A base station load prediction method and device combining deep learning and statistical methods, and electronic equipment

Abstract

The application discloses a base station load prediction method and device combining deep learning and statistical methods and electronic equipment, and belongs to the field of mobile communication. The method comprises the following steps: defining the load as the number of users entering the coverage range of the base station, decomposing the base station load historical data Y(t) into two parts: a part Y1(t) following a long-term trend and changing with time according to a specific rule, and a part Y2(t) caused by short-term random fluctuations; accurately predicting the internal law of Y1(t) changing with time, training an LSTM network to complete the prediction; using a statistical method to predict Y2(t) which has strong burstiness and real-time performance; finally, predicting the load at multiple future time points and controlling the base station sleep according to the prediction. The application solves the problem of how to minimize the error of load prediction in a high mobility scenario, predicts the load at multiple future time points through the load of the base station in the past period and the current time, and formulates a base station sleep control strategy.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of mobile communication, and more particularly relates to a base station load prediction method and device combining deep learning and statistical methods and an electronic device. BACKGROUND

[0002] In recent years, mobile communication networks have deployed a large number of base stations to improve performance. In order to ensure communication quality, base station deployment is usually planned according to the number of users and the peak of service demand, and a certain degree of redundancy is left. However, the distribution of users has a tidal effect, i.e. it changes significantly over time. If all base stations are kept in working state at all times, it will result in a large number of base stations with low utilization and serious energy waste.

[0003] To address this challenge, base station sleep control has become a key strategy to reduce energy consumption. When there are fewer users, some base stations can enter a sleep state, and users within their coverage area will switch to neighboring base stations. With the increase of high-mobility terminals (such as vehicles and drones), these terminals will pass through the coverage areas of multiple base stations and perform multiple base station handovers in a short period of time, causing rapid changes in base station load in a short period of time. This makes accurate load prediction essential for effective base station sleep control.

[0004] In existing research, base station load prediction is usually modeled as a time series prediction problem. However, in high-mobility scenarios, the load of high-frequency mobile communication networks is influenced by both long-term regularity (tidal effect) and short-term fluctuations. Existing solutions do not consider both. SUMMARY

[0005] To address the need for improvement of the prior art, the present application provides a base station load prediction method and device combining deep learning and statistical methods, thereby solving the problem of base station load prediction in high-mobility scenarios. In addition, the predicted load data can be used to solve the problem of intelligent base station sleep control strategy.

[0006] To achieve the above-mentioned purpose, according to the first aspect of the present application, a base station load prediction method combining deep learning and statistical methods is provided, comprising:

[0007] S1, according to the definition of base station load, the base station load history data Y(t) is divided into two parts: Y1(t) which follows a long-term trend and changes over time according to a certain rule, and Y2(t) which is a short-term random fluctuation; wherein Y1(t) is the average data obtained by averaging the past load data in a sliding window, and Y2(t) is the difference between Y(t) and Y1(t);

[0008] Wherein, the communication system is a multi-user multi-base station mobile communication system, the base station load is defined as the number of users in the coverage of each base station, and the load of the high-frequency mobile communication network base station is affected by two factors: the change of the load in the long time dimension caused by the tidal effect of user distribution, and the fluctuation in the short time caused by the random movement of users.

[0009] S2, according to the inherent law of Y1(t) changing with time, a long short-term memory neural network (LSTM) is trained to obtain a trained LSTM neural network to complete the first load prediction and obtain first load prediction data;

[0010] S3, according to the strong burstiness and real-time of Y2(t), a statistical method is used for the second load prediction to obtain second load prediction data;

[0011] S4, since the frequency of base station dormancy control update is much lower than the frequency of load prediction update, the base station needs to predict the load at multiple future time points, and according to the first load prediction data and the second load prediction data at multiple future time points, the base station dormancy control problem is modeled as MDP and solved, and the base station dormancy control is performed accordingly.

[0012] Preferably, the step S2 comprises:

[0013] T1, determining the structure and initial parameters of the LSTM neural network;

[0014] T2, normalizing Y1(t) to divide it into training data and test data for training and testing respectively;

[0015] T3, updating the weight values of the LSTM neural network by minimizing the loss function representing the difference between the output of the LSTM neural network and the actual data, and completing training and testing by using the training data and the test data respectively to obtain the first load prediction data.

[0016] Preferably, the LSTM neural network structure is as follows: 1) the input of the input layer is the sequence of the current time and the previous sliding window; 2) an activation function ReLu full connection layer is used after the input layer for feature extraction; 3) a plurality of recursive layers based on LSTM structure are used for analyzing and identifying the internal relevance of the input sequence; 4) the output layer is a softmax function, wherein the output value of each neuron represents the probability of the load at the next time being in the value corresponding to the neuron.

[0017] Preferably, the step S3 comprises:

[0018] R1, using kernel density estimation on Y2(t) to obtain short-term prediction value y1 and long-term prediction value y2;

[0019] R2, the prediction value y3 is obtained using the weighted moving average method for Y2(t);

[0020] R3, since the accuracy of different prediction values varies with the measurement period, a final prediction value, i.e., the second load prediction data, is obtained using dynamic weights in combination with the above three prediction values.

[0021] Preferably, the step R3:

[0022] y = ay1 + by2 + gy3

[0023] wherein y is the final prediction value, a, b and g are dynamic weights, a + b + g = 1 and are all non-negative values.

[0024] According to a second aspect of the present application, a base station load prediction device combining deep learning and statistical methods is provided, comprising:

[0025] a base station load data acquisition module, which acquires the load of a base station at a past period of time and a current time according to the number of users entering the coverage of the base station, as base station load history data;

[0026] a data preprocessing module, which is used to divide the collected base station load history data into two parts, i.e., a long-term trend part Y1(t) and a short-term random fluctuation part Y2(t), and perform preprocessing operations such as cleaning and normalization on the original data;

[0027] a load prediction module, which contains two sub-modules:

[0028] an LSTM neural network prediction sub-module, which is used to train an LSTM neural network according to Y1(t) to complete first load prediction and obtain first load prediction data, so as to capture and predict the long-term trend part of the load.

[0029] a statistical method prediction sub-module, which is used to perform second load prediction according to Y2(t) using a proper statistical method (such as exponential weighted moving average) to obtain second load prediction data, and model and predict the short-term random fluctuation part of the load.

[0030] a multi-step prediction integration module, since the frequency of base station dormancy control update is much lower than the frequency of load prediction update, this module is used to integrate the first load prediction data and the second load prediction data at multiple future time points to generate load prediction values at the multiple future time points.

[0031] a base station dormancy control module, which is used to model the base station dormancy control problem as MDP based on the integrated load prediction data and solve it to perform base station dormancy control.

[0032] According to a third aspect of the present application, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to the first aspect when executing the computer program.

[0033] Overall, compared with the prior art, the above technical solutions conceived by the present application can achieve the following beneficial effects:

[0034] 1. The base station load prediction method based on deep learning provided in the present application analyzes the influence of different factors on long-term change trend and short-term fluctuation of load, divides the load into two parts, according to the change of load in a longer time dimension caused by the tidal effect of user distribution, uses LSTM neural network for prediction, and the accuracy is better than other methods.

[0035] 2. The base station load prediction method based on statistics provided in the present application uses kernel density estimation to obtain short-term prediction value y1 and long-term prediction value y2 according to the short-term fluctuation caused by random movement of users, uses weighted moving average method to obtain prediction value y3, and combines the three prediction values through dynamic weight, which can effectively reduce prediction error and realize high-precision load prediction. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 The scene schematic diagram provided for the embodiments of the present application;

[0037] Figure 2 The system block diagram provided for the embodiments of the present application;

[0038] Figure 3 The flowchart of the base station load prediction method combining deep learning and statistical method provided for the embodiments of the present application;

[0039] Figure 4 The electronic device schematic diagram provided for the embodiments of the present application. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application. In addition, the technical features involved in each embodiment of the present application described below can be combined with each other as long as they do not conflict with each other.

[0041] In existing research, base station load prediction is usually modeled as a time series prediction problem. Prediction methods based on ARIMA model and Markov model focus on predicting load changes caused by short-term fluctuations, while prediction methods based on models such as LSTM focus on learning long-term trends of load changes. In a high mobility scenario, the load of a high frequency mobile communication network is affected by both long-term regularity (tide effect) and short-term fluctuations.

[0042] Based on this, the embodiment of the present application provides a base station load prediction method combining deep learning and statistical methods, which is used for a base station load prediction end and can solve the problem that the load in a high mobility scenario is difficult to predict, and minimize the load prediction error.

[0043] The communication system model is a multi-user multi-base station mobile communication system, as shown in Figure 1 In a target area, multiple high frequency base stations jointly provide services for multiple users. Each user can be in the coverage of one or more base stations, and is connected to one of the base stations in working state according to a certain standard (such as the highest received signal strength). There is a network controller in the area to collect various information in the network and make decisions on base station dormancy.

[0044] The load prediction scheme aims to predict the load at the next moment by the load of the base station in the past period and the current moment. In order to reflect the utilization rate of the base station in working state, the load defined here refers to the number of users entering the coverage of the base station. The proposed prediction model can be represented as follows:

[0045] Y(t+Δ)=f(Y(t),Y(t-Δ),Y(t-2Δ),...)

[0046] Where Y(t) represents the load of the base station at time t, Δ is the time interval between recording adjacent two data, and f(·) is a function mapping between the past and current load and the load at the next moment. As shown in Figure 2 The load of the base station of the high frequency mobile communication network is affected by two factors: the load changes in a long time dimension caused by the tide effect of user distribution, and the short-term fluctuations caused by random movement of users. Therefore, Y(t) can be decomposed into two parts: Y1(t) is the part of the load following the long-term trend and changing with time according to a certain rule, and Y2(t) is the part of the load changing caused by short-term random fluctuations.

[0047] The flowchart of the base station load prediction method combining deep learning and statistical methods proposed by the embodiment of the present application is shown in Figure 3 The steps S1-S4 contained in the load prediction method are described in detail as follows:

[0048] S1, according to the definition of base station load, the base station load history data Y(t) is divided into two parts Y1(t) and Y2(t) which follow long-term trends and short-term random fluctuations respectively;

[0049] Specifically, the communication system is a multi-user multi-base station mobile communication system, and the base station load is defined as the number of users in the coverage of each base station. The load of the high-frequency mobile communication network base station is affected by two factors: the load changes in the long time dimension caused by the tidal effect of user distribution, and the short-term fluctuations caused by the random movement of users.

[0050] In order to extract Y1(t) and Y2(t) from Y(t), the data of Y(t) can be filtered by averaging calculation to achieve the effect of "denoising". The "smoothed" data obtained after filtering will become Y1(t) data, and the remaining part will become Y2(t) data. Specifically, Y1(t) and Y2(t) are obtained as follows:

[0051]

[0052] Y2(t) = Y(t) - Y1(t)

[0053] Y1(t) is the data obtained by averaging the past data in a sliding window, which can reduce the disturbance of short-term events (such as vehicles and drones quickly passing through the coverage of the base station) to the data, and highlight the inherent rules of the data in the time domain. The remaining part Y2(t) corresponds to the load fluctuation in a short time (which can be negative).

[0054] S2, according to the inherent rules of Y1(t) changing with time, a LSTM neural network is trained using historical data to complete the first load prediction, and the first load prediction data is obtained.

[0055] The load prediction model is a sequential model, and the prediction value only depends on the historical load data. According to the base station load prediction model, the following LSTM neural network is constructed: 1) the input of the input layer is the sequence of the current time and the previous sliding window; 2) an all-connected layer with ReLu as the activation function is used for feature extraction after the input layer; 3) then there are several recursive layers based on LSTM structure to analyze and identify the internal relevance of the input sequence; 4) the output layer is a softmax function, and the output value of each neuron represents the probability of the load at the next time being in the value (or range) corresponding to the neuron.

[0056] The training process of the LSTM neural network involves offline hyperparameter tuning. Before inputting the data into the neural network, it is normalized as follows:

[0057]

[0058] where is the sample mean load of the input data. The normalized data is then split into 80% for training and 20% for testing.

[0059] The training of the LSTM neural network is achieved by backpropagation through time, i.e. the weights of the LSTM neural network are updated by minimizing a loss function L representing the difference between the output of the neural network and the actual data. The training is stopped when the loss cannot be further reduced, with the final monitoring of the root mean square error (RMSE).

[0060] After the training is completed, the trained LSTM network is used for online prediction.

[0061] S3, according to the strong burstiness and real-time nature of Y2(t), a statistical method is used to make the second load prediction data prediction, obtaining the second load prediction data.

[0062] Specifically, a hybrid statistical method is used to model, which combines three prediction factors.

[0063] Firstly, the Kernel Density Estimation (KDE) method is used to approximate its probability density function (PDF). This method assumes that the load distribution presents a concentration trend around the mean, i.e. the closer the load is to its mean value, the higher the corresponding probability density; on the contrary, when it is far away from the mean, the density gradually decreases. On this basis, a sampling value prediction method based on kernel probability density function is proposed to dynamically capture the time-varying characteristics of the base station load. However, since the probability density function of the base station load will be updated in real time with the continuous influx of new measurement data, it makes it challenging to accurately depict the real load distribution at a specific time point. Therefore, in order to improve the prediction accuracy and adaptability of the model, it is necessary to evaluate the performance of different kernel functions to select the optimal solution. After systematic research and comparative analysis, the results show that using the Epanechnikov kernel function for load measurement has a significant advantage over other methods. Specifically, for the probability density value of a given load x, the following formula is used to calculate:

[0064]

[0065] where x i is the i-th data point, n is the sample size, h is the bandwidth parameter, and K is the Epanechnikov kernel function. Using the kernel function, the probability density function of x as a continuous variable is obtained This function depends on n and the selection of the sample set (i.e. x1, x2,..., x n). To capture the trend of load variation over time, a sliding window technique is employed to extract data from the measurement trajectory for the calculation of To ensure that the most recent samples contain more valuable information, two prediction factors are calculated based on this: a short-term prediction factor y1 and a long-term prediction factor y2. These two prediction factors are defined according to the given sampling range, aiming to integrate the effects of short-term fluctuations and long-term trends, thereby improving the prediction accuracy. The short-term prediction factor mainly focuses on recent data patterns, while the long-term prediction factor considers more extensive historical information. This dual-prediction factor approach not only reflects the immediate change dynamics but also maintains an understanding of the overall trend, making the prediction model more robust and accurate.

[0066] However, the above sampling-based methods cannot fully capture the dependencies (i.e., autocorrelation between samples) in the collected load data. Therefore, an Exponentially Weighted Moving Average (EWMA) method is used to enhance them:

[0067] y i = y i-1 + a(x i - y i-1 )

[0068] where y i is the predicted value at time i, and a = 2n+1 is the weight parameter. Let y3 represent the prediction factor obtained by the EWMA method.

[0069] From the measurements, it can be observed that the accuracy of different prediction factors varies with the measurement period. For example, when the load fluctuates greatly, the short-term prediction factor outperforms the other two. This is because the recent load has a greater impact on the current load than the long-term load, and fluctuations further reduce the accuracy of the EWMA method. Since y1, y2, and y3 are relatively independent in prediction, dynamic weights (weights can be adjusted according to their relative importance) α, β, and γ can be used to combine these three prediction factors, and the prediction error can be reduced by optimizing the weight distribution. The final prediction value can be represented as:

[0070] y = αy1 + βy2 + γy3

[0071] where α + β + γ = 1, and all weights are non-negative. This dynamic weight distribution mechanism allows the model to flexibly adjust the contribution proportion of each prediction factor according to the actual situation, ensuring the best performance of the prediction result.

[0072] To improve the prediction accuracy in various dynamic environments, it is necessary to update the coefficients α, β, and γ over time. Specifically, at each time step i, first set a potential coefficient set containing N groups of coefficient vectors, and calculate the possible prediction values y i,1 ,y i,2..., y i,j ..., y i,J As follows:

[0073]

[0074] where (a n , b n , g n ) is the possible coefficient vector (a, b, g) at time i. Then, the error vector [E i,n , E i ,..., E i,1 ] between y i,2 and the actual delay x i,N is calculated, and n' e {1, 2,..., N} that minimizes E i,N is found:

[0075]

[0076] Finally, update {a i+1 , b i+1 , g i+1} = {a n′ , b n′ , g n′} for the next prediction.

[0077] S4, since the frequency of base station dormancy control update is much lower than that of load prediction update, the base station needs to predict the load of multiple future time points, model the base station dormancy control problem as MDP according to the first load prediction data and the second load prediction data of multiple future time points, solve the MDP, and perform base station dormancy control according to the solution.

[0078] The base station dormancy control problem is modeled as an MDP. The network controller that determines the base station dormancy control in the region is an agent that tries to solve the MDP. The system state in the MDP includes the interval in which the load of all base stations at the current time point is located, the channel state of all users (whether blocked), and the interval in which the transmission rate of all users is located. The action of the agent is the work / sleep decision of all base stations. The reward at each time point is defined as the negative value of the cost function at that time point. The purpose of the agent is to maximize the sum of the discounted rewards at the current time point and all subsequent time points.

[0079] The base station load at multiple future time points predicted by the above two methods is used as the state of the agent to perform deep reinforcement learning-based solution and formulate the base station dormancy strategy.

[0080] The base station load prediction device provided by the present application is described below, and the base station load prediction device described below can be correspondingly referred to the base station load prediction method described above.

[0081] The base station load prediction device provided by the present application is described below, and the base station load prediction device described below can be correspondingly referred to the base station load prediction method described above.

[0082] The base station load data acquisition module acquires the load of the base station in the past period and the current time according to the number of users entering the coverage of the base station, as the base station load history data.

[0083] The data preprocessing module is used for dividing the collected base station load history data into two parts of a long-term trend and a short-term random fluctuation Y1(t) and Y2(t) according to a specific law over time, and performing preprocessing operations such as cleaning and normalization on the original data.

[0084] The load prediction module includes two sub-modules.

[0085] The LSTM neural network prediction submodule is used for training the LSTM neural network according to Y1(t) to complete the first load prediction by the trained LSTM neural network, so as to obtain the first load prediction data and capture and predict the long-term trend part of the load.

[0086] The statistical method prediction submodule is used for performing the second load prediction by a proper statistical method (such as an exponential weighted moving average) according to Y2(t) to obtain the second load prediction data, and modeling and predicting the short-term random fluctuation part of the load.

[0087] The multi-step prediction integration module is used for integrating the first load prediction data and the second load prediction data of multiple future time points to generate the load prediction values of the multiple future time points, since the frequency of the base station sleep control update is much lower than the frequency of the load prediction update.

[0088] The base station sleep control module is used for modeling the base station sleep control problem as an MDP based on the integrated load prediction data and solving the MDP, so as to perform the base station sleep control.

[0089] The device provided by the present application is suitable for a multi-user multi-base station mobile communication system in all high-mobility scenarios.

[0090] The electronic device provided by the present application includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the method described in any of the above embodiments when executing the computer program.

[0091] The embodiment of the present application provides a base station load prediction system 300 combining deep learning and statistical methods, as shown in the figure, comprising a computer readable storage medium 320 and a processor 310. Figure 4

[0092] The computer readable storage medium is used for storing executable instructions.

[0093] The processor is used for reading the executable instructions stored in the computer readable storage medium, and performing the method described in any of the above embodiments.

[0094] Specifically, the processor 310 can include, for example, a general-purpose microprocessor, an instruction set processor and / or a related chipset, and / or a special-purpose microprocessor (for example, an application-specific integrated circuit (ASIC)), and the like. The processor 310 can also include on-board memory for cache use. The processor 310 can be a single processing unit for executing the method flow described according to the embodiments of the present disclosure or a plurality of processing units. Figures 1-3 The processor 310 can be a single processing unit for executing the method flow described according to the embodiments of the present disclosure or a plurality of processing units.

[0095] The computer readable storage medium 320, for example, can be any medium capable of containing, storing, communicating, propagating or transferring instructions. For example, the readable storage medium can include but is not limited to electrical, magnetic, optical, electromagnetic, infrared or semiconductor system, device, instrument or propagation medium. Specific examples of readable storage medium include magnetic storage devices such as magnetic tape or hard disk (HDD); optical storage devices such as compact disc (CD-ROM); memory such as random access memory (RAM) or flash memory; and / or wired / wireless communication links.

[0096] The computer readable storage medium 320 can include a computer program 321, which can include code / computer executable instructions that, when executed by the processor 310, cause the processor 310 to perform, for example, the method flow described in any of the above embodiments and any variations thereof.

[0097] The computer program 321 can be configured to have, for example, computer program code including computer program modules. For example, in an example embodiment, the code in the computer program 321 can include one or more program modules, for example, including modules 321A, 321B, … It should be noted that the division and number of modules are not fixed, and those skilled in the art can use appropriate program modules or program module combinations according to actual conditions, which, when executed by the processor 310, cause the processor 310 to perform, for example, the method flow described in any of the above embodiments and any variations thereof.

[0098] ​Those skilled in the art can easily understand that the above description is only the preferred embodiment of the present application, and is not intended to limit the present application. Any modification, equivalent replacement and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A base station load prediction method combining deep learning and statistical methods, characterized by, Comprising: S1, base station load history data is divided into two parts, a long-term trend and a short-term random fluctuation is the average load data in a sliding window, the difference between the two​​​​ S2, according to training the LSTM neural network to obtain a trained LSTM neural network to complete the first load prediction to obtain first load prediction data; S3, according to using statistical methods to make a second load prediction to obtain second load prediction data; comprising: R1, para Using kernel density estimation to obtain short-term prediction values and long-term prediction values ; R2, para Using a weighted moving average method to obtain the predicted value ; R3, using dynamic weights to combine the three prediction values to obtain a final prediction value, i.e., second load prediction data; S4, modeling the base station sleep control problem as MDP according to the first load prediction data and the second load prediction data of the future multiple time points, and solving the MDP, thereby performing base station sleep control.

2. The method of claim 1, wherein, The step S2 comprises: T1, determining the structure and initial parameters of the LSTM neural network; T2, pair Normalization is performed, and the data is divided into training data and test data for training and testing, respectively. T3, updating the weight values of the LSTM neural network by minimizing the loss function representing the difference between the output of the LSTM neural network and the actual data, and respectively using the training data and the test data to complete training and testing, thereby obtaining the first load prediction data.

3. The method of claim 2, wherein, The LSTM neural network structure is as follows: 1) the input of the input layer is the sequence in the current time and the previous sliding window; 2) an all-connection layer with ReLu as the activation function is followed, which is used for feature extraction; 3) a plurality of recursive layers based on the LSTM structure are followed, which are used for analyzing and identifying the relevance inside the input sequence; 4) the output layer is a softmax function, wherein the output value of each neuron represents the probability of the load at the next time being the value corresponding to the neuron.

4. The method of claim 1, wherein, In R3: wherein, y is the final prediction value, α、β and γ is the dynamic weight, α + β + γ = 1 and are non-negative values. 5.A base station load prediction device combining a deep learning method and a statistical method, characterized by, Comprising: A base station load data acquisition module, which acquires the load of the base station at a past period of time and the current time according to the number of users entering the coverage of the base station, as base station load history data; The data preprocessing module is used for dividing the collected base station load history data into two parts, i.e., a part following a long-term trend and a part fluctuating according to a specific rule over time and a short-term random fluctuation. ​​ A load prediction module, comprising two sub-modules: The LSTM neural network prediction submodule is configured to predict the first load according to the historical load data and the historical load prediction data. The LSTM neural network is trained to obtain a trained LSTM neural network, and the trained LSTM neural network is used to complete the first load prediction to obtain first load prediction data. The statistical methods prediction submodule is used to predict based on The second load forecast was performed using statistical methods to obtain the second load forecast data; including: [the following data is missing from the original text]. Short-term forecasts were obtained using kernel density estimation. and long-term forecasts ;right The predicted value was obtained using the weighted moving average method. The final predicted value, i.e., the second load prediction data, is obtained by combining the three prediction values ​​mentioned above with dynamic weights. A multi-step prediction integration module, which is used for integrating the first load prediction data and the second load prediction data of the future multiple time points; A base station sleep control module, which is used for modeling the base station sleep control problem as MDP based on the integrated load prediction data, and solving the MDP, thereby performing base station sleep control.

6. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor realizes the method of any one of claims 1 to 4 when executing the computer program.

Images