhauled to the MBS through the SBS, and the MBS is connected to the core network through optical fiber.

3. The method according to claim 2, wherein The user services are divided into three categories: ultra-reliable and low-latency communication uRLLC services, enhanced bandwidth eMBB services, and massive machine-type communication mMTC services. The establishment method of each type of user service model is as follows: (1) Use a two-layer model to model the uRLLC service. The session layer is used to describe the characteristics of the uRLLC service request initiated by the user, where the characteristics of the uRLLC service request include the service arrival time interval and the session length; the data packet layer is used to describe the characteristics of the data packets included in each uRLLC service, where the data packet characteristics include the data packet arrival interval and the data packet size; The arrival model of uRLLC services at the session layer is the ON / OFF model. Let the duration of the ON / OFF state be T on and T off , then T on and T off both follow an exponential distribution, and their means are t on and t off , and t on > t off , so the probability densities of T on and T off are respectively: When the uRLLC service source is in the ON state, it sends data packets at a fixed time interval T p The size of each data packet sent is φ p obeys a truncated Pareto distribution, and its probability density function is expressed as follows: Among them, l sg represents the minimum value of the data packet group size, which is the proportionality parameter of the Pareto distribution; h sg represents the maximum value of the group size; 1 / α sg is the shape parameter; The average size φ of uRLLC service data packets uRLLC is as follows: The number of uRLLC users within the network coverage area follows a Poisson distribution with a mean of θ uRLLC , then the average packet arrival rate λ1 of the uRLLC service is expressed as: (2) The FTP3 model is used to model the eMBB service. The FTP3 model defines that the arrival of eMBB data packets follows a Poisson distribution with a mean of D eMBB , and the size of the data packets is a fixed value φ eMBB . The eMBB users are uniformly distributed in the network coverage area with a mean size of θ eMBB . Then, the average packet arrival rate of the eMBB service is obtained as λ2 = θ eMBB ×D eMBB ; (3) The number of mMTC users within the network coverage area follows a uniform distribution with a mean of θ mMTC , and mMTC users upload data packets of a fixed size φ mMTC per second. The packet traffic of mMTC services is randomly distributed within one second. Consequently, the average packet arrival rate of mMTC services is obtained as λ3 = θ mMTC .

4. The method according to claim 2, wherein Establish an access channel model. Assume that the access and backhaul of the SBS use different frequencies for transmission, orthogonal frequency division multiplexing is used between base station cells, the wireless channel is a Rayleigh channel, and there is no interference between the user UEs in the entire system; There are three cases for the uplink access transmission of the user: If the UE is within the coverage area of the MBS but not within the coverage area of the SBS, the user directly accesses the MBS; if the UE is within the coverage area of the SBS but the SBS is congested, the user directly accesses the MBS; when the UE is within the coverage area of both the MBS and the SBS and the SBS is not congested, the user accesses the core network through the SBS; According to the Shannon formula, the theoretical transmission rate of the user wirelessly accessing the base station is: Among them, B UE represents the bandwidth evenly allocated by the base station to the users under it, and γ k (t) represents the signal-to-noise ratio between UE k and the base station, which can be specifically expressed as: where P UE represents the transmission power of the UE, and h k (t) represents the channel gain from the UE to the base station at time slot t, and follows an exponential distribution with a mean of l, and N0 is the white noise power spectral density.

5. The method according to claim 2, wherein Establish a backhaul channel model. The backhaul channel includes a millimeter-wave backhaul channel, a wireless Mesh backhaul channel, and a satellite backhaul channel. The total rate of the backhaul link for each backhaul channel is calculated as follows: (1) The probability that the defined system millimeter-wave backhaul link is a LOS transmission is p LOS (d): where α0 is the environmental occlusion factor, and d represents the distance from the receiving end to the transmitting end of the millimeter-wave link; Define the probability that the system millimeter-wave backhaul link is NLOS transmission as p NLOS (d), then: p NLOS (d) = 1 - p LOS (d) Define the transmission power of SBS as P t , then the received power of MBS is: P r = P t ·G·L(d) -1 where G represents the antenna gain coefficient; L(d) -1 represents the large-scale channel gain, which is a path loss model including shadow fading, and is denoted by PL db (d) represents its dB form: Among them, α represents the path loss value for determining the reference distance from the transmitter to the receiver; β represents the path loss exponent; represents the shadow fading loss parameter, which is a zero-mean normal Gaussian random variable with a variance of ; Define the transmit power of SBS for backhaul using millimeter wave as P t W , after passing through the transmission path, the received power at the MBS is: P r W = P t W ·G·[L LOS (d) -1 ·p LOS (d)+L NLOS (d) -1 ·(1 - p LOS (d))] Signal-to-noise ratio of the millimeter-wave link in time slot t is defined as: where N0 represents the white noise power spectral density, and B W represents the allocated bandwidth of the millimeter-wave fronthaul link; Furthermore, it can be obtained that the total rate of the SBS millimeter-wave backhaul link in the t time slot is: (2) Assume that the space-ground link uses the Ka band, and the channel fading of the SBS satellite return link is set to follow large-scale fading and shadow Rice fading. Then the channel correlation coefficient of the satellite link is as follows: where \(g\) s represents a complex Gaussian variable for Rayleigh fading, and \(\beta\) s (d) follows a lognormal distribution, and \(\alpha\) s represents the path loss coefficient of the satellite link; Signal-to-noise ratio of the t-slot SBS satellite return link is as follows: Among them, P t S represents the transmission power of SBS to the satellite; G S represents the antenna gain; represents the satellite link channel gain; represents the noise variance of additive white Gaussian noise, B S represents the transmission bandwidth of the satellite link; From the above formula, it can be seen that the total rate of the SBS satellite backhaul link in the t time slot is: (3) The SBS transmits data back to the MBS through the wireless Mesh backhaul link. Assume that during this wireless Mesh backhaul link process, it passes through n Mesh nodes M1, M2,..., M n , and the node set M = {M1, M2,..., M n}; The signal-to-interference-plus-noise ratio (SINR) status parameter of the communication link between any two nodes M in the t time slot of a wireless Mesh network a and M b is as follows: a,b (t) is: Among them, represents the transmission power of SBS at M a ; G a,b represents the antenna gain between M a and M b ; h a,b (t) represents the channel correlation coefficient between M a and M b at time slot t; B M represents the transmission channel bandwidth of the SBS wireless Mesh backhaul link; G x,x+1 represents the antenna gain between node M x and M x and its next adjacent node M x+1 ; |h x,x+1 | represents the channel correlation coefficient between M x and M x+1 at time slot t; γ is the orthogonality factor; γ0 represents the given threshold of the Mesh link signal-to-interference-plus-noise ratio parameter; The set SINR of the state parameters of the link signal-to-interference-plus-noise ratio of the connected link between any two nodes in the t-slot SBS wireless Mesh link M (t) is expressed as: SINR M (t) = {SINR 0,1 (t), SINR 1,2 (t),..., SINR n-1,n (t)} From the minimum value of SINR M (t) obtain the total rate of the SBS wireless Mesh backhaul link in time slot t:

6. The method according to claim 2, wherein Assume that Q exists at the SBS W , Q M , Q S Three backhaul queues are used to cache the user data packets received from the SUE. The queue set Q = {Q W , Q M , Q S}; where Q W , Q M , Q S respectively represent that the data packets queuing in the queue are backhauled to the MBS via millimeter wave, wireless Mesh network, and satellite respectively; Assume that a packet group consisting of N' packets of different service types arrives at the SBS. The SBS distributes the packets to the backhaul sub-queues and transmits them sequentially in the sub-queues. At the same time, assume that the packets of different services are independent of each other, and the arrival process of a batch of packets follows a Poisson distribution; According to the set queue rules, the small cell queuing model of queue Q belongs to the birth and death process, where represents that the successive arrival times of data packets follow a negative exponential distribution with parameter λ, D represents the service time, N represents the maximum length of the queue, and ∞ represents an infinite data packet source; Use p i,n to denote the probability that the length of queue Q i is n, where n ∈ {0, 1, ..., N}. According to the analysis method of queuing theory, steady-state equations for the probabilities of each state of the queue are listed for the queue states: where \(i\in\{W,M,S\}\), \(0\leq n\leq N\); the queue \(Q\) i The utilization rate \(\rho\) i of the arrival rate \(\lambda\) i and the service rate ratio \(\mu\) i ; Obtained from the above formula: where p i,0 is the idle probability of queue Q i ; According to the regularity of the queue state: Furthermore, the idle probability p of the queue Q is obtained i as follows i,0 : When the queue length is greater than N, it is considered that the queue is blocked. Therefore, for queue Q i the blocking rate is denoted as B 1,i = p i,N For queue Q i the non-blocking rate is B 0,i = 1 - B 1,i = 1 - p i,N ; Queue Q i The average queue length of is obtained by adding two parts. One part is the queuing length L increased during the process of data packets queuing in the queue i,w and the other part is the waiting length L caused during the transmission of data packets in the SBS sub-queue i,s Furthermore, the average queue length after the queue reaches the equilibrium state is obtained as follows: The effective arrival rate of the packets is the average number of users who can enter the system queue per unit time: λ i,e = λ i (1 - p i,N ) = μ i (1 - p i,0 ) According to Little's formula, the average queuing delay of the packets is: Furthermore, the average queuing delay of the packets is obtained:

7. The method according to claim 1, characterized in that Define the set as the set of data packets to be transmitted in the t-th time slot, where a m,x (t) represents the m-th data packet of service type x, and the effective backhaul rate of a m,x (t) on the backhaul link i is: Among them, represents the normalized matching degree factor, MF i,x represents the matching degree factor of a m,x (t) for queue i, represents a m,x (t) the k-th relevant backhaul network parameter in queue i, is the weight of represents the total rate of the SBS backhaul link i in time slot t, a m,x (t) represents the m-th data packet of service type x in the set of data packets to be transmitted, where x ∈ {uRLLC, eMBB, mMTC} and m ∈ {1, 2,..., M}.

8. The method according to claim 7, wherein Establish a backhaul link allocation problem with the goal of maximizing the delay tolerance elasticity value: Among them, is the delay tolerance elasticity value, T(a m,x (t)) represents the delay threshold of a m,x (t), α is the sensitivity factor, and χ x represents the service gain of a m,x (t), is the average size of the data packet, φ MMV 、φ mesh 、φ S respectively represent the data packet sizes of the millimeter wave backhaul link, the mesh backhaul link, and the satellite backhaul link; is the value of the m-th row and the i-th column of the backhaul link selection matrix , represents that a m,x (t) is assigned to the backhaul link i, represents that a m,x (t) is not assigned to the backhaul link i; d i represents the maximum tolerable delay for the backhaul link i to transmit a data packet; D max 、D min respectively represent the maximum and minimum average queuing delays of the data packet in the queue, and P m,i represents the transmission power of a m,x (t) on the backhaul link i, and P max represents the maximum value of the total instantaneous transmission power of the SBS. φ(a m,x (t)) represents the size of a m,x (t), represents the average queuing delay for a m,x (t) to access the SBS, represents the average queuing delay for a m,x (t) in the queue Q W , represents the average queuing delay for a m,x (t) in the queue Q M , represents the average queuing delay for a m,x (t) in the queue Q S ; δ i ∈{0,1} is the congestion flag corresponding to the queue Q i , and when δ i = 0, it means that the queue is congested, and when δ i = 1, it means that the queue is not congested; τ i is the set threshold; The above backhaul link allocation problem is decomposed into two sub-problems, namely backhaul link selection and backhaul power allocation, by using a two-step method. The two sub-problems are solved separately to obtain the backhaul link selection matrix and P m,i .

9. The method according to claim 8, wherein Simplify the backhaul link allocation problem into a backhaul link selection sub-problem: The solution process of the backhaul link selection sub-problem is as follows: Step 1, initialize D min , D max , d uRLLC , d eMBB , d mMTC , λ, λ = λ1 + λ2 + λ3; Step 2, initialize the data packet transmission power: Step 3, initialize the range of the allowable length of each queue and initialize λ W , λ M , λ S is 1 / 3 of the total arrival rate λ; Step 4, calculate the delay tolerance elasticity value of each data packet and queue and fill it into the delay tolerance elasticity matrix; where the element value in the m-th row and i-th column of the delay tolerance elasticity matrix is Step 5, transform the delay tolerance elasticity matrix into a square matrix to simplify the backhaul link selection sub-problem into a one-to-one assignment problem; Step 6, taking the maximized delay elasticity tolerance value as the objective, execute the Hungarian algorithm to solve the one-to-one assignment problem, obtain the fronthaul link selection matrix, and update λ W , λ M , λ S ; Step 7, calculate the average queuing delay of each queue according to the backhaul link selection matrix; Step 8, if D is satisfied i ≤d i then execute Step 10, otherwise execute Step 9; Step 9, reallocate the packet with the smallest delay tolerance elasticity value in the queue that does not satisfy D i ≤ d i to other queues that satisfy D i ≤ d i to obtain a new backhaul link selection matrix and jump to Step 4; Step 10, output the latest backhaul link allocation matrix and the maximum delay elasticity tolerance value.

10. The method according to claim 8, wherein Simplify the backhaul link allocation problem into a backhaul power allocation sub-problem: Use the Lagrange dual method to transform the backhaul power allocation sub-problem into a concave problem, construct a Lagrangian function to handle the non-linear constraint conditions, and use the gradient descent method to solve the optimal backhaul power allocation value that satisfies the KKT conditions.

Images