Multi-base-station cooperative wireless resource allocation method based on queue sensing

Abstract

The invention discloses a multi-base-station cooperative wireless resource allocation method based on queue sensing, relates to the technical field of electronic communication, and provides a low-complexity distributed resource optimization solution aiming at the problem of random arrival of URLLC service in a multi-cell cellular system. The resource management problem of dynamic URLLC task multi-time-slot downloading under dense cellular network deployment is researched, and the defect that queue stability is neglected in an existing scheme is overcome; a multi-dimensional time coupling amount is decoupled by converting an original problem into a series of online sub-problems with the maximum rate, and then a distributed algorithm with multiple parallel base stations is provided to decompose global calculation tasks to the respective base stations, so that the calculation complexity is reduced; the method can effectively adjust the compromise between the throughput and the queue cache, and has obvious advantages in the aspect of improving the long-term performance of the system.

Description

Technical Field

[0001] The present invention relates to the field of electronic communication technology, and in particular to a multi-base station collaborative wireless resource allocation method based on queue perception. Background Art

[0002] Next-generation mobile communication systems are expected to revolutionize latency and reliability for a wide range of applications. Ultra-Reliable Low-Latency Communication (URLLC) is considered a core application scenario, where immediacy of service and long-term transmission efficiency are crucial and practical issues. Typically, to meet stringent performance standards, short frame structures and finite block length (FBL) coding techniques are widely adopted. However, in harsh propagation environments, users may face severe channel challenges, making FBL coding susceptible to reduced transmission rates and an inevitable increase in decoding error probability. Therefore, mitigating the negative effects of FBL transmission has become a focus of numerous researchers. It is important to note that most current research on URLLC assumes a single-cell scenario, focusing on mitigating intra-cell interference.

[0003] On the other hand, in order to overcome the severe challenges posed by spectrum resource shortages to the immediate capacity of information transmission, increasing base station deployment to improve data transmission rates is gradually being seen as a potentially effective response strategy. This network configuration requires us to rationally allocate limited wireless resources to meet the growing demand for concurrent services. For single-cell systems, it may not be feasible to reserve a specific portion of the spectrum, so spatial frequency reuse must be fully utilized in multi-cell networks. By achieving synchronous transmission on the same sub-channel, the transmission power not only affects the intended receiver, but also causes co-channel interference to non-target receivers. In response to dense spectrum reuse and bursty demands of critical mission types, multi-cell networks must adopt efficient resource management and interference control measures to meet the service quality of long-term URLLC transmission.

[0004] Although existing work has made many improvements to URLLC resource management methods, these studies are mostly based on single-time slot, centralized, and single-cell scenarios, which cannot fully utilize the advantages of dense spectrum reuse and ignore the random nature of data service arrival. The backlog caused by the time-varying cache queue will affect the stability of the communication network. Traditional methods to solve such problems involve the use of dynamic programming, but this easily leads to unmanageable computational complexity. On the other hand, reinforcement learning schemes using Markov decisions require detailed historical training information of the dynamic system and may also lead to exponential growth of the state space. Therefore, combined with the full spectrum reuse of cellular networks and the transmission characteristics of short packet data, with the goal of improving the quality of service (QoS) of long-term URLLC and reducing the computational burden, corresponding resource allocation strategies are still quite scarce. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention provides a queue-aware multi-base station cooperative radio resource allocation method, comprising the following steps:

[0006] S1. Build a system model, including establishing a network model, a downlink transmission model, a queue model, and establishing the system's optimization problems and constraints;

[0007] S2, online conversion of multi-slot long-term effective throughput optimization problems, constructing a series of single-slot continuous optimization problems based on queue stability;

[0008] S3. Perform a queue-aware multi-base station collaborative resource allocation method, including global partitioning of user sets, distributed solution of wireless resource allocation in each cell, and a multi-cell distributed collaborative framework.

[0009] The technical solution further defined in the present invention is:

[0010] Furthermore, the method for establishing the network model in step S1 is: establishing a multi-cell space division multiple access downlink network model, which includes a cloud server, M multi-antenna base stations, and K cellular users with specific URLLC requirements; each user downloads short packet data from the server through the accessed base station; the downlink from the base station to the user is transmitted through N orthogonal subcarriers, where the bandwidth of each subcarrier is B0; each user uses multiple subcarriers, and each subcarrier is also allocated to multiple users through space division multiple access; the transmission duration of each scheduling time slot is defined as T0, where the wireless channel is a quasi-static flat fading channel, and its coefficient remains unchanged within each scheduling time slot.

[0011] As described above, in a method for allocating wireless resources among multiple base stations in cooperation based on queue perception, the method for establishing a downlink transmission model in step S1 is as follows: define the nth subcarrier in the mth base station as (m, n), and let Indicates the indicator variable of the base station and carrier allocation in the tth scheduling time slot. If the (m, n)th subcarrier is allocated to user k, there is otherwise Since each user can only choose one base station to access, Satisfy the following formula:

[0012]

[0013] Where H(·) is the sign function; let the channel coefficient of user k on the (m, n)th subcarrier in the tth scheduling time slot be recorded as Then the corresponding received signal at user k is:

[0014]

[0015] Among them, s k (t) and z0(t) represent the transmitted signal and power spectrum density respectively. Noise; precoding adopts maximum ratio transmission, that is and denote the weight of the normalized vector in the t-th scheduling time slot and the power allocated to user k on the (m, n)-th subcarrier; therefore, the received signal-to-interference-and-noise ratio of the signal is:

[0016]

[0017] in, is the interference suffered by user k on the (m, n)th subcarrier, which consists of intra-cell interference and inter-cell interference:

[0018]

[0019] For URLLC services, the transmission rate of user k on the (m, n)th subcarrier is Expressed as:

[0020]

[0021] in, is the channel dispersion factor; Q -1 (·) is the inverse function of the Gaussian Q function, written as L c Indicates the block length of channel coding; is the decoding error probability of user k; according to formula (5), the decoding error probability at a given transmission rate is written as:

[0022]

[0023] in, Therefore, the number of bits correctly transmitted by user k in the tth scheduled time slot is written as:

[0024]

[0025] Among them, R k (t) represents the number of bits correctly transmitted by user k in the tth scheduled time slot.

[0026] As described above, in a method for allocating wireless resources among multiple base stations in cooperation based on queue perception, the method for establishing a queue model in step S1 is as follows: based on queue theory, a dynamic queue scheduling model is established for each user; let Q k (t) is the queue length of the kth user's data to be transmitted at the beginning of the tth scheduling time slot, and its calculation formula is as follows:

[0027] Q k (t+1)=max{Q k (t)-S k (t), 0}+D k (t), (8)

[0028] Among them, D k (t) represents the amount of download data requested by the kth user in the tth scheduling time slot, S k (t) is the number of bits transmitted, and it is assumed that S k (0)=0 and Q k (0)=0.

[0029] As described above, in a queue-aware multi-base station cooperative wireless resource allocation method, the method for establishing the system optimization problem and constraint conditions in step S1 is: since each user can only access one base station at most, then The report is subject to the following constraints:

[0030]

[0031] The number of base station selection handovers performed by the user in adjacent scheduling time slots meets the following conditions:

[0032]

[0033] The non-negativity constraints on power allocation and transmission rate are expressed as:

[0034]

[0035] The user's transmission reliability is expressed as:

[0036]

[0037] The buffer queue constraint for each user is expressed as the requirement on the amount of data to be transmitted without exceeding the current buffer:

[0038]

[0039] The maximum instantaneous transmission power and maximum average power of each base station are P max and P avg , then the downlink power meets the following conditions:

[0040]

[0041] The stability of the communication network is guaranteed by the following constraints:

[0042]

[0043] Jointly optimize base station selection, subcarriers, power allocation, and rate adaptation to maximize the long-term time-averaged effective throughput of the system. The objective function is expressed as follows:

[0044]

[0045] Among them, R k (t) represents the number of bits correctly transmitted by user k in the tth scheduled time slot.

[0046] As described above, in a queue-aware multi-base station cooperative radio resource allocation method, in step S2, a virtual queue is introduced for the constraint shown in equation (15), and the corresponding virtual power queue is defined as:

[0047]

[0048] in,

[0049] Define a joint queue as Φ(t)={Q1(t),…,Q K (t); U1(t),…,U M (t)}, thus introducing the Lyapunov function as follows:

[0050]

[0051] Accordingly, the Lyapunov drift can be written as:

[0052]

[0053] The objective function in the original problem is maximized by minimizing the drift minus reward function. The expression of the drift minus reward function is written as:

[0054]

[0055] Where V is a non-negative constant that represents the stability of the network. and system performance The trade-off between the two is then derived; the theoretical upper bound expression of formula (21) is derived; according to Lyapunov optimization theory, the upper bound expression of the drift minus reward function is written as:

[0056]

[0057] Where B is a non-negative constant. By removing the constant term in the upper bound of Equation (22), the deterministic problem under the t-th scheduling time slot can be written as:

[0058]

[0059] Thus, the original random optimization problem is transformed into a continuous online problem.

[0060] As described above, in a method for allocating wireless resources of multiple base stations in cooperation based on queue perception, in step S3, a global partitioning method of the user set is used to determine which base station is responsible for transmitting to each user; first, an auxiliary variable y=[y 1，1 ,y 1，2 ,…,y M，K ]∈{0,1} MK , where y m，k =1 indicates that the kth user accesses the mth base station; in addition, y′ represents the base station selection indicator vector of the previous scheduling time slot; by optimizing y, the base station selection subproblem is expressed as follows:

[0061]

[0062] Among them, UT m，k =QkS m，k +VR m,k , The set of access users of base station m is defined as The problem shown in formula (24) is refined into which users are worth switching and how to switch;

[0063] Define the farthest and closest distances of user k to the base station in the current scheduling time slot as d k，max and d k，min , and all |d k，max -d k，min | The elements are arranged in ascending order; for the sorted user list, the first λ0K users are taken as the user set to be switched, and the set is defined as The remaining problem is expressed as a game based on user cluster division, and the cooperative game of any cluster is written as:

[0064]

[0065] in, Indicates that user k is in cluster The benefits gained from is the number of elements in the mth cluster; a base station selection scheme based on cooperative game is proposed, namely the priority association algorithm, which is used to solve the game between user clusters. In this algorithm, a specific base station is selected according to the change of cooperative benefits. The symbol Indicates priority relationship;

[0066] For any user and base station m≠m′, Indicates that user k prefers the m′th base station; the strict priority relationship is defined as follows:

[0067]

[0068] The above formula shows that user k is allowed to access the m′th base station and Then the new user cluster and The corresponding total system utility also increases; by adopting comparison and exchange operations, users first compare the current utility with all other options, and then decide whether to exchange based on the priority relationship.

[0069] As described above, in the queue-aware multi-base station cooperative radio resource allocation method, the distributed solution method for radio resource allocation in each cell in step S3 is as follows: given a base station y is selected, and each base station is responsible for its own resource allocation design in a distributed manner. The optimization problem at the mth base station is expressed as:

[0070]

[0071] An alternating optimization algorithm is proposed to decouple transmission rate and transmit power. First, the transmission rate is optimized under the premise of fixed transmit power. Then, the transmit power is updated based on the transmission rate of this iteration. This process is repeated until convergence.

[0072] Since the transmission rates of users on different subcarriers are independent of each other, they are only affected by the decoding error probability when other variables are fixed. Therefore, optimizing the transmission rate is equivalent to maximizing each sum term in the problem shown in Equation (27):

[0073] As described above, in the queue-aware multi-base station cooperative radio resource allocation method, in step S3, the transmission rate optimization of user k on the (m, n)th subcarrier includes the following sub-steps:

[0074] S3.1. Convert the constraint shown in formula (12) into a solvable form; by deriving formula (6) about The first-order derivative of It can be seen that the probability of decoding error increases with the increase of transmission rate; therefore, since there is a maximum transmission rate to maintain reliability when power is allocated to timing, The maximum value of is written as:

[0075]

[0076] S3.2 Optimal solution for transmission rate It is obtained by solving the following convex problem, expressed as:

[0077]

[0078] And written as:

[0079]

[0080] in, For the equation the root; is defined as:

[0081]

[0082] S3.3. Solve the sub-problem of power allocation under the given transmission rate; by introducing auxiliary variables The problem shown in formula (27) is reconstructed as follows:

[0083]

[0084] in,

[0085]

[0086] Since the above formulas are all concave functions, they are expressed in the form of differential programming;

[0087] According to the properties of f(γ, L, c), auxiliary variables are further introduced To reconstruct constraint C1 as follows:

[0088]

[0089] in, is the inter-cell interference generated by other base stations; it is approximated iteratively using a continuous convex approximation algorithm; it is defined is α m，k The solution in the i-th continuous convex approximation iteration; based on A k (α m，k ) and W k (α m，k ), C4 is strictly approximated as:

[0090]

[0091] In addition, the constraint shown in formula (35) is rewritten as follows:

[0092]

[0093] S3.4. Convert constraint C2 into a convex expression; here is equivalent to the form of l0 norm; it is approximated iteratively using the weighted l1 norm method; according to the above equivalent transformation and approximation, the problem shown in formula (32) is replaced by a convex problem form that maximizes the lower bound of the objective function, and is solved using the interior point method; wherein the solution in the i-th iteration is used as the initial point of the (i+1)-th iteration, and then a series of improved feasible solutions are generated until convergence.

[0094] As described above, in the multi-cell distributed collaborative framework of the queue-aware multi-base station cooperative radio resource allocation method, in step S3, the calculation process is performed by each base station separately, which includes two parts: performing base station selection according to the priority association algorithm, and performing power allocation and transmission rate optimization distributedly by the corresponding base stations;

[0095] At the beginning of each scheduling time slot, all base stations measure and broadcast the corresponding channel state information and the initial power allocation plan. Then, each base station temporarily fixes the power variable on its interfering link and calculates its own resource allocation plan in parallel. The base stations send the updated plan through the broadcast or backhaul link until the plan converges or the number of iterations reaches a predetermined limit. Finally, each base station performs the data transmission task based on the final calculated result.

[0096] The beneficial effects of the present invention are:

[0097] In this invention, the resource management problem of multi-slot downloading of dynamic URLLC tasks under dense cellular network deployment is studied, which makes up for the defect of ignoring queue stability in existing solutions; by converting the original problem into a series of online sub-problems with maximum sum rate to decouple the multi-dimensional time coupling quantity, and then a multi-base station parallel distributed algorithm is proposed to decompose the global computing tasks into respective base stations, thereby reducing the computational complexity; it can effectively adjust the trade-off between throughput and queue cache, and at the same time has obvious advantages in improving the long-term performance of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] Figure 1 A schematic diagram of the structure of a network model in an embodiment of the present invention;

[0099] Figure 2 It is a schematic diagram of the overall process of the present invention;

[0100] Figure 3 Schematic diagram of a curve showing changes in data queue length over time and different control parameters according to an embodiment of the present invention;

[0101] Figure 4 Graph showing the relationship between the number of users and the time-averaged effective throughput in different embodiments of the present invention;

[0102] Figure 5 2 is a diagram showing the relationship between the total size of data to be transmitted and the scheduling time slot in an embodiment of the present invention. DETAILED DESCRIPTION

[0103] This embodiment provides a queue-aware multi-BS collaborative (QAMBC) wireless resource allocation method, and uses a simulation experiment platform to verify the performance of the proposed algorithm, in which cellular users are evenly distributed within the coverage area of multiple base stations, such as Figure 1 As shown, the throughput performance of the method of this embodiment is evaluated by comparing its performance with that of existing algorithms in actual application scenarios.

[0104] This embodiment provides a queue-aware multi-base station cooperative radio resource allocation method, such as Figure 2 As shown, the following steps are included:

[0105] S1. Build a system model, including establishing a network model, a downlink transmission model and a queue model, and establishing the system's optimization problems and constraints.

[0106] The method for establishing a network model is as follows: a multi-cell Space Division Multiple Access (SDMA) downlink network model is established, which includes a cloud server, M multi-antenna base stations, and K cellular users with specific URLLC requirements; each user downloads short packet data from the server through the base station to which it accesses; the downlink from the base station to the user is transmitted through N orthogonal subcarriers, where the bandwidth of each subcarrier is B0; each user uses multiple subcarriers, and each subcarrier is also allocated to multiple users through space division multiple access to improve spectrum efficiency; the transmission duration of each scheduling period (SP) is defined as T0, where the wireless channel is a quasi-static flat fading channel, and its coefficient remains stable within each scheduling period.

[0107] The method of establishing the downlink transmission model is as follows: define the nth subcarrier in the mth base station as (m, n), let Indicates the indicator variable of the base station and carrier allocation in the tth scheduling time slot. If the (m, n)th subcarrier is allocated to user k, there is otherwise Since each user can only choose one base station to access, Satisfy the following formula:

[0108]

[0109] Where H(·) is the sign function; let the channel coefficient of user k on the (m, n)th subcarrier in the tth scheduling time slot be recorded as Then the corresponding received signal at user k is:

[0110]

[0111] Among them, s k (t) and z0(t) represent the transmitted signal and power spectrum density respectively. noise.

[0112] Precoding uses Maximum Ratio Transmission (MRT), that is, and denote the weight of the normalized vector in the t-th scheduling time slot and the power allocated to user k on the (m, n)-th subcarrier. Therefore, the received signal-to-interference-plus-noise ratio (SINR) of the signal is:

[0113]

[0114] in, is the interference suffered by user k on the (m, n)th subcarrier, which consists of intra-cell interference and inter-cell interference:

[0115]

[0116] For URLLC services, the transmission rate of user k on the (m, n)th subcarrier is As shown in the following formula, the unit is bits / s / Hz:

[0117]

[0118] in, is the channel dispersion factor; Q -1 (·) is the inverse function of the Gaussian Q function, written as L c Indicates the block length of channel coding; is the decoding error probability of user k.

[0119] According to formula (43), the decoding error probability at a given transmission rate can be written as:

[0120]

[0121] in, Therefore, the number of bits correctly transmitted by user k in the tth scheduled time slot (effective throughput) is written as:

[0122]

[0123] Among them, R k (t) represents the number of bits correctly transmitted by user k in the tth scheduled time slot.

[0124] Next, we build a transmission queue update model. Based on queuing theory, we build a dynamic queue scheduling model for each user to control the backlog of data packets within a certain range. Let Q k (t) is the queue length of the kth user's data to be transmitted at the beginning of the tth scheduling time slot. It is updated over time and is calculated as follows:

[0125] Q k (t+1)=max{Q k (t)-S k (t), 0}+D k (t), (46)

[0126] Among them, D k (t) represents the amount of download data requested by the kth user in the tth scheduling time slot, S k (t) is the number of bits transmitted, and it is assumed that Sk (0)=0 and Q k (0)=0.

[0127] The method of establishing the optimization problem and constraints of the system is: Since each user can only access one base station at most, Subject to the following constraints:

[0128]

[0129] The number of base station selection handovers performed by the user in adjacent scheduling time slots meets the following conditions:

[0130]

[0131] The non-negativity constraints on power allocation and transmission rate are expressed as:

[0132]

[0133] The user's transmission reliability is expressed as:

[0134]

[0135] The buffer queue constraint for each user is expressed as the requirement on the amount of data to be transmitted without exceeding the current buffer:

[0136]

[0137] The maximum instantaneous transmission power and maximum average power of each base station are P max and P avg , then the downlink power meets the following conditions:

[0138]

[0139] The stability of the communication network is guaranteed by the following constraints:

[0140]

[0141] This embodiment jointly optimizes base station selection, subcarriers, power allocation, and rate adaptation to maximize the long-term time-averaged effective throughput of the system. The objective function is expressed as follows:

[0142]

[0143] Among them, R k (t) represents the number of bits correctly transmitted by user k in the tth scheduled time slot.

[0144] S2. Online conversion of multi-slot long-term effective throughput optimization problem, constructing a series of single-slot continuous optimization problems based on queue stability; in order to deal with the long-term time average constraint on transmit power, a virtual queue is introduced for the constraint shown in Equation (53), and the corresponding virtual power queue is defined as:

[0145]

[0146] in,

[0147] In order to control data cache and energy consumption accumulation at the same time, a joint queue is defined as Φ(t) = {Q1(t), ..., Q K (t); U1(t),…,U M (t)}, thus introducing the Lyapunov function as follows:

[0148]

[0149] Accordingly, the Lyapunov drift can be written as:

[0150]

[0151] By continuously pushing the Lyapunov drift shown in Equation (58) to a smaller value within each SP, the stability of the network will be guaranteed.

[0152] Therefore, the objective function in the original problem can be maximized by minimizing the drift minus reward function. The expression of the drift minus reward function is written as:

[0153]

[0154] Where V represents a non-negative constant, which represents the relationship between network stability △L(Φ(t)) and system performance The trade-off between.

[0155] Then, the theoretical upper bound expression of Equation (59) is derived. According to Lyapunov optimization theory, the upper bound expression of the drift-minus-reward function is written as:

[0156]

[0157] Where B is a non-negative constant. By removing the constant term in the upper bound of Equation (60), the deterministic problem under the t-th scheduling time slot can be written as:

[0158]

[0159] Here, the original random optimization problem is transformed into a continuous online problem.

[0160] S3. Perform a queue-aware multi-base station collaborative resource allocation method, including global partitioning of user sets, distributed solution of wireless resource allocation in each cell, and a multi-cell distributed collaborative framework.

[0161] By using the global partitioning method of the user set, it is determined which base station is responsible for the transmission of each user; first, an auxiliary variable y = [y 1，1 ,y 1，2 ,…,y M，K ]∈{0,1} MK , where y m，k =1 indicates that the kth user accesses the mth base station; in addition, y′ represents the base station selection indicator vector of the previous scheduling time slot; by optimizing y, the base station selection subproblem is expressed as follows:

[0162]

[0163] Among them, UT m，k =QkS m，k +VR m,k , The set of access users of base station m is defined as The problem shown in formula (62) is refined into which users are worth switching and how to switch them.

[0164] According to empirical criteria, users tend to access the nearest base station, while users at the cell edge are more likely to suffer from higher inter-cell interference and lower intra-cell gain. Therefore, the farthest and closest distances of user k to the base station in the current scheduling time slot are defined as d and d respectively. k，max and d k，min , and all |d k，max -d k，min | The elements are arranged in ascending order; for the sorted user list, the first λ0K users are taken as the user set to be switched, and the set is defined as The remaining problem can actually be expressed as a game based on user cluster division.

[0165] Specifically, the cooperative game of any cluster can be written as:

[0166]

[0167] in, Indicates that user k is in cluster The benefits gained from is the number of elements in the mth cluster; To resolve this inter-cluster game, this embodiment proposes a base station selection scheme based on cooperative game, namely, the priority association algorithm. In this algorithm, the switching of users between different clusters is determined based on the sum of their benefits, that is, a specific base station is selected according to the change in cooperative benefit; the symbol is introduced Indicates a priority relationship.

[0168] For any user and base station m≠m′, Indicates that user k prefers the m′th base station; the strict priority relationship is defined as follows:

[0169]

[0170] The above formula shows that user k is allowed to access the m′th base station and Then the new user cluster and The corresponding total system utility also increases; by using comparison and exchange, users first compare the current utility with all other options, and then decide on the exchange operation based on the priority relationship.

[0171] The distributed solution for wireless resource allocation in each cell is as follows: given a base station y, each base station is responsible for its own resource allocation design in a distributed manner. The optimization problem under the mth base station is expressed as:

[0172]

[0173] An alternating optimization algorithm is proposed to decouple transmission rate and transmit power. First, the transmission rate is optimized under the premise of fixed transmit power. Then, the transmit power is updated based on the transmission rate of this iteration. This process is repeated until convergence.

[0174] Since the transmission rates of users on different subcarriers are independent of each other, they are only affected by the decoding error probability when other variables are fixed. Therefore, optimizing the transmission rate is equivalent to maximizing each sum term in the problem shown in Equation (65).

[0175] Without loss of generality, the transmission rate optimization of user k on the (m, n)th subcarrier includes the following steps:

[0176] S3.1. Convert the constraint shown in formula (50) into a solvable form; by deriving formula (44) about The first-order derivative of It can be seen that the probability of decoding error increases with the increase of transmission rate; therefore, since there is a maximum transmission rate to maintain reliability when power is allocated to timing, The maximum value of is written as:

[0177]

[0178] S3.2 Optimal solution for transmission rate It is obtained by solving the following convex problem, expressed as:

[0179]

[0180] And written as:

[0181]

[0182] in, For the equation the root; is defined as:

[0183]

[0184] S3.3. Under a given transmission rate, solve the subproblem about power allocation. Since this problem is non-convex in nature, we introduce the auxiliary variable The problem shown in formula (65) is reconstructed as follows:

[0185]

[0186] in,

[0187]

[0188] Since the above formulas are all concave functions, they are expressed in the form of differential programming.

[0189] According to the properties of f(γ, L, c), auxiliary variables are further introduced To reconstruct constraint C1 as follows:

[0190]

[0191] in, is the inter-cell interference generated by other base stations; it is approximated iteratively using a continuous convex approximation algorithm; it is defined is α m，k The solution in the i-th consecutive convex approximation iteration.

[0192] Based on A k (α m，k ) and W k (α m，k), C4 is strictly approximated as:

[0193]

[0194] In addition, the constraint shown in formula (73) is rewritten as follows:

[0195]

[0196] S3.4. Convert constraint C2 into a convex expression; in fact, here is equivalent to the form of the l0 norm; consider using the weighted l1 norm method to iteratively approximate it; according to the above equivalent transformation and approximation, the problem shown in Equation (70) is replaced by a convex problem form that maximizes the lower bound of the objective function, which can be efficiently solved with the help of the interior point method; wherein, the solution in the i-th iteration is used as the initial point of the (i+1)-th iteration, and then a series of improved feasible solutions are generated until convergence.

[0197] In the multi-cell distributed collaboration framework, the main computational process is performed by each base station separately and can be summarized into two parts: a) base station selection is performed based on the priority association algorithm; b) power allocation and transmission rate optimization are completed in a distributed manner by the corresponding base stations.

[0198] At the beginning of each scheduling time slot, all base stations measure and broadcast the corresponding channel state information and the initial power allocation plan. Then, each base station temporarily fixes the power variable on its interfering link and calculates its own resource allocation plan in parallel. The base stations send the updated plan through the broadcast or backhaul link until the plan converges or the number of iterations reaches a predetermined limit. Finally, each base station performs the data transmission task based on the final calculated result.

[0199] In this example, simulation experiments are used to verify the algorithm's performance. Assume that there are K users served by M base stations. The distance-varying path loss is 120.9 + 37.6log d, where d is the distance from the user to the base station in kilometers. Shadow fading is represented by a lognormal random variable with zero mean and 8 dB variance. The small-scale fading coefficient follows a Rayleigh distribution with unit variance. Each base station is located at the center of a cell with a radius of 300 meters, and users are evenly distributed within the cell.

[0200] Experiment (A): The change in the total amount of data to be transmitted in the system as the number of scheduled time slots increases.

[0201] Experiment (B): Compare the performance of similar algorithms under different parameters, increasing the average number of users per cell from 4 to 16 and the number of carriers from 9 to 15.

[0202] The experimental parameters are shown in Table 1 below.

[0203] Table 1

[0204]

[0205] This example will consider two performance indicators, namely:

[0206] 1. The change of the total amount of data to be transmitted with the increase of the number of transmission time slots under different parameters of the proposed method.

[0207] 2. Comparison of time-averaged effective throughput with similar algorithms under different parameters.

[0208] The simulation results of this example are as follows:

[0209] 1. The change of the total amount of data to be transmitted with the increase of the number of transmission time slots under different parameters of the proposed method.

[0210] like Figure 3 As shown in the figure, it can be seen that in the initial stage, the queue length increases significantly; then, as the system gradually stabilizes, the queue length no longer continues to increase, but enters a relatively stable fluctuation range; at the same time, the control parameter V has a significant impact on the total queue length; when the V value increases, the total queue length tends to decrease. This is because a larger V increases the weight of the original objective function, making the system more inclined to transmit more data, thereby reducing the amount of waiting data in the queue and the transmission delay.

[0211] 2. Comparison of time-averaged effective throughput with similar algorithms under different parameters.

[0212] like Figure 4 As shown in the figure, it can be seen that the time-averaged effective throughput of each scheme increases with the increase in the number of users. This is because the increase in the number of users will increase the chance of finding channels with higher channel gains. This phenomenon is also called user diversity. Among the three algorithms, the proposed QAMBC algorithm achieves the best performance, partly because the algorithm adopts a distributed collaboration mechanism and takes into account interference coordination, thereby achieving significant gains in multi-cell wireless networks. In contrast, the comparison algorithm only considers the maximum benefit of each cell separately, without fully considering the interference caused by other cells.

[0213] like Figure 5As shown in the figure, it can be seen that with the increase of the number of subcarriers, the effective throughput of all schemes also increases accordingly. The reason is that the increase in the number of subcarriers enables the system to process more data simultaneously, thereby improving the overall data transmission capability. In a multi-cell communication environment, the proposed algorithm performs better than other schemes. In contrast, the performance of Algorithm 2 is weaker, mainly due to its shortcomings in carrier allocation. It cannot fully utilize the advantages of multi-carriers like QAMBC, resulting in resource waste and efficiency loss in the data transmission process. This result emphasizes the importance of optimizing carrier allocation in URLLC.

[0214] In addition to the above embodiments, the present invention may also have other implementations. Any technical solution formed by equivalent replacement or equivalent transformation falls within the scope of protection required by the present invention.

Claims

1. A queue-aware multi-base station cooperative radio resource allocation method, characterized by: The following steps are involved: S1. Build a system model, including establishing a network model, a downlink transmission model, a queue model, and establishing the system's optimization problems and constraints; S2, online conversion of multi-slot long-term effective throughput optimization problems, constructing a series of single-slot continuous optimization problems based on queue stability; S3. Perform a queue-aware multi-base station collaborative resource allocation method, including global partitioning of user sets, distributed solution of wireless resource allocation in each cell, and a multi-cell distributed collaborative framework.

2. The queue-aware multi-base station cooperative wireless resource allocation method according to claim 1, characterized in that: The method for establishing the network model in step S1 is as follows: establishing a multi-cell space division multiple access downlink network model, which includes a cloud server, M multi-antenna base stations, and K cellular users with specific URLLC requirements; each user downloads short packet data from the server through the accessed base station; the downlink from the base station to the user is transmitted through N orthogonal subcarriers, where the bandwidth of each subcarrier is B0; each user uses multiple subcarriers, and each subcarrier is also allocated to multiple users through space division multiple access; defining the transmission duration of each scheduling time slot as T0, wherein the wireless channel is a quasi-static flat fading channel, and its coefficient remains unchanged within each scheduling time slot.

3. The queue-aware multi-base station cooperative wireless resource allocation method according to claim 2, characterized in that: The method for establishing the downlink transmission model in step S1 is: define the nth subcarrier in the mth base station as (m, n), let Indicates the indicator variable of the base station and carrier allocation in the tth scheduling time slot. If the (m, n)th subcarrier is allocated to user k, there is otherwise Since each user can only choose one base station to access, Satisfy the following formula: Where H(·) is the sign function; let the channel coefficient of user k on the (m, n)th subcarrier in the tth scheduling time slot be recorded as Then the corresponding received signal at user k is: Among them, s k (t) and z0(t) represent the transmitted signal and power spectrum density respectively. Noise; precoding adopts maximum ratio transmission, that is and denote the weight of the normalized vector in the t-th scheduling time slot and the power allocated to user k on the (m, n)-th subcarrier. Therefore, the received signal-to-interference-and-noise ratio of the signal is: in, is the interference suffered by user k on the (m, n)th subcarrier, which consists of intra-cell interference and inter-cell interference: For URLLC services, the transmission rate of user k on the (m, n)th subcarrier is Expressed as: in, is the channel dispersion factor; Q -1 (·) is the inverse function of the Gaussian Q function, written as L c Indicates the block length of channel coding; is the decoding error probability of user k; according to formula (5), the decoding error probability at a given transmission rate is written as: in, Therefore, the number of bits correctly transmitted by user k in the tth scheduled time slot is written as: Among them, R k (t) represents the number of bits correctly transmitted by user k in the tth scheduled time slot.

4. The queue-aware multi-base station cooperative wireless resource allocation method according to claim 1, characterized in that: The method for establishing the queue model in step S1 is: based on queueing theory, a dynamic queue scheduling model is established for each user; let Q k (t) is the queue length of the kth user's data to be transmitted at the beginning of the tth scheduling time slot, and its calculation formula is as follows: Q k (t+1)=maX{Q k (t)-S k (t),0}+D k (t), (8) Among them, D k (t) represents the amount of download data requested by the kth user in the tth scheduling time slot, S k (t) is the number of bits transmitted, and it is assumed that S k (0)=0 and Q k (0)=0.

5. The queue-aware multi-base station cooperative wireless resource allocation method according to claim 3, wherein: The method of establishing the optimization problem and constraint conditions of the system in step S1 is: since each user can only access one base station at most, Subject to the following constraints: The number of base station selection handovers performed by the user in adjacent scheduling time slots meets the following conditions: The non-negativity constraints on power allocation and transmission rate are expressed as: The user's transmission reliability is expressed as: The buffer queue constraint for each user is expressed as the requirement on the amount of data to be transmitted without exceeding the current buffer: The maximum instantaneous transmission power and maximum average power of each base station are P max and P avg , then the downlink power meets the following conditions: The stability of the communication network is guaranteed by the following constraints: Jointly optimize base station selection, subcarriers, power allocation, and rate adaptation to maximize the long-term time-averaged effective throughput of the system. The objective function is expressed as follows: Among them, R k (t) represents the number of bits correctly transmitted by user k in the tth scheduled time slot.

6. The queue-aware multi-base station cooperative radio resource allocation method according to claim 5, characterized in that: In step S2, a virtual queue is introduced for the constraint shown in formula (15), and the corresponding virtual power queue is defined as: in, Define a joint queue as Φ(t)={Q1(t),…,Q K (t); U1(t),…,U M (t)}, thus introducing the Lyapunov function as follows: Accordingly, the Lyapunov drift can be written as: The objective function in the original problem is maximized by minimizing the drift minus reward function. The expression of the drift minus reward function is written as: Where V is a non-negative constant that represents the stability of the network. and system performance The trade-off between the two is then derived; the theoretical upper bound expression of formula (21) is derived; according to Lyapunov optimization theory, the upper bound expression of the drift minus reward function is written as: Where B is a non-negative constant. By removing the constant term in the upper bound of Equation (22), the deterministic problem under the t-th scheduling time slot can be written as: Thus, the original random optimization problem is transformed into a continuous online problem.

7. The queue-aware multi-base station cooperative wireless resource allocation method according to claim 1, characterized in that: In step S3, a global partitioning method of the user set is used to determine which base station is responsible for transmission to each user; first, an auxiliary variable y=[y 1，1 ,y 1，2 ,…,y M，K ]∈{0,1} MK , where y m，k =1 indicates that the kth user accesses the mth base station; in addition, y′ represents the base station selection indicator vector of the previous scheduling time slot; by optimizing y, the base station selection subproblem is expressed as follows: in, The set of access users of base station m is defined as The problem shown in formula (24) is refined into which users are worth switching and how to switch; Define the farthest and closest distances of user k to the base station in the current scheduling time slot as d k,max and d k,min , and all |d k，max -d k，min | The elements are arranged in ascending order; for the sorted user list, the first λ0K users are taken as the user set to be switched, and the set is defined as The remaining problem is expressed as a game based on user cluster division, and the cooperative game of any cluster is written as: in, Indicates that user k is in cluster The benefits gained from is the number of elements in the mth cluster. A base station selection scheme based on cooperative game, namely the priority association algorithm, is proposed to solve the game between user clusters. In this algorithm, specific base stations are selected according to the changes in cooperative benefits. The symbol > is introduced to represent the priority relationship. For any user and base station m≠m′, Indicates that user k prefers the m′th base station; the strict priority relationship is defined as follows: The above formula shows that user k is allowed to access the m′th base station and Then the new user cluster and The corresponding total system utility also increases; by adopting comparison and exchange operations, users first compare the current utility with all other options, and then decide whether to exchange based on the priority relationship.

8. The queue-aware multi-base station cooperative radio resource allocation method according to claim 7, characterized in that: The distributed solution method for wireless resource allocation in each cell in step S3 is: given a base station y, each base station is responsible for its own resource allocation design in a distributed manner. The optimization problem under the mth base station is expressed as: An alternating optimization algorithm is proposed to decouple transmission rate and transmit power. First, the transmission rate is optimized under the premise of fixed transmit power. Then, the transmit power is updated based on the transmission rate of this iteration. This process is repeated until convergence. Since the transmission rates of users on different subcarriers are independent of each other, they are only affected by the decoding error probability when other variables are fixed. Therefore, optimizing the transmission rate is equivalent to maximizing each sum term in the problem shown in Equation (27):

9. The queue-aware multi-base station cooperative radio resource allocation method according to claim 8, characterized in that: In step S3, the transmission rate optimization of user k on the (m, n)th subcarrier includes the following sub-steps: S3.

1. Convert the constraint shown in formula (12) into a solvable form; by deriving formula (6) about The first-order derivative of It can be seen that the probability of decoding error increases with the increase of transmission rate; therefore, since there is a maximum transmission rate to maintain reliability when power is allocated to timing, The maximum value of is written as: S3.2 Optimal solution for transmission rate It is obtained by solving the following convex problem, expressed as: And written as: in, For the equation the root; is defined as: S3.

3. Solve the sub-problem of power allocation under the given transmission rate; by introducing auxiliary variables The problem shown in formula (27) is reconstructed as follows: in, Since the above formulas are all concave functions, they are expressed in the form of differential programming; According to the properties of f(γ, L, c), auxiliary variables are further introduced To reconstruct constraint C1 as follows: in, is the inter-cell interference generated by other base stations; it is approximated iteratively using a continuous convex approximation algorithm; it is defined is α m，k The solution in the i-th continuous convex approximation iteration; based on A k (α m，k ) and W k (α m，k ), C4 is strictly approximated as: In addition, the constraint shown in formula (35) is rewritten as follows: S3.

4. Convert constraint C2 into a convex expression; here is equivalent to the form of l0 norm; it is approximated iteratively using the weighted l1 norm method; according to the above equivalent transformation and approximation, the problem shown in formula (32) is replaced by a convex problem form that maximizes the lower bound of the objective function, and is solved using the interior point method; wherein the solution in the i-th iteration is used as the initial point of the (i+1)-th iteration, and then a series of improved feasible solutions are generated until convergence.

10. The queue-aware multi-base station cooperative radio resource allocation method according to claim 1, characterized in that: In the multi-cell distributed cooperation framework of step S3, the calculation process is completed by each base station separately, which includes two parts: performing base station selection according to the priority association algorithm and completing power allocation and transmission rate optimization distributedly by the corresponding base stations; At the beginning of each scheduling time slot, all base stations measure and broadcast the corresponding channel state information and the initial power allocation plan. Then, each base station temporarily fixes the power variable on its interfering link and calculates its own resource allocation plan in parallel. The base stations send the updated solutions through broadcast or backhaul links until the solution converges or the number of iterations reaches a predetermined limit; finally, each base station performs the data transmission task based on the final calculated result.

Images