Large-scale MIMO hybrid precoding method based on improved particle swarm optimization algorithm

Abstract

The invention provides a large-scale MIMO (Multiple-Input Multiple-Output) hybrid precoding method based on an improved particle swarm optimization algorithm, relates to a hybrid precoding design method, and aims at realizing high-speed communication in a large-scale MIMO wireless communication system, so that the invention aims at a large-scale MIMO system to maximize the spectrum efficiency of the system, and aims at solving the problem that the traditional particle swarm optimization algorithm is required to realize high-speed communication in the large-scale MIMO wireless communication system. The phase tracking method based on the improved particle swarm optimization (IPSO) algorithm is provided by improving a PSO (Particle Swarm Optimization) algorithm and adopting homogenized population distribution and a self-adaptive inertia weight and a speed factor on this basis, the method converts an analog precoding matrix recovery problem into an unconstrained phase tracking problem, and on the basis of the above process, the phase tracking method based on the improved particle swarm optimization (IPSO) algorithm is provided. According to the invention, the system performance index is obviously improved.

Description

Technical Field

[0001] The present invention relates to a hybrid precoding design method, and is particularly applicable to the technical field of millimeter wave wireless communication systems, and specifically is a large-scale MIMO hybrid precoding method based on an improved particle swarm optimization algorithm. Background Art

[0002] With the growing demand for wireless communications, the millimeter wave band (30GHz-300GHz) has attracted widespread attention as an emerging frequency resource. Compared with the currently crowded low-frequency bands, the millimeter wave band provides a large amount of undeveloped spectrum resources that can support higher data rates and capacities. However, millimeter wave communications also face some unique challenges, such as poor diffraction capability and severe path loss, which bring new problems to system design. An important technology in millimeter wave communications is large-scale MIMO technology. Because the millimeter wave wavelength is short, many antennas can be deployed in a small space to form high-gain directional beams, which can concentrate energy to compensate for the loss on the transmission path. Large-scale MIMO can significantly improve spectrum efficiency through spatial division multiplexing. As the number of antennas increases, the capacity of the channel system is greatly improved.

[0003] The traditional pure digital precoding method completes all precoding calculations in the baseband processing link, and needs to configure an independent RF link for each antenna. Although the hardware architecture of this method is simple and can achieve the theoretically optimal system performance parameters, the system power consumption is large, the hardware cost is high, and the implementation complexity is also high. In contrast, the pure analog precoding technology mainly uses a phase shift network to beamform the antenna array at the RF front end, and only a small number of RF links are required to connect the baseband processor. Therefore, the hardware complexity and power consumption of pure analog precoding are relatively low. The performance and flexibility of pure analog precoding technology are seriously affected. In order to strike a balance between performance and complexity and achieve a signal transmission efficiency closer to the optimal digital precoding technology, the hybrid precoding technology combines the above two methods. The hybrid precoding technology uses digital precoding in the baseband part and analog precoding in the RF part. However, the spectrum efficiency of hybrid precoding is currently low and does not meet the requirements of the development of communication technology. Summary of the invention

[0004] Purpose of the invention: To address the shortcomings of the above-mentioned large-scale MIMO precoding method based on millimeter waves, a large-scale MIMO hybrid precoding method is provided to effectively improve the spectrum efficiency.

[0005] Technical solution: To achieve the above-mentioned invention object, the present invention provides a large-scale MIMO hybrid precoding method based on an improved particle swarm optimization algorithm, and the specific steps are as follows:

[0006] Step 1: Create the Saleh-Valenzuela channel, assuming that there are N downlink cl There are N scattering clusters, each of which consists of N ray The narrowband millimeter wave channel is H, and uniform planar array (UPA) is used at both the transmitting and receiving ends.

[0007] Step 2: Determine the optimization target of the system. After the signal passes through the hybrid precoding and narrowband block fading channel from the transmitter, the received signal generated is: The received signal then passes through the receiving end mixer

[0008] It can be expressed as:

[0009]

[0010] Where n is noise, s is data flow, F RF F BB represents the analog and digital encoders, ρ is the received power and the spectral efficiency of the millimeter wave communication system is taken as the optimization target;

[0011] Step 3: Decouple the system optimization target into the receiving end and the transmitting end. Taking the transmitting end as an example, after performing singular value decomposition on the channel, the optimal precoding matrix is ​​its right singular value The problem is simplified to a constrained two-variable problem;

[0012] Step 4: Improve the particle swarm optimization algorithm, use the alternating iteration method to find the optimal values ​​of analog precoding and digital precoding respectively, use the least square method to fix the analog precoding, and solve the digital precoding. Fix the digital precoding, and use the improved particle swarm optimization algorithm to track the phase of the analog precoding matrix;

[0013] Step 5: Recover the analog precoding and digital precoding matrices according to system constraints;

[0014] Further, under specific channels and antenna arrays:

[0015]

[0016] In the formula, N cl scattering clusters, each containing N ray Transmission paths, N t 、N r Represents the number of antennas at the transmitting end and the receiving end respectively. ij represents the fading coefficient of the jth transmission path in the i-th scattering cluster, satisfying the power The complex Gaussian distribution of Represent the azimuth and elevation angles of the receiving and transmitting ends respectively, denote the normalized antenna response vectors of azimuth and elevation at the receiving and transmitting ends, respectively.

[0017]

[0018] The antenna array size N = WQ, W is the number of array elements deployed on the y-axis, Q is the number of array elements deployed on the z-axis, λ is the millimeter wave wavelength, d is the interval between array elements, φ and θ represent the azimuth and elevation angles, respectively, and m and n are the position indexes of the array elements in the array plane, satisfying 0≤m≤W, 0≤n≤Q.

[0019] Further steps 2 and 3 are as follows:

[0020] The optimization goal of the system, maximizing spectrum efficiency, is:

[0021]

[0022] After separating the transmitter and the receiver, the optimal precoding matrix F can be obtained by performing singular value decomposition on the channel matrix. opt , and then the optimization goal of the system becomes:

[0023]

[0024] The elements of the analog precoding matrix satisfy the constant modulus constraint.

[0025] The specific steps for optimizing the particle swarm are as follows:

[0026] The present invention uses logistic mapping to initialize the population, and the specific operation is as follows:

[0027]

[0028] Where α is a constant, ranging from 0 to 4. Represent the positions of particles at the current and next moments respectively. Before the improvement, the distribution of the population was often too concentrated in certain places and relatively sparse in the edge areas, which may cause the algorithm to fall into the local optimum and the search convergence speed will also slow down. After the improvement, the population distribution has become more uniform and exists in almost all areas.

[0029] The present invention adopts adaptive coefficients. In the early stage of iteration, a larger coefficient is configured to accelerate the global search. In the later stage of iteration, it is changed to a smaller coefficient to accelerate the convergence speed. The present invention processes the inertia weight coefficient as follows:

[0030] In this formula, wmax and w min Represents the upper and lower limits of the inertia coefficient, t max Represents the maximum number of iterations. The improved inertia coefficient presents a convex function distribution at the beginning of the iteration and a concave function distribution in the later stage.

[0031] The speed factors c1 and c2 represent the impact of the individual optimal position and the population optimal position on the particle speed, respectively. Traditional PSO usually sets them as fixed values. In order to expand the search range in the initial stage, so that particles can find the optimal solution in the entire space, and pay more attention to the convergence speed in the later stage, this paper proposes an adaptive adjustment method: in the early stage, c1 should take a larger value; in the later stage, take a smaller value, and c2 is the opposite.

[0032]

[0033] Furthermore, the steps for improving the particle swarm tracking simulation precoding matrix phase are as follows:

[0034] Step 1: By fixing F RF To solve F BB , here we use the least squares method to solve F BB ;

[0035] Step 2: Fix F BB , for F RF Use Euler's formula instead Initialize the necessary parameters of the population, use the position information of the particles to replace the angle information of the simulated precoding, and calculate the fitness of each population;

[0036] Step 3: According to the improved speed update and position update formula, adjust the fitness of each particle in the population, and use this method to determine the optimal position of each particle and the optimal position of the population;

[0037] Step 4: Repeat step 3) until the maximum number of iterations is reached or the current threshold is less than the previous threshold or the initial given threshold, then the iteration ends;

[0038] Step 5: Restore the optimal position of the final population to F RF Phase information of

[0039] The present invention proposes a large-scale MIMO hybrid precoding method based on an improved particle swarm optimization algorithm, and the beneficial effects are as follows:

[0040] 1. The invention extracts the elements of the analog precoding matrix as corresponding phases, and transforms the original constant modulus constrained analog precoding matrix recovery problem into an unconstrained phase tracking problem, which is beneficial to improving the system spectrum efficiency.

[0041] 2. The use of adaptive weight coefficients, adaptive learning factors, and chaotic population initialization greatly reduces the computational complexity and accelerates the algorithm convergence speed compared to the traditional particle swarm optimization algorithm.

[0042] 3. The accuracy of the method of the present invention is higher than that of the other three traditional solutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 Flowchart of precoding design method;

[0044] Figure 2 System model diagram;

[0045] Figure 3 N t =64,N r =64, The spectrum efficiency comparison chart below;

[0046] Figure 4 N t =124,N r =32, The following is a spectrum efficiency comparison chart. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention will be described in detail below with reference to the accompanying drawings;

[0048] Reference Manual Attached Figure 1 To Attachment Figure 4 First, the practical application scenario of the present invention is introduced: a point-to-point fully connected large-scale MIMO millimeter wave communication system is considered, and the transmitting end is equipped with RF links, and the receiving end is equipped with RF links to meet The hybrid precoder consists of a digital encoder F BB and analog encoder F RF Combining, where The analog encoder only performs beamforming and does not change the signal energy, so the modulus value of each element in its precoding matrix is ​​the same, that is, |[F RF ] ij |=1. In order to meet the power constraint, the hybrid precoding matrix needs to satisfy This design can achieve efficient and stable multi-stream transmission in complex wireless communication environments.

[0049] To achieve the above-mentioned object of the invention, a large-scale MIMO hybrid precoding method based on an improved particle swarm optimization algorithm is provided, and the specific steps are as follows:

[0050] Step 1: Create the Saleh-Valenzuela channel model, assuming that there are N cl There are N scattering clusters, each of which consists of N ray The present invention adopts a uniform planar array (UPA) at both the transmitting end and the receiving end;

[0051] Step 2: Determine the optimization target of the system. After the signal passes through the hybrid precoding and narrowband block fading channel from the transmitter, the received signal generated is:

[0052]

[0053] Where n is noise, s is data flow, F RF F BB represents the analog and digital encoder, and ρ is the received power. The signal passes through the receiving end hybrid combiner W RF W BB After that, it can be expressed as:

[0054]

[0055] Taking the spectrum efficiency of the millimeter wave communication system as an optimization target;

[0056] Step 3: Decouple the system optimization target into the receiving end and the transmitting end. Taking the transmitting end as an example, after performing singular value decomposition on the channel, the optimal precoding matrix is ​​its right singular value The problem is simplified to a constrained two-variable problem;

[0057] Step 4: Improve the particle swarm optimization algorithm, use the alternating iteration method to find the optimal values ​​of analog precoding and digital precoding respectively, use the least square method to fix the analog precoding, and solve the digital precoding. Fix the digital precoding, and use the improved particle swarm optimization algorithm to track the phase of the analog precoding matrix;

[0058] Step 5: Restore the analog precoding and digital precoding matrices according to system constraints.

[0059] Furthermore, the specific channels and antenna arrays are as follows:

[0060]

[0061] In the formula, N cl scattering clusters, each containing N ray Transmission paths, N t 、N r Represents the number of antennas at the transmitting end and the receiving end respectively. ijrepresents the fading coefficient of the jth transmission path in the i-th scattering cluster, satisfying the power The complex Gaussian distribution of Represent the azimuth and elevation angles of the receiving and transmitting ends respectively, denote the normalized antenna response vectors of azimuth and elevation at the receiving and transmitting ends, respectively.

[0062]

[0063] The antenna array size N = WQ, W is the number of array elements deployed on the y-axis, Q is the number of array elements deployed on the z-axis, λ is the millimeter wave wavelength, d is the interval between array elements, φ and θ represent the azimuth and elevation angles, respectively, and m and n are the position indexes of the array elements in the array plane, satisfying 0≤m≤W, 0≤n≤Q.

[0064] Further steps 2 and 3 are as follows:

[0065] The optimization goal of the system, maximizing spectrum efficiency, is:

[0066]

[0067] After separating the transmitter and the receiver, the optimal precoding matrix F can be obtained by performing singular value decomposition on the channel matrix. opt , and then the optimization goal of the system becomes:

[0068]

[0069] The elements of the analog precoding matrix satisfy the constant modulus constraint.

[0070] The specific steps for optimizing the particle swarm are as follows:

[0071] In the traditional PSO algorithm, the population is initialized randomly according to the requirements of the system model. The present invention uses logistic mapping to initialize the population. The specific operation is as follows:

[0072]

[0073] Where a is a constant, ranging from 0 to 4. Represent the positions of particles at the current and next moments respectively. Before the improvement, the distribution of the population was often too concentrated in certain places and relatively sparse in the edge areas, which may cause the algorithm to fall into the local optimum and the search convergence speed will also slow down. After the improvement, the population distribution has become more uniform and exists in almost all areas.

[0074] The present invention adopts adaptive coefficients. In the early stage of iteration, a larger coefficient is configured to accelerate the global search. In the later stage of iteration, it is changed to a smaller coefficient to accelerate the convergence speed. The present invention processes the inertia weight coefficient as follows:

[0075] In this formula, w max and w min Represents the upper and lower limits of the inertia coefficient, t max Represents the maximum number of iterations. The improved inertia coefficient presents a convex function distribution at the beginning of the iteration and a concave function distribution in the later stage.

[0076] The speed factors c1 and c2 represent the impact of the individual optimal position and the population optimal position on the particle speed, respectively. Traditional PSO usually sets them as fixed values. In order to expand the search range in the initial stage, so that particles can find the optimal solution in the entire space, and pay more attention to the convergence speed in the later stage, this paper proposes an adaptive adjustment method: in the early stage, c1 should take a larger value; in the later stage, take a smaller value, and c2 is the opposite.

[0077]

[0078] Furthermore, the steps for improving the particle swarm tracking simulation precoding matrix phase are as follows:

[0079] Step 1: By fixing F RF To solve F BB , here we use the least squares method to solve F BB ;

[0080] Step 2: Fix F BB , for F RF Use Euler's formula instead Initialize the necessary parameters of the population, use the position information of the particles to replace the angle information of the simulated precoding, and calculate the fitness of each population;

[0081] Step 3: According to the improved speed update and position update formula, adjust the fitness of each particle in the population, and use this method to determine the optimal position of each particle and the optimal position of the population;

[0082] Step 4: Repeat step 3) until the maximum number of iterations is reached or the current threshold is less than the previous threshold or the initial given threshold, then the iteration ends;

[0083] Step 5: Restore the optimal position of the final population to F RF phase information.

[0084] It should be noted that those skilled in the art should be aware that, based on the principles of the present invention, it is not necessary to list all implementation methods, and obvious changes or modifications based on this are still within the protection scope of the present invention.

[0085] Figure 3 and Figure 4 It shows that: population size n = 20, maximum number of iterations t max =300, maximum inertia weight w max =0.9, minimum inertia weight w min =0.6, random factors r1=rand(1), r2=rand(1). Data stream N transmitted by the system s =3, both the transmitting and receiving ends are equipped with UPA antenna arrays. The performance of the following algorithms under different signal-to-noise ratios (SNRs) is compared: OMP algorithm, MO algorithm, PSO algorithm before improvement, IPSO algorithm after improvement, and optimal unconstrained OPT algorithm.

[0086] It can be seen that the algorithm of the present invention is superior to the compared algorithms in many cases.

[0087] It should be noted that those skilled in the art should be aware that, based on the principles of the present invention, it is not necessary to list all implementation methods, and obvious changes or modifications based on this are still within the protection scope of the present invention.

Claims

1. A large-scale MIMO hybrid precoding method based on an improved particle swarm optimization algorithm, wherein the specific steps of MIMO hybrid precoding are as follows: Step 1: Create the Saleh-Valenzuela channel, assuming that there are N downlink cl There are N scattering clusters, each of which consists of N ray The narrowband millimeter wave channel is H, and uniform planar array (UPA) is used at both the transmitting and receiving ends. Step 2: Determine the optimization target of the system. After the signal passes through the hybrid precoding and narrowband block fading channel from the transmitter, the received signal generated is: Where n is noise, s is data flow, F RF F BB represents the analog and digital encoder, ρ is the received power, and the received signal is then passed through the receiving end hybrid combiner It can be expressed as: Taking the spectrum efficiency of the millimeter wave communication system as an optimization target; Step 3: Decouple the sender and receiver for system optimization goals; Step 4: Improve the particle swarm optimization algorithm, use the alternating iteration method to find the optimal values ​​of analog precoding and digital precoding respectively, use the least square method to fix the analog precoding, solve the digital precoding, fix the digital precoding, and use the improved particle swarm optimization algorithm to track the phase of the analog precoding matrix; Step 5: Restore the analog precoding and digital precoding matrices according to system constraints.

2. The large-scale MIMO hybrid precoding method based on an improved particle swarm optimization algorithm according to claim 1, characterized in that: The specific channels and antenna arrays are as follows: In the formula, N cl scattering clusters, each of which contains N ray Transmission paths, N t 、N r Respectively represent the number of antennas at the transmitting end and the receiving end, α ij represents the fading coefficient of the jth transmission path in the i-th scattering cluster, satisfying the power The complex Gaussian distribution of and Represent the azimuth and elevation angles of the receiving and transmitting ends respectively, and denote the normalized antenna response vectors of azimuth and elevation at the receiving and transmitting ends, respectively; The antenna array size N = WQ, W is the number of array elements deployed on the y-axis, Q is the number of array elements deployed on the z-axis, λ is the millimeter wave wavelength, d is the interval between array elements, φ and θ represent the azimuth and elevation angles, respectively, and m and n are the position indexes of the array elements in the array plane, respectively, satisfying 0≤m≤W, 0≤n≤Q.

3. The large-scale MIMO hybrid precoding method based on the improved particle swarm optimization algorithm according to claim 2, characterized in that: Steps 2 and 3 are as follows: The optimization goal of the system, maximizing spectrum efficiency, is: The channel matrix singular value decomposition is performed, and then the optimal precoding matrix F is obtained by decoupling the transmitter and the receiver. opt , the optimization goal of the system also changes to The elements of the analog precoding matrix must follow the constant modulus constraint.

4. The large-scale MIMO hybrid precoding method based on an improved particle swarm optimization algorithm according to claim 1 or 3, characterized in that: The specific steps for optimizing the particle swarm are as follows: In the particle swarm algorithm, the formula for particle update is: and Represent the positions of particles at the current and next moments respectively, Indicates the speed update at the next moment, where the speed update formula is: In the formula represents the current velocity direction of the dth dimension of the i-th particle, Indicates the updated speed direction, parameter w indicates the inertia coefficient, c1 and c2 indicate the speed factors, r1 and r2 are random coefficients, usually randomly taking values ​​between 0 and 1. represents the optimal position of the current particle i in the dth dimension, pg d (t) represents the optimal position of all particles in the dth dimension at present; In the traditional PSO algorithm, the population is initialized randomly according to the requirements of the system model, and the logistic mapping is used to initialize the population. The specific operations are as follows: Where a is a constant, ranging from 0 to 4; Adopt adaptive coefficients. In the early stage of iteration, configure larger coefficients to speed up global search. In the later stage of iteration, change them to smaller coefficients to speed up convergence. The inertia weight coefficients are processed as follows: In this formula, w max and w min Represents the upper and lower limits of the inertia coefficient, t max Indicates the maximum number of iterations. The improved inertia coefficient shows a convex function distribution in the early stage of iteration, and a concave function distribution in the later stage; The speed factors c1 and c2 represent the effects of the individual optimal position and the population optimal position on the particle speed, respectively; Adaptive adjustment method: In the early stage, c1 should take a larger value; in the later stage, take a smaller value, and c2 should take the opposite 5. According to the large-scale MIMO hybrid precoding method based on the improved particle swarm optimization algorithm of claim 4, its notable feature is that the improved particle swarm tracking simulation precoding matrix phase step is as follows: Step 1: By fixing F RF To solve F BB , here we use the least squares method to solve F BB ; Step 2: Fix F BB , for F RF Use Euler's formula instead Initialize the necessary parameters of the population, use the position information of the particles to replace the angle information of the simulated precoding, and calculate the fitness of each population; Step 3: According to the optimized speed update and position update formula, adjust the fitness of each particle in the population, and use this method to determine the optimal position of each particle and the optimal position of the population; Step 4: Repeat step 3 until the maximum number of iterations is reached or the current threshold is less than the previous threshold or the initial given threshold, then the iteration ends; Step 5: Restore the optimal position of the final population to F RF Phase information of Finally, the digital precoding matrix is ​​restored through analog precoding and system power constraints, completing the design of hybrid precoding.

Images