Single intelligent metasurface beam forming method based on particle swarm optimization

Abstract

The invention discloses a particle swarm algorithm-based single intelligent metasurface beam forming method, which comprises the following steps of: firstly, generating an IRS position phase shift vector and a BS position beam forming vector by using a phase regulation matrix at an IRS position and a BS position, taking the generated vector as a single particle, then randomly generating a particle swarm, initializing each particle, calculating the fitness of each particle, and finally forming a single intelligent metasurface beam according to the particle swarm. And updating individual experience and group experience from the fitness, continuously carrying out iterative search to enable an optimal solution with the maximum reachable rate at the user, and recording the optimal solution. According to the method, large-scale global search can be carried out, local search can be finely carried out, the possibility of falling into a local optimal solution is reduced, excellent performance is shown in the non-convex optimization problem, and good performance is achieved in solving the joint optimization IRS and BS beam forming problem.

Description

Technical Field

[0001] The present invention belongs to the field of beamforming, and in particular relates to a single intelligent metasurface beamforming method based on a particle swarm algorithm. Background Art

[0002] 5G and above wireless technologies are designed to meet the application needs of vertical fields such as smart grids and smart cities. This requires a significant increase in network capacity and ubiquitous wireless connectivity for a large number of devices. To achieve these goals, many technologies have been developed in the past few years, such as millimeter wave (mmWave) communications, massive multiple input multiple output (MIMO) systems, and ultra-dense networks (UDN). However, these technologies come with huge implementation costs and huge energy consumption. In recent years, in order to improve network capacity, an energy-saving and low-cost technology has been developed - Intelligent Reflecting Surface (IRS), also known as Reconfigurable Smart Surface (RIS) or Large Smart Surface (LIS), which is a metasurface equipped with a large number of low-cost passive elements. The elements at the IRS can reflect the incident signal with a given phase shift, which can be optimized so that the reflected signal and the direct signal are constructively increased at the user (UE). Due to its cost-effective and energy-efficient artificial planar passive radio array structure, it has become a promising technology to improve the spectrum and energy efficiency of wireless networks. The functions of IRS / RIS are useful in a variety of scenarios, such as signal strength enhancement, anti-interference, and ensuring signal confidentiality.

[0003] At present, some research works have studied the beamforming design of BS and / or IRS. In order to maximize the energy efficiency of the system, IRS beamforming and power allocation are jointly optimized. There is also an IRS enhanced orthogonal frequency division multiplexing system proposed to maximize the downlink achievable rate by optimizing the beamforming at the IRS and the power allocation at the BS. In addition, the transmit power at the BS is minimized by jointly optimizing the BS and IRS beamforming. These works all consider the perfect channel state information (CSI) of the BS. However, estimating the CSI between the IRS and the BS or between the IRS and the UE is a challenge. Some studies have proposed possible solutions to this problem, however, they have the following possible disadvantages: the IRS needs to be equipped with multiple radio frequency (RF) chains; the IRS needs to perform not only phase control but also amplitude control, which increases the complexity of the system; the channel must be sparse, which is usually the case only in millimeter wave and LOS-dominated sub-6GHz systems]; (4) the training overhead increases with the number of elements of the IRS and / or BS.

[0004] In the application scenario of single IRS and multiple users, the optimization of the communication rate of the system is a non-convex problem. The existing methods have the disadvantages of high difficulty in solving, high computational complexity and slow convergence speed. Therefore, an optimization method is needed to achieve beamforming optimization and maximize the communication rate of the system while satisfying the BS transmission power constraint. Summary of the invention

[0005] In view of the shortcomings of existing research, the present invention proposes a single intelligent metasurface beamforming method based on particle swarm algorithm, and performs scene modeling according to the characteristics of the actual communication system and analyzes the optimization results, thereby solving the technical problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a single intelligent metasurface beamforming method based on particle swarm algorithm, the method comprising the following steps:

[0007] Step 1: Select an application scenario, select a downlink system, and model the channel as a model that includes large-scale path loss and small-scale fading;

[0008] Step 2: Set each unit at the smart metasurface to the maximum signal reflection, and obtain the phase shift vector at the smart metasurface and the beamforming vector at the base station according to the channel modeling and the reflection unit of the smart metasurface;

[0009] Step 3: The phase shift vector at the smart metasurface and the beamforming vector at the base station are combined into a vector pair. This vector pair is used as a particle and the particle swarm algorithm is used for joint optimization to obtain a matrix that maximizes the achievable rate at the user during the iteration process.

[0010] Step 4: Restore the matrix obtained by the particle swarm optimization algorithm to the beamforming vector at the base station and the phase shift vector at the smart metasurface.

[0011] Furthermore, the application scenario is a scenario where there is no obstruction in the path from the base station to the reflecting surface to the user;

[0012] Furthermore, the downlink system includes a base station having N transmitting antennas and equipped with a uniform linear array (ULA), a smart metasurface equipped with a uniform planar array (UPA) with M reflecting units, and a single antenna user.

[0013] Furthermore, the channel modeling includes channel modeling among a base station, an intelligent reflection surface and a user;

[0014] Furthermore, the phase shift vector at the smart metasurface and the beamforming vector at the base station are specifically column vectors of angles at the smart metasurface and the base station.

[0015] Furthermore, the joint optimization using the particle swarm optimization algorithm is specifically as follows:

[0016] Randomly generate L particles, or beamforming vector pairs. The beamforming vector pair is also the "position" of the particle in the particle swarm algorithm nomenclature. Each particle includes velocity information and position information. The position and velocity of the particle are initialized and randomly generated.

[0017] For each particle, if the transmission power at the base station is less than the transmission power constraint, the fitness of the particle is:

[0018]

[0019] in represents the channel vector from the intelligent metasurface (IRS) to the user (UE) (IRS-UE), is the channel matrix from the base station (BS) to the intelligent metasurface (IRS) (BS-IRS), is the channel vector from the base station to the user terminal (BS-UE), is the beamforming vector at the BS, where θ=[θ1,...,θ M ] represents the phase shift vector, θ m ∈[0,2π),β m ∈[0,1] is the amplitude reflection coefficient.

[0020] Update p from the calculated fitness best and g best .

[0021] where p best is the best fitness obtained by each particle up to the current iteration, g best It is the best fitness obtained by all particles up to the current iteration.

[0022] Generate L new particles, update the speed and position of each particle, generate L particles, and the group of the next iteration consists of the particles generated in this step. Update the beam forming of the base station and the smart metasurface to form a new vector pair (or particle).

[0023] Furthermore, the formula for updating the velocity and position of the particle is specifically:

[0024]

[0025]

[0026] in, and They represent the speed and position of the particle during iteration respectively. l1 and l2 are particle learning factors, which respectively define the influence of individual and collective experience on particle motion. ω is the inertial speed weight, which defines the influence of the current speed on the update speed of each particle.

[0027] Furthermore, the inertia speed weight is specifically:

[0028]

[0029] Among them, ω max and ω min are the maximum and minimum values ​​of ω.

[0030] Check whether the stopping condition is met. If it is met, return the most suitable particle, otherwise continue to iterate.

[0031] Beneficial effects of the present invention:

[0032] Compared with traditional methods, the particle swarm algorithm jointly optimizes the beamforming at the base station and the smart metasurface. The algorithm generates a large number of particles for optimization, thus having a strong global search capability and a more refined local search capability. It has the advantages of fast convergence speed and low algorithm complexity, and has good performance in solving the beamforming problem at the joint base station and the smart metasurface.

[0033] The performance of the particle swarm algorithm is closely related to the update of speed and position, so the present invention continuously adjusts the update coefficient during the iteration process to further enhance the method's ability to search for the optimal value.

[0034] This method can perform both large-scale global search and refined local search, and reduce the possibility of falling into the local optimal solution, showing relatively excellent performance in this non-convex optimization problem. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 A flow chart of a beamforming method of a particle swarm algorithm provided in an embodiment of the present invention;

[0036] Figure 2 A simulation diagram of an application scenario provided by an embodiment of the present invention;

[0037] Figure 3 This is a performance comparison chart for different numbers of IRS reflection units provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The present invention proposes a single intelligent metasurface beamforming method based on a particle swarm algorithm. The specific implementation mode of the present invention is further described in detail below in conjunction with the accompanying drawings, and scenario modeling is performed according to the characteristics of an actual communication system, and the optimization results are analyzed.

[0039] like Figure 1 As shown, it is a flow chart of a single intelligent metasurface beamforming method based on a particle swarm algorithm provided in an embodiment of the present invention; the specific process of the present invention includes:

[0040] like Figure 2 As shown, we consider a downlink system consisting of a BS with N transmit antennas equipped with a uniform linear array (ULA), an IRS equipped with a uniform planar array (UPA) with M reflective elements, and a single-antenna UE.

[0041] The signal received by the terminal can be written as

[0042]

[0043] in represents the channel vector from IRS to UE (IRS-UE), is the channel matrix from BS to IRS (BS-IRS), is the channel vector from the base station to the terminal (BS-UE), is the beamforming vector at the BS, s is the transmitted data, is an independent random variable with zero mean and unit variance, n is the additive Gaussian white noise at the user, and the power is σ 2 . where θ=[θ1,...,θ M ] represents the phase shift vector, θ m ∈[0,2π),β m ∈[0,1] is the amplitude reflection coefficient. For simplicity, we consider β m =1, i.e., maximum signal reflection for each element of IRS.

[0044] The channel model we use includes large-scale path loss and small-scale fading. The model includes channel modeling between the base station (BS), intelligent reflective surface (IRS) and user (User). Large-scale path loss describes the power attenuation of the signal during propagation. The following formula is used to calculate large-scale path loss: This is the path loss at the reference distance. Path loss between the base station (BS) and the intelligent reflecting surface (IRS): The coordinates of the base station and IRS are (x bs ,y bs ) and (x irs ,y irs ), the distance between the two The large-scale path loss from the base station to the IRS is calculated by the following formula: Path loss between base station (BS) and user (User): The user's location is (x user ,yuser ), the distance between the base station and the user Large-scale path loss from base station to user Path loss between the intelligent reflecting surface (IRS) and the user: the distance between the intelligent reflecting surface and the user Large-scale path loss from IRS to users

[0045] Small-scale fading is used to describe the rapid changes of signals caused by multipath effects and interference during propagation. We use the Rayleigh fading model and represent it by generating random complex numbers that follow a normal distribution.

[0046] Small-scale fading from base station (BS) to intelligent reflection surface (IRS): The channel matrix G from base station to IRS is expressed as: Where M and N represent the number of antennas of IRS and base station respectively. Small-scale fading from base station (BS) to user (User): The channel vector from base station to user is Small-scale fading from intelligent reflecting surface (IRS) to user: Channel vector from IRS to user The signal-to-noise ratio (SNR) at the final user (UE) is:

[0047]

[0048] Our main goal is to maximize the achievable rate by jointly optimizing the beamforming at the base station (BS) and the phase shift at the IRS while considering the transmit power constraint. This non-convex optimization problem can be formulated as:

[0049]

[0050] Since this optimization problem is non-convex, there is no standard way to solve it. Therefore, we propose a new method based on PSO technique, which is described below.

[0051] The above non-convex optimization problem is solved using a method based on the particle swarm algorithm. The particle swarm algorithm (PSO) is a stochastic optimization method inspired by the social behavior of animals. It is effective for the optimization of non-convex problems with continuous variables and has low computational complexity.

[0052] The PSO algorithm searches for the optimal solution through the cooperation between particles. Each particle has memory and can transmit information with other particles. In the PSO algorithm, each particle has two attributes: speed and position. The position is the value of the independent variable, while the speed represents the change of the independent variable. The quality of the particle position is evaluated by the fitness function, which is the optimization objective function.

[0053] In the particle swarm algorithm, the so-called "swarm" consists of L particles, where each particle represents a possible solution to the optimization problem. These particles move in the search space at different speeds. Their movement depends on individual experience (g best ) and group experience (p best ), which is obtained by evaluating the position of each particle through the fitness function. For each iteration, the velocity and position of each particle are updated according to the following formula:

[0054]

[0055] where t∈{1,...,N it} is the current iteration of particle i∈{1,...,L}, and Respectively represent the speed and position of particle i at iteration t, l1 and l2 are particle learning factors, which respectively define the impact of individual and collective experience on particle motion, and the function rand() returns a uniformly distributed random number between 0 and 1. Among them, ω is the inertial velocity weight, which defines the impact of the current velocity on the update speed of each particle, which is:

[0056]

[0057] where ω max and ω min are the maximum and minimum values ​​of ω.

[0058] The PSO algorithm is used to solve the above optimization problem. The velocity and position of each particle are updated using the above formula to generate L particles. In order to better understand the proposed method, its steps are described in detail below.

[0059] 1. Randomly generate L particles, or beamforming vector pairs, (ω l ,Θ l ),l=1,...,L. The beamforming vector pair is also the “position” of the particle in the Particle Swarm Optimization nomenclature.

[0060] 2. For each particle, if the transmission power at the BS is less than the transmission power constraint, the fitness of the particle is

[0061] 3. Update p from the fitness calculated in the previous step best and g best :p best is the best fitness obtained by each particle up to the current iteration, g bestis the best fitness obtained by all particles up to the current iteration. In our approach, the fittest particle is the largest one. These parameters define the individual and collective experience of the particles.

[0062] 4. In this step, L new particles are generated as shown below. The velocity and position of each particle are updated separately to generate L particles. The group of the next iteration consists of the particles generated in this step, that is, from this operation, the beamforming of BS and IRS is updated to form new vector pairs (or particles).

[0063] 5. Check whether the stopping condition is met. If so, return the most suitable particle (w best ,θ best ). Otherwise, go to step 2.

[0064] In our proposed method, the BS to be measured and the IRS beamforming vector pair are defined at the BS, and the IRS beamforming is sent by the BS to the controller, such as Figure 2 shown.

[0065] Compared with traditional methods, the particle swarm optimization algorithm has stronger global search capability, fast convergence speed, low algorithm complexity, etc. It has good performance in solving the joint optimization of IRS and BS beamforming problems.

[0066] Figure 3 The performance comparison diagram under different numbers of IRS reflection units provided by the embodiment of the present invention shows that when the number of IRS reflection units increases, the achievable rate at the user also increases, and the convergence speed of the algorithm is faster.

[0067] The above embodiments are used to illustrate the present invention rather than to limit the present invention. Any modification and change made to the present invention within the spirit of the present invention and the protection scope of the claims shall fall within the protection scope of the present invention.

Claims

1. A single intelligent metasurface beamforming method based on particle swarm algorithm, characterized in that: The following steps are involved: Step 1: Select an application scenario, select a downlink system, and model the channel as a model that includes large-scale path loss and small-scale fading; Step 2: Set each unit at the smart metasurface to the maximum signal reflection, and obtain the phase shift vector at the smart metasurface and the beamforming vector at the base station according to the channel modeling and the reflection unit of the smart metasurface; Step 3: The phase shift vector at the smart metasurface and the beamforming vector at the base station are combined into a vector pair. This vector pair is used as a particle and the particle swarm algorithm is used for joint optimization to obtain a matrix that maximizes the achievable rate at the user during the iteration process. Step 4: Restore the matrix obtained by the particle swarm optimization algorithm to the beamforming vector at the base station and the phase shift vector at the smart metasurface.

2. According to the particle swarm algorithm-based single-intelligent metasurface beamforming method of claim 1, it is characterized in that: The application scenario is a scenario where the path from the base station to the reflecting surface to the user is unobstructed.

3. According to the particle swarm algorithm-based single-intelligent metasurface beamforming method of claim 1, it is characterized in that: The downlink system includes a base station with N transmitting antennas, equipped with a uniform linear array (ULA), an intelligent metasurface equipped with a uniform planar array (UPA) with M reflection units, and a single-antenna user, and the channel modeling includes channel modeling among the base station, the intelligent reflection surface and the user.

4. The single-intelligent metasurface beamforming method based on particle swarm algorithm according to claim 1, characterized in that: The phase shift vector at the smart metasurface and the beamforming vector at the base station are specifically column vectors of angles at the smart metasurface and the base station.

5. The single-intelligent metasurface beamforming method based on particle swarm algorithm according to claim 1, characterized in that: The joint optimization using the particle swarm optimization algorithm is specifically as follows: Randomly generate L particles, or beamforming vector pairs. The beamforming vector pair is also the "position" of the particle in the particle swarm algorithm nomenclature. Each particle includes velocity information and position information. The particle position and velocity are initialized and randomly generated. For each particle, if the transmission power at the base station is less than the transmission power constraint, the fitness of the particle is: in represents the channel vector from the intelligent metasurface (IRS) to the user (UE) (IRS-UE), is the channel matrix from the base station (BS) to the intelligent metasurface (IRS) (BS-IRS), is the channel vector from the base station to the user terminal (BS-UE), is the beamforming vector at the BS, where θ=[θ1,...,θ M ] represents the phase shift vector, θ m ∈[0,2π),β m ∈[0,1] is the amplitude reflection coefficient; Update p from the calculated fitness best and g best ; where p best is the best fitness obtained by each particle up to the current iteration, g best The best fitness obtained by all particles up to the current iteration; Generate L new particles, update the speed and position of each particle, generate L particles, and the group of the next iteration consists of the particles generated in this step. Update the beam forming of the base station and the smart metasurface to form a new vector pair (or particle).

6. The single-intelligent metasurface beamforming method based on particle swarm algorithm according to claim 5, characterized in that: The formula for updating the velocity and position of the particle is specifically: in, and They represent the speed and position of the particle during iteration respectively. l1 and 2 are particle learning factors, which define the influence of individual and collective experience on particle motion respectively. ω is the inertial speed weight, which defines the influence of the current speed on the update speed of each particle.

7. The single-intelligent metasurface beamforming method based on particle swarm algorithm according to claim 6, characterized in that: The inertia speed weight is specifically: Among them, ω max and ω min are the maximum and minimum values ​​of ω; Check whether the stopping condition is met. If it is met, return the most suitable particle, otherwise continue to iterate.

Images