Method for optimizing user experience rate in wireless communication network based on genetic algorithm

Abstract

The present invention proposes a method for optimizing the user experience rate of a wireless communication network based on a genetic algorithm, which solves the problem that the current optimization method is inefficient and the optimization effect is difficult to guarantee. The implementation steps of the method are: generating a training set; constructing an MLP mapping model of the user access distance and the base station's existing network parameters, and training the MLP mapping model; calculating the optimal value of the downtilt angle of the base station's existing network parameters through a genetic algorithm based on the MLP mapping model; adjusting the downtilt angle of the base station's existing network parameters to the optimal value. The present invention calculates the optimal value of the downtilt angle of the base station's existing network parameters through a genetic algorithm to improve the user experience rate, so it does not need to rely on manual processing, and can intelligently provide an optimization plan, significantly improving the efficiency of optimizing the user experience rate.

Description

Technical Field

[0001] The present invention relates to the field of communications technology, and more particularly to a method for optimizing user experience rates in wireless communication networks based on a genetic algorithm. The present invention can be applied to optimize the experience rates of all users in real time in an actual wireless communication network environment. Background Art

[0002] User Experience Rate (UXR) refers to the minimum transmission rate a user can experience in an actual wireless communication network environment. It directly reflects the user's actual experience when using services. Wireless communication network performance indicators include UXR, user peak rate, mobility, end-to-end latency, connection density, service density, average spectrum efficiency, and energy efficiency. UXR is a key indicator of wireless communication network performance. Optimizing UXR can improve the user experience when using services in wireless communication networks. The key process for optimizing UXR can be summarized as data collection, performance analysis, and optimization solution implementation and testing. Currently, most UXR optimization methods for wireless communication networks are developed by professional technicians who are well-versed in the relevant theoretical knowledge of wireless communication networks, have a thorough understanding of their system operations and mechanical structure, and have a clear understanding of existing network configuration parameters. Therefore, through appropriate parameter adjustments, these technicians can improve UXR in wireless communication networks. However, most existing optimization methods rely on manual or semi-manual processing, and optimization results largely depend on the skills and experience of professional engineers. This approach is inefficient and difficult to guarantee.

[0003] In their paper "Research and Practice on Improving 5G User Perceived Rate" (Communication Management and Technology, 2021(1):46-48), Fan Zhongyang et al. proposed a method for optimizing user experience rate through collaborative investigation. The steps of this method are as follows: first, by analyzing the 5G user experience rate calculation formula, the hard conditions that need to be met for high rate are summarized, and the overall idea of ​​improving low rate is given; second, collaborative investigation is carried out from the terminal, air interface, base station, transmission, core network, server and other professional fields, and a collaborative optimization plan is formulated. The collaborative optimization plan is carried out from four dimensions: structure, quality, scheduling and maintenance. Among them, the structure category has 1 indicator, involving wireless (planning) professional fields; the quality category has 4 indicators, involving terminal, wireless (air interface base station, planning, maintenance) and other professional fields; the scheduling category has 4 indicators, involving terminal, wireless (air interface base station), transmission, core network, server and other professional fields; the maintenance category has 2 indicators, involving wireless (air interface base station), core network and other professional fields; the third step is to implement the collaborative optimization plan to verify the optimization effect of user experience rate. The shortcoming of this method is that since this method summarizes the hard conditions that need to be met for high speed by analyzing the 5G user experience rate calculation formula, this method of giving an idea of ​​improving the user experience rate by analyzing the 5G user experience rate calculation formula relies on manual processing, which is inefficient and the optimization effect is difficult to guarantee. Summary of the Invention

[0004] The purpose of the present invention is to address the shortcomings of the above-mentioned existing technologies and propose a method for optimizing the user experience rate of a wireless communication network based on a genetic algorithm, aiming to solve the problems that most current optimization methods rely on manual or semi-manual processing, have low optimization efficiency, and are difficult to guarantee optimization effects.

[0005] To achieve the above objectives, the technical approach of the present invention is to analyze the relationship between user experience rate and user access distance using real-world data, construct an MLP mapping model between user access distance and existing network configuration parameters, and then use a genetic algorithm to determine the optimal value for the downtilt angle in the existing network configuration parameters to optimize the user access distance and ultimately improve the user experience rate. Because the genetic algorithm is used to determine the optimal value for the downtilt angle in the existing network configuration parameters to provide an optimization solution for user experience rate, it eliminates the need for manual processing and can intelligently provide an optimization solution, significantly improving the efficiency of optimizing user experience rate. This overcomes the problem of existing optimization methods that rely on manual processing, resulting in low efficiency and difficulty in ensuring effectiveness.

[0006] The technical solution adopted by the present invention includes the following steps:

[0007] Step 1: Generate training set;

[0008] Step 2: Build an MLP mapping model of user access distance and base station network parameters, and train the MLP mapping model;

[0009] Step 3: Calculate the optimal value of the downtilt angle in the existing base station network parameters using a genetic algorithm based on the MLP mapping model;

[0010] Step 3.1: Binary encode the downtilt angle in the existing network parameters of the base station and construct an initial population;

[0011] Step 3.2: Use the trained MLP mapping model to evaluate the fitness of individual users.

[0012] Step 3.3, select the user individual with the highest fitness in the current population as the parent individual;

[0013] Step 3.4, crossover the chromosomes of the parent individuals according to the uniform crossover rule to obtain the crossover population;

[0014] Step 3.5, randomly exchange the values ​​of two genes in the chromosomes of individuals in the crossover population to obtain the exchanged population;

[0015] Step 3.6: Determine whether the downtilt angle of the individual user in the population after the exchange has reached the local optimum. If so, perform a decoding operation on the local optimum downtilt angle to obtain the optimal value and then execute step 4. Otherwise, execute step 3.2.

[0016] Step 4: Adjust the downtilt angle of the existing network parameters in the base station to the optimal value.

[0017] Compared with the prior art, the present invention has the following advantages:

[0018] Since the present invention analyzes the relationship between the user experience rate and the user access distance through real collected data, and constructs an MLP mapping model of the user access distance and the existing network configuration parameters, and then uses a genetic algorithm to obtain the optimal value of the downtilt angle in the existing network configuration parameters to optimize the user access distance, the user experience rate is ultimately improved. This overcomes the problem that the existing technology relies on manual processing, resulting in low efficiency and difficult to ensure the effect. Compared with the existing technology, the present invention can intelligently provide an optimization solution for the user experience rate, significantly improve the efficiency of optimizing the user experience rate, save human resources, and can ensure the optimization effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart for implementing the present invention. DETAILED DESCRIPTION

[0020] The following combination Figure 1 , the implementation steps of the embodiment of the present invention are further described in detail.

[0021] Step 1: Generate a training set.

[0022] The training set is a sample set consisting of M users. The data of each user in the sample set is cleaned and standardized in turn. Each sample contains the user's Q-dimensional base station network parameter information S m and the distance y between each user and the base station (m) , S m Is a vector with 1 row and Q columns, y (m) As the information of each user's base station network parameters S m The label, S m Indicates that the m-th user contains Q-dimensional features related to the base station network engineering parameters, M>3000, Q>3, 1≤m≤M.

[0023] The cleaning refers to deleting the null values ​​and abnormal values ​​in each column of the base station's existing network engineering parameter related information and the user access distance.

[0024] The standardization refers to calculating the average value and standard deviation of each column of the base station network engineering parameter related information and the user access distance, and scaling the data so that the data falls into a small specific range.

[0025] The training set used in this embodiment of the present invention has undergone data cleaning. This involves removing null and outlier values ​​from every column of information related to base station network parameters and user access distances. For example, the presence of negative values ​​in antenna height data is clearly an anomaly. Therefore, data cleaning is necessary, primarily removing null and outlier values. After cleaning, the training set contains M sample data, ensuring data quality and accuracy.

[0026] After data cleaning, the training set selected in the embodiments of the present invention requires data normalization. Data normalization calculates the mean and standard deviation of each column in the base station's existing network parameters and the user access distance, scaling the data so that it falls within a small, specific range. This aims to eliminate dimensional differences between different features and ensure that each feature has a relatively equal weight influencing the model during model training. Data normalization also aids in training deep learning models, reducing fluctuations during gradient descent, thereby accelerating model convergence and improving model performance to a certain extent.

[0027] In the embodiment of the present invention, M=8800, Q=6, and each sample contains 6-dimensional base station network engineering parameter related information of the user, wherein the base station network engineering parameter related information includes antenna height, azimuth angle, downtilt angle, electrical downtilt angle, coverage range, and site distance.

[0028] Step 2: Build an MLP mapping model of user access distance and base station network parameters, and train the MLP mapping model.

[0029] The number of neurons in the input layer of the MLP network depends on the characteristic dimension of the training sample. The number of layers and the number of neurons in the hidden layer need to be adjusted and determined according to the actual situation. It will directly affect the fitting ability of the model. The number of neurons in the output layer depends on the specific task. The output layer produces the final prediction result. If the task to be solved is regression, the prediction result is one-dimensional. If the task to be solved is classification, the prediction result is the corresponding category. Therefore, the present invention constructs a mapping model of user access distance and existing network configuration parameters through the MLP network. The number of neurons in the input layer depends on the dimension of the existing network configuration parameters. The user access distance is a continuous indicator. The task solved by the MLP mapping model is regression. The number of nodes in the output layer should be 1.

[0030] The MLP mapping model is a 6-layer MLP network structured as follows: an input layer, a first hidden layer, a second hidden layer, a third hidden layer, a fourth hidden layer, and an output layer connected in series. The number of nodes in the input layer is set to q, and the number of neurons in the first to fourth hidden layers is set to 256, 128, 64, and 32, respectively. The input layer and each hidden layer are implemented using the ReLU activation function. The number of nodes in the output layer is set to 1, and the output layer is implemented using the linear activation function, where q = Q. In the embodiment of the present invention, q = 6.

[0031] The steps of training the MLP mapping model are as follows:

[0032] In the first step, the parameters of the MLP network, including weights and biases, are randomly initialized. The weights are initialized through a random number generator, and the biases are initialized to 0.

[0033] In the second step, the training set is input into the MLP mapping model, and the Adam optimizer is used to perform forward propagation and backpropagation training according to the stochastic gradient descent method. The model parameters are iteratively updated until the loss function of the model converges, and the trained MLP mapping model is obtained.

[0034] The loss function is as follows:

[0035]

[0036] Among them, L MSE represents the loss function of the model, m represents the sample number in the training set, M represents the total number of samples in the training set, that is, the total number of users, and y (m) represents the real value of the access distance of the mth user, Indicates the predicted value of the access distance of the mth user.

[0037] The loss function of the MLP mapping model of the embodiment of the present invention on the training set gradually decreases and tends to be stable with the iterative update of the network parameters, indicating that the performance of the MLP mapping model gradually improves and tends to be stable. After the training is completed, an MLP mapping model with good performance can be obtained.

[0038] Step 3: Based on the MLP mapping model and using a genetic algorithm, the optimal value of the downtilt angle in the existing network working parameters of the base station is calculated.

[0039] Step 3.1: Binary encode the downtilt angle in the existing network parameters of the base station and construct an initial population.

[0040] The downtilt angle refers to the angle between the radiation direction of the base station antenna and the ground.

[0041] The binary code is a binary symbol set consisting of 0 and 1. The encoding and decoding operations are simple and easy, and it is convenient to implement genetic operations such as crossover and mutation in genetic algorithms.

[0042] In step 3.2, the fitness of individual users is evaluated using the trained MLP mapping model.

[0043] In genetic algorithms, fitness is used to evaluate the quality of an individual in solving a problem. The chromosomes of individuals with high fitness are more likely to be inherited to the next generation.

[0044] The fitness of an individual is evaluated through the trained MLP mapping model, and the output user access distance of the MLP mapping model is the fitness value of each individual user.

[0045] Step 3.3: Select the user individual with the highest fitness in the current population as the parent individual.

[0046] Step 3.4: Crossover the chromosomes of the parent individuals according to the uniform crossover rule to obtain the crossover population.

[0047] In each generation of the population, each individual is traversed and the crossover rate is used to determine whether to perform a crossover operation. The results show that setting the crossover rate between 0.4 and 0.99 can achieve better evolutionary performance.

[0048] The uniform crossover rule means that the chromosomes of the parent individuals are crossed to obtain new individuals. After the crossover operation is performed, the fitness of the chromosomes of the new individuals is compared with that of the parent chromosomes, and the individuals with greater fitness are selected to be added to the post-crossover population.

[0049] Step 3.5: Randomly exchange the values ​​of two genes in the chromosomes of individuals in the post-crossover population to obtain the post-exchange population.

[0050] Step 3.6: Determine whether the downtilt angle of the individual user in the population after the exchange has reached the local optimum. If so, perform a decoding operation on the local optimum downtilt angle to obtain the optimal value and then execute step 4. Otherwise, execute step 3.2.

[0051] In the embodiment of the present invention, the binary code length is set to 10, the size of the initialization population is set to 200, the crossover rate is 0.8, and the mutation rate during random exchange is 0.003.

[0052] Step 4: Adjust the downtilt angle of the existing network parameters in the base station to the optimal value.

[0053] By adjusting the downtilt angle in the existing network parameters of the base station to the optimal value calculated by the genetic algorithm, the user access distance will be maximized, improving the coverage capability of the base station, and the user experience rate can be significantly improved, which can improve the user's real experience of the quality of the wireless communication network.

[0054] The technical effects of the present invention are further illustrated below in conjunction with simulation experiments.

[0055] 1. Simulation experiment conditions:

[0056] The hardware platform for the simulation experiment of the present invention is: processor Intel(R) Core(TM) i3-8100 CPU, main frequency 3.60GHz, memory 12GB; graphics card Intel(R) UHD Graphics 630, 512GB solid state drive.

[0057] The software platform for the simulation experiment of the present invention is: Windows 10 operating system and Python 3.9.7.

[0058] The data set used in the simulation experiment of the present invention is a sample set consisting of 8,800 users. Each sample contains the user's 6-dimensional base station network engineering parameter related information and access distance. The 6-dimensional base station network engineering parameter related information includes antenna height, azimuth, downtilt angle, electrically adjusted downtilt angle, coverage range, and site distance.

[0059] 2. Simulation experiment and content results analysis:

[0060] The simulation experiment of the present invention uses the genetic algorithm of the present invention to calculate the downtilt angle of the input base station network parameters, obtain the optimal downtilt angle value, and adjust the base station downtilt angle to the optimized optimal value to increase the user access distance, improve the base station coverage capability, and further improve the user experience rate. After the simulation of the present invention was actually operated by a communication operator, the comparison of the user experience rate before and after optimization was obtained, as shown in Table 1.

[0061] Table 1 Comparison of user experience rate optimization effects of the present invention

[0062] The present invention User Experience Rate Before optimization 33.006Mbps After optimization 105.796Mbps

[0063] As shown in Table 1, the user experience rate is increased by 221% after adopting the present invention, indicating that the downtilt angle optimization solution calculated by the genetic algorithm provided by the present invention can effectively improve the user experience rate and provide a new idea for optimizing the performance of wireless communication networks.

Claims

1. A method for optimizing user experience rate in a wireless communication network based on a genetic algorithm, characterized in that: The optimal value of the downtilt angle in the existing base station network parameters is obtained through a genetic algorithm. The specific steps of this method include the following: Step 1: Generate training set; Step 2: Build an MLP mapping model of user access distance and base station network parameters, and train the MLP mapping model; Step 3: Calculate the optimal value of the downtilt angle in the existing base station network parameters using a genetic algorithm based on the MLP mapping model; Step 3.1: Binary encode the downtilt angle in the existing network parameters of the base station and construct an initial population; Step 3.2: Use the trained MLP mapping model to evaluate the fitness of individual users. Step 3.3, select the user individual with the highest fitness in the current population as the parent individual; Step 3.4, crossover the chromosomes of the parent individuals according to the uniform crossover rule to obtain the crossover population; Step 3.5, randomly exchange the values ​​of two genes in the chromosomes of individuals in the crossover population to obtain the exchanged population; Step 3.6: Determine whether the downtilt angle of the individual user in the population after the exchange has reached the local optimum. If so, perform a decoding operation on the local optimum downtilt angle to obtain the optimal value and then execute step 4. Otherwise, execute step 3.

2. Step 4: Adjust the downtilt angle of the existing network parameters in the base station to the optimal value.

2. The method for optimizing user experience rate of a wireless communication network based on a genetic algorithm according to claim 1, characterized in that: The training set described in step 1 contains a sample set consisting of M users. The data of each user in the sample set is cleaned and standardized in turn. Each sample contains the user's Q-dimensional base station network parameter information S m and the distance y between each user and the base station (m) , S m Is a vector with 1 row and Q columns, y (m) As the information of each user's base station network parameters S m The label, S m Indicates that the m-th user contains Q-dimensional features related to the base station network engineering parameters, M>3000, Q>3, 1≤m≤M.

3. The method for optimizing user experience rate of a wireless communication network based on a genetic algorithm according to claim 2, characterized in that: The information related to the existing network engineering parameters of the base station includes antenna height, azimuth, downtilt angle, electrically adjusted downtilt angle, coverage range, and site distance.

4. The method for optimizing user experience rate of a wireless communication network based on a genetic algorithm according to claim 2, characterized in that: Cleaning means deleting the empty values ​​and abnormal values ​​in each column of the base station's existing network engineering parameter information and the user access distance.

5. The method for optimizing user experience rate of a wireless communication network based on a genetic algorithm according to claim 2, characterized in that: The standardization refers to calculating the average value and standard deviation of each column of the base station network engineering parameter related information and the user access distance, and scaling the data so that the data falls into a small specific range.

6. The method for optimizing user experience rate of a wireless communication network based on a genetic algorithm according to claim 2, characterized in that: The MLP mapping model described in step 2 is a 6-layer MLP network, whose structure is composed of an input layer, a first hidden layer, a second hidden layer, a third hidden layer, a fourth hidden layer, and an output layer connected in series in sequence; the number of nodes in the input layer is set to q, and the number of neurons in the first to fourth hidden layers is set to 256, 128, 64, and 32 respectively. The input layer and each hidden layer are implemented by the ReLU activation function, the number of nodes in the output layer is set to 1, and the output layer is implemented by the linear activation function, where q = Q.

7. The method for optimizing user experience rate of a wireless communication network based on a genetic algorithm according to claim 1, characterized in that: The steps for training the MLP mapping model described in step 2 are as follows: The first step is to randomly initialize the parameters of the MLP network, including weights and biases. The weights are initialized through a random number generator, and the biases are initialized to 0. In the second step, the training set is input into the MLP mapping model, and the Adam optimizer is used to perform forward propagation and backpropagation training according to the stochastic gradient descent method. The model parameters are iteratively updated until the loss function of the model converges, and the trained MLP mapping model is obtained. The loss function is as follows: Among them, L MSE represents the loss function of the model, m represents the sample number in the training set, M represents the total number of samples in the training set, that is, the total number of users, and y (m) represents the true value of the access distance of the mth user, Indicates the predicted value of the access distance of the mth user.

8. The method for optimizing user experience rate of a wireless communication network based on a genetic algorithm according to claim 1, characterized in that: The downtilt angle mentioned in step 3.1 refers to the angle between the base station antenna radiation direction and the ground.

9. The method for optimizing user experience rate of a wireless communication network based on a genetic algorithm according to claim 1, characterized in that: In step 3.2, the fitness of the individual is evaluated by the trained MLP mapping model. The output user access distance of the MLP mapping model is the fitness value of each individual user.

10. The method for optimizing user experience rate of a wireless communication network based on a genetic algorithm according to claim 1, characterized in that: The uniform crossover rule described in step 3.4 means that the chromosomes of the parent individuals are crossed to obtain new individuals. After the crossover operation is performed, the fitness of the chromosomes of the new individuals is compared with that of the parent chromosomes, and the individuals with larger fitness are selected to join the post-crossover population.

Images