Genetic user scheduling method for massive MIMO low-orbit satellite communication system

Abstract

The present invention proposes a genetic user scheduling method for a large-scale MIMO low-orbit satellite communication system. A large-scale uniform planar antenna array is configured on the satellite base station side. The base station groups users by using the statistical channel state information of each user. Users in the same group are scheduled to use the same time-frequency resources, and users in different groups use different time-frequency resources. First, a mathematical model for user scheduling is given. Maximizing the downlink achievable average sum rate is used as the design criterion. A 0-1 indicator variable is introduced to represent the user's scheduling state, and the sum rate maximization problem is established as a 0-1 integer programming model. A genetic algorithm is used to solve the problem. First, the solution to the problem is encoded into a chromosome of a certain length. After selection, crossover and mutation operations, the offspring individual with the highest fitness can be found after a certain number of iterations, that is, the optimal solution to the problem. The present invention greatly improves the transmission rate of a large-scale MIMO low-orbit satellite communication system.

Description

Technical Field

[0001] The invention belongs to the field of communication, and in particular relates to a user scheduling method based on a genetic algorithm and utilizing statistical channel state information in low-orbit satellite communication using a large-scale antenna array. Background Art

[0002] In the massive MIMO low-orbit satellite communication system, the base station deploys a large-scale antenna array to serve multiple users simultaneously. The use of massive MIMO technology can effectively reduce interference between users and greatly improve the energy efficiency and spectrum efficiency of wireless communication systems. At the same time, the maximum signal-to-leakage-to-noise ratio precoding and maximum signal-to-noise ratio receiver that uses statistical channel state information can effectively avoid the difficulty of obtaining instantaneous channel state information.

[0003] The number of users within the coverage of a massive MIMO low-orbit satellite communication system is usually much greater than the number of antennas in the base station, making it impossible for the satellite to serve all users within the coverage area in the same time-frequency resource. To solve this problem, users can be scheduled so that users in the same group use the same time-frequency resources, and users in different groups use different time-frequency resources. Therefore, it is very important to design an efficient user scheduling algorithm to reasonably group users.

[0004] Common user scheduling algorithms include exhaustive search methods, which exhaustively search all difficult user combinations, but at the same time bring huge computational burdens, making short-term scheduling extremely difficult in actual implementation and only applicable to scenarios with a small number of users. In the greedy algorithm, the greedy method achieves better sum rate performance and reduces the complexity of the algorithm. However, the greedy algorithm has the probability of falling into a local optimal solution, resulting in the system sum rate performance not being able to reach the optimal. This method avoids the problem from falling into a local optimal solution, greatly improves the system sum rate performance, and has important practical significance. Summary of the invention

[0005] Purpose of the invention: For a low-orbit satellite communication system using massive MIMO, the present invention provides a genetic user scheduling algorithm using statistical channel state information to improve the system's sum rate performance.

[0006] Technical solution: Consider a large-scale MIMO low-orbit satellite communication system with M satellite-side base station antennas and K single-antenna users. The satellite base station uses the spatial angle information of each single-antenna user to group users in the coverage area. Users scheduled in the same group use the same time-frequency resources to communicate wirelessly with the satellite base station, and users scheduled in different groups use different time-frequency resources to communicate wirelessly with the satellite. The satellite base station uses the statistical channel state information of users scheduled in the same group (including the spatial angle information of each user and the average channel energy) to calculate the downlink precoding vector of each user in the group:

[0007]

[0008] in is the energy normalization coefficient such that v k With v i are the antenna array response vectors of the uniform planar array of user k and , respectively, i is the channel energy of user i, ρ k is the downlink signal-to-noise ratio of user k, I M is a unit matrix with dimension M. In the maximum signal leakage noise ratio precoding transmission system based on statistical channel state information, the downlink traversal and rate R of the spatial angle user scheduling method dl for

[0009]

[0010] Among them, G is the number of user groups, q k With q i are the transmission signal energies allocated to user k and user i, respectively, and g k is the channel gain of user k, A g is the set of users in group g, σ k is the channel noise variance of user k, is the precoding vector of user i.

[0011] A total of K users in the beam coverage area are scheduled in G groups, R k,g is the downlink reachable sum rate of user g in the g-th user group, q k,g is the power of user k in group g, q sum is the total transmission power constraint of the downlink. In this problem, a binary scheduling indicator variable c is introduced k,g To indicate the relationship between a specified user k and a user group g, the values ​​of the indicator variables and their respective meanings are

[0012]

[0013] The downlink reachable traversal and rate of user k in user group g is

[0014]

[0015] In the user scheduling problem of massive MIMO low-orbit satellite communication system, the system downlink and rate maximization problem is established as a 0-1 integer programming problem:

[0016]

[0017] And there are constraints It shows that any user k can only be scheduled in one group at most. In order to simplify the analysis and feasibility in the actual system, it is assumed that the power distribution between users is uniform. Then the above spectrum efficiency maximization problem is simplified to

[0018]

[0019] The user scheduling algorithm based on genetic algorithm first randomly initializes I groups of possible user scheduling combinations in turn, and each user group has M randomly grouped scheduling users. Since the goal is to maximize the downlink average sum rate, the average sum rate of each group is calculated respectively, and the average sum rate is used as the fitness of each user group. Several groups with higher fitness are selected for crossover, that is, the two parent user groups exchange several users in their respective groups to generate two subgroups. In addition, in each iteration, the user group with the largest sum rate is selected, and the users in the group are changed based on this user group. Several users in the user group are randomly replaced to generate J new mutated user groups, and IJ-1 user groups are selected from the original user group set to join. The end condition of the algorithm is that the algorithm reaches the maximum number of iterations L set by the algorithm setter. After the algorithm iteration ends, the group with the largest sum rate among the user groups output in the last iteration is selected as the final user group result of the scheduling, and the genetic user scheduling algorithm is executed G-1 times to obtain G groups, and the final scheduling results obtained by each algorithm iteration must be different. The algorithm flow of this method is:

[0020] a) Initialization parameters, number of user groups G, number of users K, user channel direction vector v k,g , indicator variable Precoding vector User group set S g ,g=1,...,G, the encoding length of the chromosome is the maximum number of users M in each group, and the crossover probability α C , mutation probability β M , terminate the evolutionary algebra N iter ;

[0021] b) Randomly generate an initial population, randomly initialize a user group set, which contains I possible user scheduling combinations, and calculate the average and rate of each user group as the fitness parameter of each individual;

[0022] c) Select the first L < k user groups with the highest fitness in the population set for crossover, randomly select several intersection points in the two crossover user groups, exchange the users in the two groups, generate two new offspring user groups, and increase the current number of iterations by 1: n = n + 1;

[0023] d) Randomly select IJ-1 user groups from the initial user set and the user group A with the highest fitness max And J mutation groups are merged to generate a new generation of individuals;

[0024] e) If the number of iterations n is less than the termination evolution generation n≤N gen , then continue to start the next iteration from step b), otherwise terminate the iteration and execute the next step;

[0025] f) Select the user group with the largest fitness in the set, that is, the user group with the largest average and rate, as the g-th user scheduling group result, and calculate the next user grouping from step b) until G user grouping results are obtained.

[0026] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0027] 1. Statistical channel state information is used to design the downlink precoder. Compared with instantaneous channel state information, slowly changing statistical channel state information is easier to obtain, so that the reciprocity of the channel can be utilized in the TDD system, and a large amount of training and feedback load is avoided in the FDD system, reducing the complexity of the system.

[0028] 2. The introduction of 0-1 indicator variables transforms the user scheduling problem of rate maximization into a 0-1 integer programming problem, and the assumption of uniform power distribution among users simplifies the problem.

[0029] 3. The genetic algorithm is used to solve the 0-1 integer programming problem, avoiding the problem of falling into the local optimal solution in the greedy algorithm, and greatly improving the system's sum rate performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 The figure is a flow chart of a method according to an embodiment of the present invention.

[0031] Figure 2 This is a flow chart of a genetic user scheduling algorithm using statistical channel state information in an embodiment of the present invention.

[0032] Figure 3 4 is a graph comparing the sum and rate performances of different scheduling algorithms in an embodiment of the present invention. DETAILED DESCRIPTION

[0033] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0034] The specific steps of the embodiment of the present invention are described below in conjunction with specific scenarios:

[0035] 1) Massive MIMO low-orbit satellite signal transmission model and precoding

[0036] In the massive MIMO low-orbit satellite communication system, the satellite is equipped with a large-scale antenna array to simultaneously serve a large number of single-antenna users. The satellite is equipped with a uniform planar array, which has M x With M y Antennas, M x With M y are the number of antennas in the x-axis and y-axis directions of the uniform planar array. Without loss of generality, assume that the antenna spacing in the x-axis and y-axis directions is 0.5 times the wavelength, M x With M y are all even numbers. The channel between the satellite and user k is

[0037]

[0038] Among them, P k represents the number of confidences of user k, represents the complex gain of the user, ν k,p represents the Doppler shift, τ k,p represents the transmission delay, v k,p is the antenna array response vector of the uniform planar array. Note that the channel model in equation (1) is applicable to the scenario where the position of the low-orbit satellite does not change significantly relative to user k. Therefore, it is assumed that the physical parameter of the channel P_{k}, ν k,p , τ k,p Unchanged over time.

[0039] In downlink transmission, a time-frequency resource block of a low-orbit satellite simultaneously serves K single-antenna users. The user set is represented by K = {0, 1, ..., K-1}. After the base station performs linear precoding, the received signal of user k∈K is

[0040]

[0041] Among them, the subcarrier and symbol subscripts are omitted to simplify the symbols, q kis the transmitted signal energy allocated to user k. k is a standardized precoding vector that satisfies s k is the signal sent to user k, with a mean of 0 and a variance of 1. k is an additive circularly symmetric complex Gaussian noise with a mean of 0 and a variance of σ k , that is, z k ~CN(0,σ k ).

[0042] In downlink multi-user MIMO transmission, the signal-to-leakage-plus-noise-ratio (SLNR) is widely used as a convenient and efficient design criterion. In the precoding method based on the maximum SLNR criterion, the SLNR of user k is k for

[0043]

[0044] in, is the downlink signal-to-noise ratio of user k. Then, the precoder that maximizes the signal-to-noise ratio of user k is

[0045]

[0046] In the above formula, (·) * represents the conjugate operation, is the energy normalization coefficient In the above formula, the precoder based on the maximum signal leakage and noise ratio needs to obtain the instantaneous channel state information of the user, but it is usually difficult to obtain the instantaneous channel state information. Therefore, the precoder based on statistical channel state information is studied, and the signal leakage and noise ratio is rewritten as the average signal leakage and noise ratio:

[0047]

[0048] in is the downlink signal-to-noise ratio (SNR) of user k. k The maximum downlink precoding vector is

[0049]

[0050] in is the energy normalization coefficient such that The corresponding maximum ASLNR is

[0051]

[0052] In the maximum signal-leakage-to-noise ratio precoding transmission system based on statistical channel state information, the downlink traversal and rate of the spatial angle user scheduling method are:

[0053]

[0054] The user scheduling algorithm based on genetic algorithm first randomly initializes I groups of possible user scheduling combinations in turn, and each user group has M randomly grouped scheduling users. Since the goal is to maximize the downlink average sum rate, the average sum rate of each group is calculated respectively, and the average sum rate is used as the fitness of each user group. Several groups with higher fitness are selected for crossover, that is, the two parent user groups exchange several users in their respective groups to generate two subgroups. In addition, in each iteration, the user group with the largest sum rate is selected, and the users in the group are changed based on this user group. Several users in the user group are randomly replaced to generate J new mutated user groups, and IJ-1 user groups are selected from the original user group set to join. The end condition of the algorithm is that the algorithm reaches the maximum number of iterations L set by the algorithm setter. After the algorithm iteration ends, the group with the largest sum rate among the user groups output in the last iteration is selected as the final user group result of the scheduling, and the genetic user scheduling algorithm is executed G-1 times to obtain G groups, and the final scheduling results obtained by each algorithm iteration must be different. The algorithm flow of this method is:

[0055] Step 1: Initialize parameters, user group number G, user number K, user channel direction vector v k,g , indicator variable Precoding vector User group set S g ,g=1,…,G, the encoding length of the chromosome is the maximum number of users M in each group, and the crossover probability α C , mutation probability β M , terminate the evolutionary algebra N iter ;

[0056] Step 2: Randomly generate an initial population, randomly initialize a user group set, which contains I possible user scheduling combinations, and calculate the average and rate of each user group as the fitness parameter of each individual;

[0057] Step 3: Select the first L < k user groups with the highest fitness in the population set for crossover, randomly select several intersection points in the two crossover user groups, exchange the users in the two groups, generate two new offspring user groups, and increase the current number of iterations by 1: n = n + 1;

[0058] Step 4: Randomly select IJ-1 user groups from the initial user set and the user group A with the highest fitness maxAnd J mutation groups are merged to generate a new generation of individuals;

[0059] Step 5: If the number of iterations n is less than the number of generations of evolution n≤N, terminate the evolution gen , then continue to start the next iteration from step b), otherwise terminate the iteration and execute the next step;

[0060] Step 6: Select the user group with the largest fitness in the set, that is, the user group with the largest average and rate, as the g-th user scheduling group result, and calculate the next user grouping from step b) until G user grouping results are obtained.

[0061] Figure 3 The sum rate performance of the genetic user scheduling algorithm and the semi-orthogonal user scheduling algorithm are compared when the number of antennas is 128. GA is the genetic user scheduling algorithm, SA is the simulated annealing user scheduling algorithm, SAUG is the spatial angle user scheduling method, Semi-Orthogonal is the semi-orthogonal user scheduling method, and Greedy Algorithm is the greedy user scheduling method. Among them, the user scheduling method based on the genetic algorithm has the best sum rate performance.

Claims

1. A genetic user scheduling method for a massive MIMO low-orbit satellite communication system, characterized in that: The satellite base station groups users in the coverage area, calculates the downlink precoding vector and downlink average sum rate of each group, and takes maximizing the system downlink average sum rate as the design criterion for user scheduling. A 0-1 indicator variable is introduced to represent the user's scheduling state, and the downlink average sum rate maximization problem is established as a 0-1 integer programming problem. A genetic algorithm is used to solve the 0-1 integer programming problem to obtain the user grouping result. The satellite base station uses the spatial angle information of each single-antenna user to group users in the coverage area. Users scheduled in the same group use the same time-frequency resources to communicate wirelessly with the satellite base station, and users scheduled in different groups use different time-frequency resources to communicate wirelessly with the satellite. The satellite base station side calculates the downlink precoding vector of each user in the group using the statistical channel state information of the users scheduled in the same group; The statistical channel state information includes the spatial angle information of user k And its average channel energy γ k , is obtained by uplink channel detection or through feedback information from each user; The downlink precoding vector is: in,(·) * represents the conjugate operation, is the energy normalization coefficient such that v k With v i are the antenna array response vectors of the uniform planar array for user k and user i, respectively, i is the channel energy of user i, ρ k is the downlink signal-to-noise ratio of user k, I M is the M-dimensional unit matrix; Downlink average sum rate R dl for Among them, G is the number of user groups, q k With q i are the transmission signal energies allocated to user k and user i, respectively, and g k is the channel gain of user k, A g is the set of users in group g, σ k is the channel noise variance of user k, is the precoding vector of user i; Introducing a 0-1 indicator variable to represent the user's scheduling status, the downlink average and rate maximization problem is established as a 0-1 integer programming problem. The specific steps include: A total of K users in the beam coverage area are scheduled in G groups, R k,g is the downlink sum rate of user k in the g-th user group, q k,g is the power of the users in group g, q sum is the total transmission power constraint of the downlink; the binary scheduling indicator variable c k,g Indicates the relationship between a specified user k and a user group g. The values ​​of the indicator variables and their respective meanings are The downlink sum rate of user k in user group g is Among them, q k,g With q i,g are the transmission signal energies allocated to user k and user i in group g, respectively. k,g is the channel gain of user k in group g, σ k,g is the channel noise variance of user k in group g, and are the precoding vectors of user i and user k in group g respectively; The downlink average and rate maximization problem is formulated as a 0-1 integer programming problem: Constraints This indicates that any user k can be scheduled in at most one group; A genetic algorithm is used to solve the 0-1 integer programming problem. The specific steps include: first, randomly initialize I groups of possible user scheduling combinations in turn, and each user group has M randomly grouped scheduling users; calculate the downlink average sum rate of each group respectively, and use the downlink average sum rate as the fitness of each user group, and select several groups with high fitness for crossover, that is, the two parent user groups exchange several users in their respective groups to generate two subgroups; in each iteration, the user group with the largest downlink average sum rate is selected, and the users in the group are changed based on the user group, and several users in the group are randomly replaced to generate J new mutated user groups, and IJ-1 user groups are selected from the original user group set to join; the end condition of the algorithm is that the algorithm reaches a predetermined maximum number of iterations L; after the algorithm iteration ends, the group with the largest downlink average sum rate in the user group output in the last iteration is selected as the final user group result of the scheduling, and the genetic user scheduling algorithm is executed G-1 times to obtain G groups, and the final scheduling results obtained by each algorithm iteration must be different; The algorithm flow of genetic algorithm is: a) Initialization parameters, number of user groups G, number of users K, user channel direction vector v k,g , indicator variable Precoding vector User group set S g ,g=1,...,G, the encoding length of the chromosome is the maximum number of users M in each group, and the crossover probability α C , mutation probability β M , terminate the evolutionary algebra N iter ; b) Randomly generate an initial population, randomly initialize a user group set, which contains I possible user scheduling combinations, and calculate the average and rate of each user group as the fitness parameter of each individual; c) Select the first L < k user groups with the highest fitness in the population set for crossover, randomly select several intersection points in the two crossover user groups, exchange the users in the two groups, generate two new offspring user groups, and increase the current number of iterations by 1: n = n + 1; d) Randomly select IJ-1 user groups from the initial user set and the user group A with the highest fitness max And J mutation groups are merged to generate a new generation of individuals; e) If the number of iterations i is less than the number of termination evolution generations n≤N gen , then continue to start the next iteration from step b), otherwise terminate the iteration and execute the next step; f) Select the user group with the largest fitness in the set, that is, the user group with the largest average and rate, as the g-th user scheduling group result, and calculate the next user grouping from step b) until G user grouping results are obtained.

Images