CN109547076A - Mixing precoding algorithms in the extensive MIMO of millimeter wave based on DSBO - Google Patents

Abstract

The invention discloses a DSBO-based hybrid precoding algorithm in millimeter-wave massive MIMO, which includes the following steps: first, the design problem of the precoding matrix is transformed into an optimization problem with the starting point of maximizing the reachability and rate of the system, and then using the proposed The DSBO intelligent search algorithm based on DSBO solves the optimization problem and obtains the precoding matrix that can maximize the reachability and rate of the system. The advantages of the present invention are: in the partial connection structure millimeter wave massive MIMO system, compared with the existing BSA-based hybrid precoding algorithm and pure analog precoding algorithm, the present invention can significantly improve the reachability, rate and rate of the system. Reduce the bit error rate of the system.

Description

Hybrid Precoding Algorithm Based on DSBO in Millimeter-Wave Massive MIMO

technical field

The present invention relates to the technical field of communication system precoding, in particular to a hybrid precoding algorithm of a partially connected structure millimeter wave massive MIMO system.

Background technique

The millimeter wave frequency band includes frequencies of 30-300GHz, which can obtain a high bandwidth, thereby greatly improving the transmission rate of the communication system, and has become one of the key technologies of 5G. The smaller wavelength of mmWave makes it possible to install a large number of antennas in a small aperture, and the resulting array gain can compensate for the path loss of mmWave. Taking advantage of this feature, mmWave massive MIMO systems use large antenna arrays at the base station. , thus allowing the transmission of multiple data streams, making the system have a higher data transmission rate. In the millimeter-wave massive MIMO system, the precoding technology uses the state information of the channel to adjust the transmission strategy at the transmitter and equalize at the receiver, so that the user can better obtain the antenna multiplexing gain and improve the system capacity.

In the digital precoding technology, the required number of radio frequency links is equal to the number of transmit antennas, resulting in high radio frequency link cost and hardware loss when it is applied in massive MIMO systems. To solve this problem, some literatures propose a pure analog precoding scheme. This scheme replaces the RF link in digital precoding with inexpensive, low-cost analog phase shifters. However, since the phase shifter can only control the phase of the transmitted signal, the performance of this scheme is much lower than that of digital precoding. In order to balance the hardware cost and system performance, some literatures propose a hybrid precoding scheme. The scheme performs low-dimensional digital precoding in the baseband, and is connected to the phase shifter through a small number of radio frequency links, which effectively reduces the hardware cost of the base station. The traditional hybrid precoding scheme is based on a fully connected structure, which can fully utilize the gain of the antenna, but requires a large number of phase shifters and adders at the same time, resulting in high system hardware cost and power consumption. In response to this problem, some scholars have studied a hybrid precoding scheme based on a partial connection structure, that is, each radio frequency link is fixedly connected to a limited number of transmitting antennas through a phase shifter, which greatly reduces the number of required phase shifters. And no adder is needed, which reduces hardware loss and implementation complexity. However, due to the limitation of this structure, the precoding matrix will be limited to the form of a block diagonal matrix, so the performance will be lower than that of the fully connected structure. The existing hybrid precoding algorithm of partial connection structure based on BSA (Bird Swarm Algorithm, bird flock algorithm) adopts the classic ZF (Zero-Forcing, zero-forcing) digital precoding algorithm at the base station, and uses BSA in the analog precoding part. The algorithm obtains the optimal analog precoding matrix, which reduces the design complexity, but the performance of the algorithm is poor. The SBO (satin bowerbird optimization) algorithm is a swarm intelligent search algorithm proposed by Seyyed Hamid Samareh Moosavi et al in 2017, inspired by the courtship behavior of satin blue gardener birds. Although the SBO algorithm has superior performance in solving high-dimensional complex function problems, the SBO algorithm It has the disadvantage of being easily trapped in local optimum. Therefore, it is urgent to invent a hybrid precoding algorithm for a partially connected millimeter-wave massive MIMO system based on DSBO (Satin Bowerbird Optimization Based on Dynamic Mutation Probability), so that the system can obtain a higher reachable sum rate and a lower bit error rate.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a hybrid precoding algorithm based on DSBO in millimeter-wave massive MIMO that enables the system to obtain higher reachable sum rate and lower bit error rate.

To achieve the above object, the present invention adopts the following technical scheme: a DSBO-based hybrid precoding algorithm in millimeter-wave massive MIMO, comprising the following steps:

Step (1): Transform the design problem of the optimal analog precoding matrix into the following optimization problem:

θ=argmintr(F _RF (θ) ^H H ^* H ^T F _RF (θ)) ^-1

st0<θ _j <2π,j=1,2,…,N _t

where H is the channel matrix, vector f _i ∈ C ^M×1 , i=1,2,..., elements in N N is the number of radio frequency chains, M is the number of transmit antennas connected to each radio frequency chain; F _RF is about The only change, the element θ _j =θ _i+(l-1)N =θ _il is the phase of the jth phase shifter, where j=1,2,...,N _t , i=1,2,..., N, l=1,2,...,M,N _t =MN is the number of transmit antennas of the millimeter-wave massive MIMO system, (·) ^H , (·) ^T , (·) ^-1 , (·) ^* respectively The conjugate transpose, transpose, inverse, and conjugate of the matrix, tr( ) represents the trace of the matrix, tr(F _RF (θ) ^H H ^* H ^T F _RF (θ)) ^-1 is the optimization problem the objective function;

Step (2): Initialize the parameters of the DSBO algorithm: the number of pavilions nPop=50, the length of the pavilion N _t , t is the number of iterations, set t=0 during initialization, and the maximum number of iterations MaxIt=250, the general expression defining the pavilion is: That is, each pavilion corresponds to a θ vector; the θ vector at the 0th iteration of nPop length N _t is randomly generated, denoted as Calculate the objective function value corresponding to each theta ⁰ vector which is

Step (3): use To calculate the fitness of nPop pavilions, the greater the fitness, the more attractive the pavilion is to female birds; when t<MaxIt, calculate fitness

in

Step (4): According to the calculation result of step (3), select the pavilion that most attracts female birds among the nPop pavilions in the current iteration, that is, the plan with the largest fitness, and take it as the current best plan, denoted as θ′, as Other males build gazebos to provide experience and the right direction;

Step (5): When building the pavilion, the male bird will improve his pavilion by imitating the best pavilion in history and the best pavilion at present. move Will Substitute into the following formula to get

in

in, respectively represent the current iteration The k-th element of and the new scheme after the change The k-th element of , θ _elite,k , θ _k ′ represent the k-th element of the historical best program θ _elite and the current best program θ′ respectively, prob′ represents the fitness value of the current best program θ′, α =0.94;

While males are busy building their gazebos, they may be attacked by other animals, so in an iterative process, it is necessary to A certain mutation is applied according to the probability. If rand<pMutation (rand represents a random number uniformly distributed in the interval [0,1]), then the Substitute into the following formula to calculate the scheme after applying mutation:

Where δ=z×(var _max -var _min );

Among them, var _max and var _min represent the upper limit and lower limit of each element in the vector θ, respectively, z=0.02, N(0,1) represents the random number obeying the standard normal distribution; pMutation is the mutation probability, and pMutation is set as the random number The variable for which the number of iterations t decreases nonlinearly, according to the following formula:

Among them, pMutation(t) represents the value of pMutation at the t-th iteration, pMutation _max and pMutation _min represent the maximum and minimum mutation probability, respectively, MaxIt is the maximum number of iterations of the algorithm, and exp represents an exponential function based on a natural constant;

Step (6): according to the objective function value, evaluate nPop existing schemes, keep the θ vector with better performance, and record as θ _elite ; when t<MaxIt, if but The number of iterations t=t+1, when the number of iterations is less than MaxIt, return to step (3), otherwise go to step (7);

Step (7): when the algorithm reaches the maximum number of iterations, that is, when t=MaxIt, the optimal vector θ _elite is output;

Step (8): generate an optimal analog precoding matrix F _RF from θ _elite obtained in step (7); that is, F _RF =diag{f ₁ ,f ₂ ,...,f _N }; where,

Further, in the aforementioned hybrid precoding algorithm based on DSBO in millimeter-wave massive MIMO, in step (5), var _max is taken as 2π, and var _min is taken as 0.

Further, in the aforementioned DSBO-based hybrid precoding algorithm in millimeter-wave massive MIMO, in step (5), pMutation _max is taken as 1/8, and pMutation _min is taken as 1/30.

Through the implementation of the above technical solutions, the beneficial effects of the present invention are: the present invention changes the mutation probability pMutation in the SBO algorithm from a fixed value to a variable that decreases nonlinearly with the number of iterations, and proposes the DSBO algorithm based on the dynamic mutation probability for the first time, and uses In order to solve the optimization problem in the present invention; at the same time, the function tr(F _RF ^H H ^* H ^T F _RF ) ^-1 is used as the objective function of the DSBO algorithm, and each pavilion corresponds to a scheme of the search object θ. The optimal vector θ generates an analog precoding matrix; so that in the partially connected structure millimeter wave massive MIMO system, compared with the existing BSA-based hybrid precoding algorithm and pure analog precoding algorithm, the present invention can significantly improve Improve system reachability and rate and reduce system bit error rate.

Description of drawings

Figure 1 shows a model of a millimeter-wave massive MIMO system with a partial connection structure.

FIG. 2 is a schematic diagram showing the system reachability and rate comparison of each precoding algorithm when the number of transmit antennas is 128.

FIG. 3 is a schematic diagram showing the system reachability and rate comparison of each precoding algorithm when the number of transmit antennas is 256.

FIG. 4 is a schematic diagram showing the comparison of the system reachability and rate of each precoding algorithm when the number of radio frequencies is different.

FIG. 5 is a schematic diagram showing the comparison of bit error rates of various precoding algorithms.

Detailed ways

The technical solutions of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

The DSBO-based hybrid precoding algorithm in millimeter-wave massive MIMO includes the following steps:

Step (1): Transform the design problem of the optimal analog precoding matrix into the following optimization problem:

θ=argmintr(F _RF (θ) ^H H ^* H ^T F _RF (θ)) ^-1

st0<θ _j <2π,j=1,2,…,N _t

where H is the channel matrix, vector f _i ∈ C ^M×1 , i=1,2,..., elements in N N is the number of radio frequency chains, M is the number of transmit antennas connected to each radio frequency chain; F _RF is about The only change, the element θ _j =θ _i+(l-1)N =θ _il is the phase of the jth phase shifter, where j=1,2,...,N _t , i=1,2,..., N, l=1,2,...,M,N _t =MN is the number of transmit antennas of the millimeter-wave massive MIMO system, (·) ^H , (·) ^T , (·) ^-1 , (·) ^* respectively The conjugate transpose, transpose, inverse, and conjugate of the matrix, tr( ) represents the trace of the matrix, tr(F _RF (θ) ^H H ^* H ^T F _RF (θ)) ^-1 is the optimization problem the objective function;

In this embodiment, the specific conversion process for converting the design problem of the optimal analog precoding matrix into an optimization problem is:

(1) System model and channel model

The present invention considers the hybrid precoding problem in the downlink multi-user millimeter wave massive MIMO system, and the base station adopts a partial connection mode, as shown in Figure 1; the transmitter is configured with N radio frequency links, and each radio frequency link is fixedly connected to M For the phase shifter, the number of transmit antennas is N _t =MN, the transmit data stream is N _s , and the number of receive antennas is K. The transmitter first performs digital precoding on N _s data streams through the digital precoder, and then transmits it to the analog precoding part through the radio frequency link for analog precoding. After the analog precoding is completed, the data is mapped to the transmitting antenna and sent to the transmitter. Receiving antenna; in this system, the user's received signal y∈C ^K×1 can be expressed as follows:

y=HF _RF F _BB s+n (1)

in, In order to transmit signals, N _s ≤ N; H=[h ₁ , h ₂ ,...,h _K ] ^T ∈ C ^K×MN is the channel matrix, here it is assumed that the H matrix has passed the user side downlink channel estimation and channel state feedback process Obtained; F _RF ∈ C ^MN×N is an analog precoding matrix in the form of F _RF =diag{f ₁ ,f ₂ ,…,f _N }, where f _n ∈ C ^M×1 , n=1,2,… ,N, f _ni represents the ith element in the vector f _n , i=1,2,...,M; is the digital precoding matrix, where In the present invention, F _BB is obtained by ZF precoding algorithm; n∈C ^K×1 represents additive white Gaussian noise, that is, n～CN(0,σ ² I _K ), σ ² represents variance, and I _K represents K-order unit matrix. F _RF and F _BB satisfy the total transmit power constraint, i.e.

Due to the limited spatial selectivity of the millimeter-wave frequency band due to the higher path loss and the high antenna correlation caused by the closely-arranged antenna array, the traditional channel model is no longer suitable for the millimeter-wave massive MIMO system; the present invention considers The extended Saleh-Valenzuela model commonly used in millimeter-wave massive MIMO systems captures the characteristics of millimeter-wave channels well. The channel matrix is as follows:

Among them, N _t is the number of transmitting antennas of the base station, K is the number of receiving antennas, L is the number of millimeter-wave scattering beam paths, δ _i is the gain of the i-th scattering beam path, θ _i ∈ [0, 2π] represents the departure angle and arrival angle of the i-th path, respectively, α _BS (θ _i ) represents the user antenna array response vector and the base station antenna array response vector respectively; the form of these two expressions is determined by the distribution mode of the antenna array; common antenna arrays include uniform linear arrays and uniform planar arrays; the present invention Using a uniform linear array, α _BS (θ _i ) is expressed as follows:

Where λ represents the wavelength of the electromagnetic wave, and d represents the distance between the antennas.

(2) Obtaining the objective function

The reachable sum rate of K users in the hybrid precoding system of the present invention can be represented by the following formula:

Among them, γ _k is the signal-to-interference-noise ratio of the kth user, which is expressed as follows:

Since the millimeter-wave massive MIMO system in the present invention is based on a partial connection structure, the analog precoding matrix will be limited to the form of a block-diagonal matrix, as shown below:

In the formula, the vector f _i ∈ C ^M×1 , i=1,2,…, elements in N It can be seen that the analog precoding matrix F _RF is a random vector Uniquely varying matrix with elements θ _j = θ _i+(l-1)N = θ _il , j=1,2,...,N _t , i=1,2,...,N,l=1,2 ,...,M is the phase of the jth phase shifter; the baseband digital precoding of the present invention adopts the classical ZF precoding algorithm, in this case, the reachable sum rate of the system can be expressed as the following formula:

Among them, R represents the reachable sum rate of the system, and snr _k represents the signal-to-noise ratio of the kth user; it can be seen from this formula that the reachable sum rate of the system is at most equivalent to tr(F _RF ^H H ^* H ^T F _RF ) ^-1 The minimum value is reached, so we take tr(F _RF ^H H ^* H ^T F _RF ) ^-1 as the objective function of the algorithm of the present invention, and the design problem of hybrid precoding is equivalent to the optimization problem expressed by the following formula:

Step (2): Initialize the parameters of the DSBO algorithm: the number of pavilions nPop=50, the length of the pavilion N _t , t is the number of iterations, set t=0 during initialization, and the maximum number of iterations MaxIt=250, the general expression defining the pavilion is: That is, each pavilion corresponds to a θ vector; the θ vector at the 0th iteration of nPop length N _t is randomly generated, denoted as Calculate the objective function value corresponding to each theta ⁰ vector which is

Step (3): use To calculate the fitness of nPop pavilions, the greater the fitness, the more attractive the pavilion is to female birds; when t<MaxIt, calculate fitness

in

Step (4): According to the calculation result of step (3), select the pavilion that most attracts female birds among the nPop pavilions in the current iteration, that is, the plan with the largest fitness, and take it as the current best plan, denoted as θ′, as Other males build gazebos to provide experience and the right direction;

Step (5): When building the pavilion, the male bird will improve his pavilion by imitating the best pavilion in history and the best pavilion at present. move Will Substitute into the following formula to get

in

in, respectively represent the current iteration The k-th element of and the new scheme after the change The k-th element of , θ _elite,k , θ _k ′ represent the k-th element of the historical best program θ _elite and the current best program θ′ respectively, prob′ represents the fitness value of the current best program θ′, α =0.94;

While males are busy building their gazebos, they may be attacked by other animals, so in an iterative process, it is necessary to A certain mutation is applied according to the probability. If rand<pMutation (rand represents a random number uniformly distributed in the interval [0,1]), then the Substitute into the following formula to calculate the scheme after applying mutation:

Where δ=z×(var _max -var _min );

Among them, var _max and var _min respectively represent the upper limit and lower limit of each element in the vector θ. In this embodiment, var _max is taken as 2π, var _min is taken as 0, z=0.02, and N(0,1) means obeying A random number from a standard normal distribution; pMutation is the mutation probability, and pMutation is set to a variable that decreases nonlinearly with the number of iterations t, according to the following formula:

Among them, pMutation(t) represents the value of pMutation at the t-th iteration, and pMutation _max and pMutation _min represent the maximum and minimum mutation probability respectively. In this embodiment, pMutation _max is taken as 1/8, and pMutation _min is taken as 1/ 30, MaxIt is the maximum number of iterations of the algorithm, and exp represents an exponential function based on a natural constant;

Step (6): according to the objective function value, evaluate nPop existing schemes, keep the θ vector with better performance, and record as θ _elite ; when t<MaxIt, if but The number of iterations t=t+1, when the number of iterations is less than MaxIt, return to step (3), otherwise go to step (7);

Step (7): when the algorithm reaches the maximum number of iterations, that is, when t=MaxIt, the optimal vector θ _elite is output;

Step (8): generate an optimal analog precoding matrix F _RF from θ _elite obtained in step (7); that is, F _RF =diag{f ₁ ,f ₂ ,...,f _N }; where,

The advantages of the present invention are: the present invention changes the mutation probability pMutation in the SBO algorithm from a fixed value to a variable that decreases nonlinearly with the number of iterations, and proposes the DSBO algorithm based on dynamic mutation probability for the first time, and is used to solve the optimization in the present invention At the same time, the function tr(F _RF ^H H ^* H ^T F _RF ) ^-1 is used as the objective function of the DSBO algorithm, each pavilion corresponds to a scheme of the search object θ, and finally the searched optimal vector θ is used to generate a simulation precoding matrix; so that in the partially connected millimeter wave massive MIMO system, compared with the existing BSA-based hybrid precoding algorithm and pure analog precoding algorithm, the present invention can significantly improve the system reachability and rate and reduce The bit error rate of the system.

The beneficial effects of this hybrid precoding algorithm on improving the system capacity and bit error rate performance of the partially connected millimeter-wave massive MIMO system are described below through simulation. It enables the partially connected millimeter wave massive MIMO system to have higher system reachability and lower bit error rate.

Simulation results:

For a millimeter-wave massive MIMO system with a partially connected structure in a single cell, the algorithm of the present invention is compared with the hybrid precoding algorithm based on BSA and the pure analog precoding algorithm. The simulation parameters are: the number of scatterers of the millimeter wave beam L=10, the base station transmit power is 1w, the antenna spacing d=λ/2, λ is the millimeter wave wavelength, and the modulation method adopts 4QAM modulation.

Figure 2 and Figure 3 respectively show the performance comparison of each precoding algorithm system reachability and rate when the system is N _t = MN = 128, N = K = 32 and N _t = MN = 256, N = K = 32 . It can be seen from the two figures that the trends in Figure 2 and Figure 3 are similar, that is, the performance of each algorithm increases with the increase of the signal-to-noise ratio and the number of transmit antennas, and the performance of the DSBO algorithm of the present invention is much better than the hybrid precoding based on BSA. The performance of the algorithm and the pure analog precoding algorithm; when the number of transmit antennas increases to 256, the performance advantage of the DSBO algorithm of the present invention is more obvious.

Figure 4 shows the relationship between the reachable sum rate of each precoding algorithm and the number of radio frequency links when the system is N _t = MN = 200, K = 10, and the transmit signal-to-noise ratio snr = 10dB. It can be seen that the system reachability and rate of the three partial connection structure precoding algorithms all increase with the increase of the number of radio frequencies, and finally approach the performance of the pure digital ZF precoding algorithm. In addition, the performance of the DSBO algorithm proposed in the present invention is better than that of the hybrid precoding algorithm based on BSA.

Fig. 5 shows the relationship between the systematic bit error rate of each precoding algorithm and the signal-to-noise ratio when the system is N _t =MN=128 and N=K=32. It can be seen from the figure that the system bit error rate performance of the DSBO algorithm of the present invention is better than that of the hybrid precoding algorithm based on BSA, the gain is about 2.5dB, and the bit error rate is much lower than that of the pure analog precoding algorithm.

In terms of complexity, the complexity of the DSBO algorithm of the present invention mainly comes from The calculation of this formula will be calculated twice in each iteration of the algorithm, that is, the calculation and Therefore, when the algorithm reaches the maximum number of iterations MaxIt, the total complexity can be expressed as O(2nPopMaxItN _t N ² ). Among them, N is the number of radio frequency links. The complexity of the hybrid precoding algorithm based on BSA can be expressed as O(popTN _t N ² ), where pop and T represent the initial population number and the maximum number of iterations of the BSA algorithm, respectively. It can be seen that the complexity of the algorithm of the present invention is the same as that of the BSA algorithm. The algorithmic complexity is in the same order of magnitude.

Claims (3)

1. The mixed precoding algorithm based on DSBO in millimeter wave large-scale MIMO is characterized in that: the method comprises the following steps:

step (1): converting the design problem of the optimal simulation precoding matrix into the following optimization problem:

θ＝argmintr(F_RF(θ)^HH^*H^TF_RF(θ))^-1

s.t.0<θ_j＜2π,j＝1,2,…,N_t

wherein, H is a channel matrix,vector f_i∈C^M×1I is an element of 1,2, …, NN is the number of radio frequency links, and M is the number of transmitting antennas connected with each radio frequency link; f_RFAboutThe only change, element θ_j＝θ_i+(l-1)N＝θ_ilIs the phase of the jth phase shifter, where j is 1,2_t，i＝1,2,…,N，l＝1,2,…,M，N_tMN is the number of millimeter wave large-scale MIMO system transmitting antennas, (·)^H、(·)^T、(·)^-1、(·)^*Respectively representing the conjugate transpose, inverse and conjugate of the fetch matrix, tr (-) represents the trace of the fetch matrix, tr (F)_RF(θ)^HH^*H^TF_RF(θ))^-1An objective function for the optimization problem;

step (2): initializing DSBO algorithm parameters: pavilion number nPop 50, pavilion length N_tT is iteration number, when initializing, t is set to 0, and the maximum iteration number maxti is 250, and a general expression of the wayside pavilion is defined as follows:namely, each pavilion corresponds to one theta vector; randomly generating nPop pieces with length of N_tThe theta vector at iteration 0 of (2), is notedCalculate each theta⁰Objective function value corresponding to vectorNamely, it is

And (3): by usingCalculating the fitness of the nPop kiosks, wherein the larger the fitness is, the larger the attraction of the kiosks to female birds is; when t is<MaxIt time, calculateIs adapted to

Wherein

And (4): selecting the pavilion which attracts the female birds most from the nPop pavilions of the current iteration, namely the scheme with the maximum fitness, according to the calculation result of the step (3), and taking the pavilion as the current optimal scheme, namely marking the scheme as theta' to provide experience and correct direction for constructing the pavilions by other male birds;

and (5): when the pavilion is built, the male bird can imitate the historical best pavilion and the current best pavilion to improve the pavilion, and according to the characteristic, the male bird moves in the direction of the recorded historical best scheme and the current best schemeWill be provided withSubstituted into the formula to obtain

Wherein

Wherein,respectively represent the current iterationThe kth element of (1) and the new scheme after the changeThe k-th element of (a), theta_elite,k、θ_k' respective representation of the historical best solution θ_eliteAnd the kth element of the current best solution θ ', prob ' represents the fitness value of the current best solution θ ', α ═ 0.94;

while the males are busy building the pavilion, they may be attacked by other animals, and thus, in an iterative process, it is necessary to do soApplying a certain mutation by probability if rand<pMutition (rand denotes the interval [0,1 ]]Internally uniformly distributed random numbers), then will beSubstituting the following calculation to obtain the scheme after mutation application:

wherein δ is zx (var)_max-var_min)；

Wherein, var_max、var_minRespectively representing the upper limit and the lower limit of each element in the vector theta, z is 0.02, and N (0,1) represents a random number which follows a standard normal distribution; pMutation is the mutation probability, pThe Mutation is set to a variable that decreases non-linearly with the number of iterations t, following the following equation:

wherein, pMutation (t) represents the value of pMutation at the t iteration, pMutation_max、pMutation_minRespectively representing maximum mutation probability and minimum mutation probability, wherein MaxIt is the maximum iteration number of the algorithm, and exp represents an exponential function with a natural constant as a base;

and (6): evaluating the nPop acquired schemes according to the objective function values, reserving the theta vector with better performance, and recording the theta vector as theta_elite(ii) a When t < MaxIt,if it is notThenWhen the iteration time t is equal to t +1, returning to the step (3) when the iteration time is less than MaxIt, otherwise, entering the step (7);

and (7): when the algorithm reaches the maximum iteration number, namely t is MaxIt, the optimal vector theta is output_elite；

And (8): theta obtained from step (7)_eliteGenerating an optimal analog precoding matrix F_RF(ii) a I.e. F_RF＝diag{f₁,f₂,…,f_N}; wherein,

2. the DSBO-based hybrid precoding algorithm in millimeter wave massive MIMO according to claim 1, wherein: in step (5), var_maxIs taken to be 2 pi, var_minTake 0.

3. The DSBO-based hybrid precoding algorithm in millimeter wave massive MIMO according to claim 1, wherein: in step (5), pMutition_maxTaken as 1/8, pMutition_minTaken as 1/30.