A user grouping method for millimeter wave massive MIMO system based on genetic algorithm

Abstract

The present invention provides a method for user grouping of a millimeter wave massive MIMO system based on a genetic algorithm, which adopts a survival of the fittest genetic mechanism to perform user grouping. From a genetic point of view, dynamically adjusting the user grouping method to ultimately achieve stable grouping is a genetic evolutionary process, in which different user groups constitute different elites in a generation, and the genes (or sub-individuals) of the elites are crossed to produce new offspring, and the parent generation and offspring are simultaneously subjected to preferential treatment to complete the natural selection mechanism, and finally a stable user grouping is obtained. Compared with the early heuristic method, the present invention has a greater advantage in performance; compared with the greedy method, its complexity is significantly reduced while the performance is similar.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication technology, and in particular to a millimeter wave large-scale MIMO system user grouping method based on genetic algorithm. Background Art

[0002] With the development of society and the advancement of science and technology, people's demand for mobile communication traffic and connections is also growing. As one of the key technologies for future mobile communications, millimeter-wave (mmWave) massive MIMO (Multiple-Input and Multiple-Output, MIMO) technology is booming. Among them, although the millimeter wave band has higher path loss, its shorter wavelength makes massive MIMO technology easier to implement. At the same time, the gain brought by the large-scale antenna array also makes up for its disadvantage of high path loss to a certain extent, making millimeter wave technology and massive MIMO technology complement each other.

[0003] Considering the dual demands of the Internet of Things for the number of connections and bandwidth in the future, a single multiple access technology can no longer meet the huge connection demand. In this context, it has become a realistic choice to mix multiple multiple access technologies to provide more connections and optimize user allocation for different multiple access technologies to improve the system and rate. A common cooking is to combine orthogonal multiple access technologies such as time division multiple access (TDMA) and frequency division multiple access (FDMA) with space division multiple access (ADMA). However, although space division multiple access technology can bring huge multi-user rate gain, it is easily affected by user channel correlation. User groups with strong correlation will seriously affect their rate performance. In response to this situation, many scholars have carried out research on user grouping, among which algorithms represented by semi-orthogonal user grouping and greedy grouping have achieved good results, but have not yet reached the limit of grouping gain. Summary of the invention

[0004] Technical issues

[0005] To solve the above problems, the present invention provides a millimeter wave massive MIMO system user grouping method based on genetic algorithm, which greatly improves the system sum rate in the hybrid multiple access system by reasonably allocating users.

[0006] Technical Solution

[0007] In order to achieve the above object, the present invention provides the following technical solutions:

[0008] A method for grouping users in a millimeter wave massive MIMO system based on a genetic algorithm comprises the following steps:

[0009] S1: Establish a mathematical model for user grouping in mmWave massive MIMO systems;

[0010] S2: Maximizing the system and rate problem based on establishing groups;

[0011] S3: Use genetic algorithm to give a grouping algorithm for the problem.

[0012] Furthermore, the step S1 specifically includes the following sub-steps:

[0013] S1.1: Configure a mmWave all-digital time-division duplex massive MIMO system with N-antenna ULAs serving K single-antenna users, where the base station unit has learned the user's channel information through channel estimation;

[0014] S1.2: Establish a data transmission model, assuming that the downlink channel is H = [h1 T ,h2 T ,...,h K T ] T , where h k is the channel from N-antenna ULA to user k, which is modeled according to the Saleh-Valenzuela millimeter wave channel model:

[0015]

[0016] in is the complex gain of the l-th path of user k, which is composed of path loss and shadow fading, L is the number of multipaths, l = 0 represents the line-of-sight component, l ≥ 1 represents the non-line-of-sight component, is the array response vector of N-antenna ULA, expressed as:

[0017]

[0018] in, is the arrival angle of path l of user k, (·) T stands for transpose;

[0019] Let p k As its precoding vector, the data rate it receives is:

[0020]

[0021] Where p is the total power of the transmitted signal, σ 2 is the noise variance, log represents logarithmic operation, and |·| represents modulus operation;

[0022] S1.3: Use orthogonal multiple access to divide K users into G>1 groups on average, and use SDMA to serve users in the same group. is the entire set of users, then the grouping rule can be described as:

[0023]

[0024] S1.4: Grouping The rate of user k in is expressed as:

[0025]

[0026] Where p is the total power of the transmitted signal, h k Downlink channel of user k, σ 2 is the noise variance, then the group The sum rate is expressed as:

[0027]

[0028] Assume grouping The resource allocation coefficient in the orthogonal multiple access system is Then the system and rate can be expressed as:

[0029]

[0030] S2: Establish system maximization and rate problem:

[0031]

[0032] S3.1: Set the number of iterations ITER, the number of elites E, the number of groups G, the number of offspring S, the number of crossovers Mut and the threshold γ;

[0033] S3.2: First, generate the initial parent generation, that is, for each elite e, randomly generate groups according to the number of groups G

[0034] S3.3: Start iteration, and perform the following operations in each iteration:

[0035] S3.3.1: First, generate S offspring for each parent elite e. The specific operations on the offspring s are as follows:

[0036] S3.3.1.1: First, let the offspring s be And randomly select the crossover number mut∈{1,2,...,Mut};

[0037] S3.3.1.2: In each crossover, the group number and the two users to be crossovered are randomly selected to complete the mutation;

[0038] S3.3.1.3: Calculate the sum rate R based on the grouping information of the offspring s;

[0039] S3.3.2: After completing the above operations for all parents, consider all the parents and their children and retain the E elites with the largest sum rate;

[0040] S3.4: If the optimization degree is less than the threshold value γ for multiple consecutive times, the iteration ends and the grouping method with the largest sum rate is the final grouping method. Otherwise, jump to S to perform the next iteration on the E elites obtained in the previous step.

[0041] Beneficial Effects

[0042] The present invention discloses a method for user grouping of a millimeter wave massive MIMO system based on a genetic algorithm, which adopts a survival of the fittest genetic mechanism to perform user grouping. From a genetic point of view, dynamically adjusting the user grouping method to ultimately achieve stable grouping is a genetic evolutionary process, in which different user groups constitute different elites in a generation, and the genes (or sub-individuals) of the elites are crossed to produce new offspring, and the parent generation and offspring are simultaneously subjected to preferential treatment to complete the natural selection mechanism, and finally a stable user grouping is obtained. Compared with the early heuristic method, the present invention has a greater advantage in performance; compared with the greedy method, its complexity is significantly reduced while the performance is similar. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A flow chart of a method for grouping users in a millimeter-wave massive MIMO system based on a genetic algorithm provided by the present invention.

[0044] Figure 2 It is a comparison diagram of sum rate simulation under ZF precoding of an embodiment of the present invention.

[0045] Figure 3 It is a comparison diagram of sum rate simulation under MMSE precoding of an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The technical solution provided by the present invention will be described in detail below in conjunction with specific embodiments. It should be understood that the following specific implementation methods are only used to illustrate the present invention and are not used to limit the scope of the present invention.

[0047] In order to achieve the above object, the present invention provides the following technical solutions:

[0048] A method for grouping users of a millimeter wave massive MIMO system based on a genetic algorithm, comprising:

[0049] S1: Establish a mathematical model for user grouping in mmWave massive MIMO systems;

[0050] S2: Maximizing the system and rate problem based on establishing groups;

[0051] S3: Use genetic algorithm to give a grouping algorithm for the problem.

[0052] Furthermore, the step S1 specifically includes the following sub-steps:

[0053] S1.1: Configure a millimeter-wave all-digital time-division duplex massive MIMO system with N=128-antenna ULA serving K single-antenna users. The digital precoding method considers ZF (Zero-forcing Beamforming, ZF) and MMSE (Minimum Mean Square Error, MMSE) group number G=4. Considering the actual three-sector layout of the base station, that is, each array only serves users in the front area of ​​about 120°, the range of user generation is limited to [1 / 6π, 5 / 6π];

[0054] S1.2: Establish a data transmission model, assuming that the downlink channel is H = [h1 T ,h2 T ,...,h K T ] T , where h k is the channel from N-antenna ULA to user k, which is modeled according to the Saleh-Valenzuela millimeter wave channel model:

[0055]

[0056] in is the complex gain of the l-th path of user k, which is composed of path loss and shadow fading. L is the number of multipaths. l = 0 represents the line-of-sight component, and l ≥ 1 represents the non-line-of-sight component. The path loss is:

[0057] PL(dB)=-35.4+26log 10 (d)+20log 10 (f c )

[0058] Where d is the distance from the user to the base station in meters, and f c = 30 GHz is the carrier frequency. Shadow fading has a standard deviation of 3 dB. is the array response vector of N-antenna ULA, expressed as:

[0059]

[0060] in, is the arrival angle of path l of user k, (·) T Represents transpose.

[0061] Let p k As its precoding vector, the data rate it receives is:

[0062]

[0063] Where p is the total power of the transmitted signal, set to p = 50dBm, σ 2 The noise variance is set to σ 2 =-174dBm / Hz, log represents logarithmic operation, |·| represents modulus operation;

[0064] S1.3: The present invention considers SDMA (such as ADMA) mixed with ideal TDMA. Different user groups serve different time domain or frequency domain resources to eliminate interference. SDMA technology is used in the same group of users. Therefore, it is necessary to reasonably group users to achieve weak correlation between channels of the same group of users to improve resource utilization efficiency. Considering that the user grouping method divides K users into G>1 groups on average, let is the entire set of users, then the grouping rule can be described as:

[0065]

[0066] S1.4: Grouping The rate of user k in is expressed as:

[0067]

[0068] Where p is the total power of the transmitted signal, h k Downlink channel of user k, σ 2 is the noise variance, then the group The sum rate is expressed as:

[0069]

[0070] Assume grouping The resource allocation coefficient in the orthogonal multiple access system is Then the system and rate can be expressed as:

[0071]

[0072] S2: Establish system maximization and rate problem:

[0073]

[0074] S3 includes the following steps:

[0075] S3.1: Set the number of iterations ITER, the number of elites E, the number of groups G, the number of offspring S, the number of crossovers Mut and the threshold γ;

[0076] S3.2: First, generate the initial parent generation, that is, for each elite e, randomly generate groups according to the number of groups G

[0077] S3.3: Start iteration, and perform the following operations in each iteration:

[0078] S3.3.1: First, generate S offspring for each parent elite e. The specific operations on the offspring s are as follows:

[0079] S3.3.1.1: First, let the offspring s be And randomly select the crossover number mut∈{1,2,...,Mut};

[0080] S3.3.1.2: In each crossover, the group number and the two users to be crossovered are randomly selected to complete the mutation;

[0081] S3.3.1.3: Calculate the sum rate R based on the grouping information of the offspring s;

[0082] S3.3.2: After completing the above operations for all parents, consider all the parents and their children and retain the E elites with the largest sum rate;

[0083] S3.4: If the optimization degree is less than the threshold value γ for multiple consecutive times, the iteration ends and the grouping method with the largest sum rate is the final grouping method. Otherwise, jump to S to perform the next iteration on the E elites obtained in the previous step.

[0084] Figure 2 , 3 (Using ZF and MMSE precoding respectively) is a comparison chart of the simulation results of this specific implementation method and the results of common algorithms. In the figure, the random grouping algorithm is used as the baseline for grouping algorithm comparison, and the semi-orthogonal grouping algorithm curve is the rate of the first group using the semi-orthogonal grouping algorithm, which is also the upper limit of the semi-orthogonal method. The rates of various grouping algorithms improved based on the semi-orthogonal grouping algorithm cannot exceed this limit.

[0085] Simulation experiments show that this algorithm performs well in both ZF and MMSE scenarios. In ZF, it is slightly lower than the semi-orthogonal grouping algorithm when the number of users is low, but since it is only the first group of results, the slight advantage is not of reference value. As the number of users increases, the performance of this algorithm gradually surpasses and suppresses the performance of other algorithms until it reaches the peak. The greedy grouping algorithm has partially surpassed it, and the three algorithms are significantly better than the random grouping algorithm. In addition, this algorithm and the greedy grouping algorithm have achieved great performance improvements; in MMSE, similar to the conclusion of ZF, it is slightly surpassed by the semi-orthogonal grouping algorithm in the first half, has a slight advantage over the greedy grouping algorithm near the peak, and is in a state of suppression for the semi-orthogonal grouping algorithm and the random grouping algorithm. After exceeding the peak, it is gradually surpassed by the greedy grouping algorithm.

Claims

1. A method for user grouping in a millimeter wave massive MIMO system based on genetic algorithm, characterized in that: The following steps are involved: S1: Establish a mathematical model for user grouping in mmWave massive MIMO systems; S2: Maximizing the system and rate problem based on establishing groups; S3: Using genetic algorithm to give a grouping algorithm for the problem; The step S1 specifically includes the following sub-steps: S1.1: Configure a mmWave all-digital time-division duplex massive MIMO system with N-antenna ULAs serving K single-antenna users, where the base station unit has learned the user's channel information through channel estimation; S1.2: Establish a data transmission model, assuming that the downlink channel is H = [h1 T ,h2 T ,...,h K T ] T , where h k is the channel from N-antenna ULA to user k, which is modeled according to the Saleh-Valenzuela millimeter wave channel model: in is the complex gain of the l-th path of user k, which is composed of path loss and shadow fading, L is the number of multipaths, l = 0 represents the line-of-sight component, l ≥ 1 represents the non-line-of-sight component, is the array response vector of N-antenna ULA, expressed as: in, is the arrival angle of path l of user k, (·) T stands for transpose; Let p k As its precoding vector, the received data rate is: Where p is the total power of the transmitted signal, σ 2 is the noise variance, log represents logarithmic operation, and |·| represents modulus operation; S1.3: Use orthogonal multiple access to divide K users into G>1 groups on average, and use SDMA to serve users in the same group. is the entire set of users, then the grouping rule is described as: S1.4: Grouping The rate of user k in is expressed as: Where p is the total power of the transmitted signal, h k Downlink channel of user k, σ 2 is the noise variance, then the group The sum rate is expressed as: Assume grouping The resource allocation coefficient in the orthogonal multiple access system is S g >0, Then the system and rate can be expressed as: The step S2 is specifically described as follows: Distribute resources evenly among groups. The problem of maximizing the system and rate is stated as: The step S3 specifically includes the following sub-steps: S3.1: Set the number of iterations ITER, the number of elites E, the number of groups G, the number of offspring S, the number of crossovers Mut and the threshold γ; S3.2: First, generate the initial parent generation, that is, for each elite e, randomly generate groups according to the number of groups G S3.3: Start iteration, and perform the following operations in each iteration: S3.3.1: First, generate S offspring for each parent elite e. The specific operations on the offspring s are as follows: S3.3.1.1: First, let the offspring s be And randomly select the crossover number mut∈{1,2,...,Mut}; S3.3.1.2: In each crossover, the group number and the two users to be crossovered are randomly selected to complete the mutation; S3.3.1.3: Calculate the sum rate R based on the grouping information of the offspring s; S3.3.2: After completing the above operations for all parents, consider all the parents and their children and retain the E elites with the largest sum rate; S3.4: If the optimization degree is less than the threshold value γ for multiple consecutive times, the iteration ends and the grouping method with the largest sum rate is the final grouping method. Otherwise, jump to S to perform the next iteration on the E elites obtained in the previous step.

2. The method for user grouping in a millimeter wave massive MIMO system based on genetic algorithm according to claim 1, characterized in that: The precoding vector p in step S1.2 k Generated by linear precoding methods MRT, ZF or MMSE.

3. The method for user grouping in a millimeter wave massive MIMO system based on genetic algorithm according to claim 1, characterized in that: The elites cross over within genes, that is, individuals are produced by allowing different groups within a gene to exchange members with each other.

4. The method for user grouping in a millimeter wave massive MIMO system based on genetic algorithm according to claim 1, characterized in that: When completing mutation, calculation and rate selection, in order to ensure that the excellent genes of the parents in the genetic algorithm can be retained and not eliminated, the parents and children are allowed to participate in the selection at the same time in the preferential processing, that is, the sum of the rates of the parents and children is uniformly sorted, and the top E elites with the largest sum of rates are retained.

Images