CN116258052B - Road loss prediction method, device and electronic equipment - Google Patents

Abstract

The invention provides a method and a device for predicting path loss and electronic equipment, and relates to the technical field of wireless communication, wherein the method comprises the following steps: generating a plurality of environment samples according to scene information of the wireless propagation environment; generating path loss simulation data and a plurality of environmental features corresponding to each environmental sample according to the plurality of environmental samples; training according to the environmental characteristics and the path loss simulation data to obtain a first model; changing the environmental characteristics input into the first model, analyzing the influence degree of different environmental characteristics on the path loss prediction result, and obtaining an importance analysis result; and selecting target environmental characteristics to train the first model according to the importance analysis result, and generating a target path loss prediction result. The scheme of the invention can be suitable for various application scenes and has the advantages of high precision and low calculation complexity.

Description

Road loss prediction method, device and electronic equipment

Technical Field

The present invention belongs to the technical field of wireless communications, and in particular relates to a path loss prediction method, device and electronic equipment.

Background Art

The mainstream road loss modeling methods currently include statistical modeling methods and deterministic modeling methods. Among them, statistical modeling methods, such as the ABG (Alpha-Beta-Gamma) model, use frequency and the distance between Tx (Transmitter) and Rx (Receiver) for modeling. It has the advantages of simple and fast calculation, but does not consider the scenario information of the actual application environment. The deterministic modeling method is an RT (Ray-Tracing) method implemented using detailed map information. It can provide more accurate prediction results, but it has the disadvantages of high computational complexity and high implementation cost.

Summary of the invention

The purpose of the embodiments of the present invention is to provide a path loss prediction method, device and electronic device, so as to solve the problem that the path loss prediction method in the prior art does not take the actual application environment into consideration and has high computational complexity.

In order to achieve the above object, an embodiment of the present invention provides a path loss prediction method, comprising:

Generate multiple environment samples according to the scene information of the wireless propagation environment;

According to the multiple environmental samples, generating road loss simulation data and multiple environmental features corresponding to each environmental sample;

Training is performed according to environmental characteristics and road loss simulation data to obtain a first model;

Changing the environmental characteristics input into the first model, analyzing the influence of different environmental characteristics on the road loss prediction result, and obtaining the importance analysis result;

According to the importance analysis result, target environment features are selected to train the first model to generate a target road loss prediction result.

Optionally, the method further comprises:

Analyze the accuracy and computational complexity of the target path loss prediction results.

Optionally, the accuracy and computational complexity of the path loss prediction results are analyzed, including:

Calculating the root mean square error of the target path loss prediction result;

Obtaining the training time and prediction time of the first model;

Analyzing the accuracy of the target path loss prediction result according to the root mean square error;

The computational complexity of the target path loss prediction result is analyzed according to the training duration and the prediction duration.

Optionally, multiple environment samples are generated according to scene information of the wireless propagation environment, including:

Generate multiple environment samples of preset space size according to the scene information;

Each environmental sample includes the physical positions and sizes of a plurality of scatterers, and different environmental samples include at least one item of the number, size and physical position of the scatterers that is different.

Optionally, generating path loss simulation data corresponding to each environment sample according to a plurality of environment samples includes:

By setting communication parameters and the physical positions of the transmitter and the receiver in the wireless propagation environment, channel simulation is performed on multiple environment samples respectively to generate the path loss simulation data.

Optionally, based on the multiple environment samples, multiple environment features corresponding to each environment sample are generated, including:

A plurality of environmental features are generated according to the quantity features, volume features, distance features, offset features and non-line-of-sight features of a plurality of scatterers in each environmental sample.

Optionally, after generating the path loss simulation data and the multiple environmental features corresponding to each environmental sample according to the multiple environmental samples, the method further includes:

Constructing a mapping data set between environmental features and the road loss simulation data;

Dividing the mapping data set into a training data set and a test data set;

Perform training according to the training data set to generate the first model;

The first model is tested using the test data set.

Optionally, changing the environmental features input into the first model, analyzing the influence of different environmental features on the road loss prediction result, and obtaining the importance analysis result includes:

Changing the environmental characteristics input into the first model to obtain a road loss prediction result;

The model interpretation method is used to analyze the influence of different environmental characteristics on the road loss prediction results, and the importance analysis results are obtained.

Optionally, the method further comprises:

Obtaining the correlation between each environmental feature and the road loss simulation data according to a cumulative distribution function between each environmental feature and the road loss simulation data;

According to the correlation, the target path loss prediction result corresponding to the target environmental feature is verified.

Optionally, the method further comprises:

Using a probability decay coefficient to reduce the participation of the environmental feature with the highest importance analysis result;

The target environmental feature is selected according to the participation degree of each environmental feature.

Optionally, according to the importance analysis result, selecting target environment features to train the first model to generate a target road loss prediction result includes:

The first model is trained using the target environment characteristics and the road loss simulation data corresponding to the target environment characteristics to generate the target road loss prediction result.

The embodiment of the present invention further provides a road loss prediction device, comprising:

A first generating module, used to generate a plurality of environment samples according to scene information of a wireless propagation environment;

A second generating module is used to generate road loss simulation data and multiple environmental features corresponding to each environmental sample according to multiple environmental samples;

A first acquisition module, used for training according to environmental characteristics and road loss simulation data to obtain a first model;

A second acquisition module is used to change the environmental characteristics input into the first model, analyze the influence of different environmental characteristics on the road loss prediction result, and obtain the importance analysis result;

The third generation module is used to select target environment features to train the first model according to the importance analysis result, and generate a target road loss prediction result.

Optionally, the device further comprises:

The analysis module is used to analyze the accuracy and computational complexity of the target path loss prediction result.

Optionally, the analysis module is specifically used for:

Calculating the root mean square error of the target path loss prediction result;

Obtaining the training time and prediction time of the first model;

Analyzing the accuracy of the target path loss prediction result according to the root mean square error;

The computational complexity of the target path loss prediction result is analyzed according to the training duration and the prediction duration.

Optionally, the first generating module is specifically used for:

Generate multiple environment samples of preset space size according to the scene information;

Each environmental sample includes the physical positions and sizes of a plurality of scatterers, and different environmental samples include at least one item of the number, size and physical position of the scatterers that is different.

Optionally, the second generating module is specifically used to:

By setting communication parameters and the physical positions of the transmitter and the receiver in the wireless propagation environment, channel simulation is performed on multiple environment samples respectively to generate the path loss simulation data.

Optionally, the second generating module is specifically used to:

A plurality of environmental features are generated according to the quantity features, volume features, distance features, offset features and non-line-of-sight features of a plurality of scatterers in each environmental sample.

Optionally, the device further comprises:

A construction module, used to construct a mapping data set between environmental features and the road loss simulation data;

A partitioning module, used for partitioning the mapping data set into a training data set and a test data set;

A training module, used for performing training according to the training data set to generate the first model;

A testing module is used to test the first model using the test data set.

Optionally, the second obtaining module is specifically used for:

Changing the environmental characteristics input into the first model to obtain a road loss prediction result;

The model interpretation method is used to analyze the influence of different environmental characteristics on the road loss prediction results, and the importance analysis results are obtained.

Optionally, the device further comprises:

An acquisition module, used for acquiring the correlation between each environmental feature and the road loss simulation data according to a cumulative distribution function between each environmental feature and the road loss simulation data;

A verification module is used to verify the target path loss prediction result corresponding to the target environmental feature according to the correlation.

Optionally, the device further comprises:

A reduction module, used for reducing the participation of the environmental feature with the highest importance analysis result by using a probability attenuation coefficient;

The selection module is used to select the target environmental feature according to the participation degree of each environmental feature.

Optionally, the third generating module is specifically used for:

The first model is trained using the target environment characteristics and the road loss simulation data corresponding to the target environment characteristics to generate the target road loss prediction result.

An embodiment of the present invention further provides an electronic device, comprising: a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the path loss prediction method as described above when executing the computer program.

An embodiment of the present invention further provides a readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the path loss prediction method as described above are implemented.

The above technical solution of the present invention has at least the following beneficial effects:

In the above scheme, multiple environmental samples are generated according to the scene information of the wireless propagation environment; based on the multiple environmental samples, path loss simulation data and multiple environmental features corresponding to each environmental sample are generated; training is performed according to the environmental features and the path loss simulation data to obtain a first model; the environmental features input into the first model are changed, and the degree of influence of different environmental features on the path loss prediction results is analyzed to obtain an importance analysis result; based on the importance analysis result, target environmental features are selected to train the first model to generate target path loss prediction results, and the characteristic information of the specific wireless propagation environment is directly mapped with the path loss simulation data to construct a high-precision and low-computational-complexity model, so as to efficiently and accurately predict the path loss data for specific application scenarios.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of a flow chart of a method for predicting a path loss according to an embodiment of the present invention;

FIG2 is a schematic diagram of a first model according to an embodiment of the present invention;

FIG3 is a block diagram of a road loss prediction device according to an embodiment of the present invention.

DETAILED DESCRIPTION

In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

The embodiments of the present invention provide a path loss prediction method, device and electronic device to address the problems that the path loss prediction method in the prior art does not take the actual application environment into consideration and has high computational complexity.

As shown in FIG1 , an embodiment of the present invention provides a path loss prediction method, including:

Step 101, generating a plurality of environment samples according to scene information of a wireless propagation environment;

It should be noted that the scene information of the real wireless propagation environment can be obtained through actual measurement, map assistance, and automatic collection by computer vision related technologies. The scene information includes but is not limited to: space size, physical location of scatterers, size of scatterers, Tx (Transmiter) location, and Rx (Receiver) location.

It should also be noted that the spatial size of the wireless propagation environment is the size of the effective area between the transmitter and the receiver.

Step 102, generating road loss simulation data and multiple environmental features corresponding to each environmental sample according to the multiple environmental samples;

Here, a ray tracing channel simulation tool is used to generate channel parameters corresponding to each environment sample, that is, to generate path loss simulation data.

In addition, since radio waves encounter scatterers of different sizes at different physical locations and produce different propagation ray combinations, the factors affecting the path loss caused by the distribution of scatterers in different environmental samples under the wireless propagation mechanism are analyzed to obtain environmental characteristics.

It should be noted that the embodiment of the present invention directly maps the characteristic information of a specific propagation environment with the path loss simulation data, thereby alleviating the dependence on the propagation scenario type and supporting the path loss prediction of a specific propagation environment.

Step 103, training is performed according to environmental characteristics and road loss simulation data to obtain a first model;

It should be noted that environmental features are used as input and road loss simulation data as output to pre-train the preset model and obtain the first model. Here, the preset model is the RF (Radom Forest) model. RF is an integrated classifier algorithm that constructs multiple decision trees and merges them to obtain better output results. It is mainly trained through Bootstrap (a statistical method) and random feature subspace. Bootstrap is used to randomly select samples from the data set with replacement, and the feature subspace is used for random feature selection. When a new input sample is passed in for prediction, the sample will be passed to each decision tree in the forest, and the average value of these trees will be taken as the output result. Since multiple decision trees can be trained at the same time, the RF model has a high degree of effective parallelization.

It should also be noted that by comparing different machine learning methods, the most suitable machine learning algorithm is selected under the premise of balancing accuracy and computational complexity. When the accuracy difference is not large, the RF model is selected as the preset model because the training time of the RF model is much shorter than the prediction model based on network training.

Step 104, changing the environmental features input into the first model, analyzing the influence of different environmental features on the road loss prediction result, and obtaining the importance analysis result;

Here, the SHARP (Shapely Additive Explanation) method is used to observe the influence of the input into the first model when the environmental characteristics change on the path loss prediction result output by the first model.

Step 105: According to the importance analysis result, target environment features are selected to train the first model to generate a target road loss prediction result.

Here, it should be noted that the first model is improved through the importance analysis result to obtain a road loss prediction model, so that the target environment characteristics are input into the road loss prediction model to output the target road loss prediction result.

The embodiment of the present invention generates multiple environmental samples according to the scene information of the wireless propagation environment; generates path loss simulation data and multiple environmental features corresponding to each environmental sample according to the multiple environmental samples; performs training according to the environmental features and the path loss simulation data to obtain a first model; changes the environmental features input into the first model, analyzes the degree of influence of different environmental features on the path loss prediction result, and obtains an importance analysis result; selects target environmental features according to the importance analysis result to train the first model, generates target path loss prediction results, directly maps the characteristic information of the specific wireless propagation environment with the path loss simulation data, and constructs a model with high precision and low computational complexity, so as to efficiently and accurately predict the path loss data for specific application scenarios.

Optionally, the path loss prediction method of the embodiment of the present invention further includes:

Analyze the accuracy and computational complexity of target path loss prediction results.

Here, the first model is improved through the importance analysis results to obtain a road loss prediction model, so that the target environmental characteristics are input into the road loss prediction model and the target road loss prediction result is output; the target road loss prediction result output by the road loss prediction model is compared with the road loss prediction results output by other models to analyze the accuracy and computational complexity of the target road loss prediction result.

It should be noted that other models include but are not limited to the ABG model (Alpha-Beta-Gamma, a statistical path loss model), the RT model, the unimproved RF model and other neural network models, such as RNN, CNN, SVM, Boosting, Bagging, etc.

Optionally, the accuracy and computational complexity of the target path loss prediction result are analyzed, including:

Calculate the root mean square error of the target road loss prediction result;

Get the training time and prediction time of the first model;

Analyze the accuracy of target road loss prediction results based on the root mean square error;

According to the training time and prediction time, the computational complexity of the target path loss prediction results is analyzed.

It should be noted that the accuracy of the target path loss prediction result is analyzed by comparing the root mean square error of the target path loss prediction result of the embodiment of the present invention with the path loss prediction result output by other models. Here, at the first communication frequency (6GHz) and the second communication frequency (28GHz), the root mean square error is calculated respectively based on the target path loss prediction result of the embodiment of the present invention and the path loss prediction result output by other models, as well as the corresponding path loss simulation data, and a comparative analysis is performed. The root mean square error of the target path loss prediction result of the embodiment of the present invention at the first communication frequency is 3.15dB, and the root mean square error of the target path loss prediction result at the second communication frequency is 3.14dB, which is improved compared with the ABG model.

It should also be noted that the road loss prediction results of other models are also trained using the environmental characteristics and road loss simulation data in the embodiment of the present invention to output road loss prediction results.

Furthermore, by comparing the training time and prediction time of the first model of the embodiment of the present invention with other models, the computational complexity of the target path loss prediction result is obtained. After comparative analysis, the training time of the first model of the embodiment of the present invention is about 1.3s, and the prediction time is about 0.004s, which are shorter than those of the RT model.

Optionally, step 101, generating a plurality of environment samples according to scene information of a wireless propagation environment, includes:

Generate multiple environment samples of preset space size according to scene information;

Each environmental sample includes the physical positions and sizes of a plurality of scatterers, and different environmental samples include at least one item of the number, size and physical position of the scatterers that is different.

It should be noted that the preset space size is determined according to the scene information, and is generally the same as the space size corresponding to the scene information. A 3D modeling tool, such as Blender or Google SketchUp, is used to generate multiple environment samples with the same space size as that corresponding to the scene information.

Taking the lower left corner of the wireless propagation environment as the origin, set the coordinates of Tx to (x _t ,y _t ,z _t ) and the coordinates of Rx to (x _r ,y _r ,z _r ). Since multiple environment samples have different numbers of scatterers, and these scatterers have different positions and sizes, here, the coordinates of the center point of the scatterer and the length, width, and height are set to (x _j ,y _j ,z _j ), l _j ,w _j ,z _j ; the length, width, and height of the scatterer are randomly generated within a certain range. The generation principle is to randomly generate rectangular scatterers inside the wireless propagation environment, excluding the boundary of the wireless propagation environment and the positions of Tx and Rx, and there is no overlap between different scatterers.

Taking the spatial size of the wireless propagation environment of 15×10×3m ³ as an example, a total of 5408 environmental samples are generated, of which 4566 environmental samples are used to generate training data sets for model training. These 4566 environmental samples are samples with 3, 5, 6, 7, 9, 10 and 11 random scatterers; 842 environmental samples are used to generate test data sets for model testing. These 842 environmental samples are samples with 4, 8 and 12 random scatterers.

Optionally, step 102, generating path loss simulation data corresponding to each environment sample according to multiple environment samples, includes:

By setting the communication parameters and the physical locations of Tx and Rx in the wireless propagation environment, channel simulation is performed on multiple environmental samples to generate path loss simulation data.

Here, multiple environment samples are input into the WirelessInsite simulation software one by one, and channel simulation is performed by setting the same communication parameters, Tx and Rx physical positions to generate path loss simulation data. The communication parameters include the first communication frequency (6GHz) and the second communication frequency (28GHz), and the path loss simulation data under the two communication frequencies are generated accordingly.

Optionally, based on the multiple environment samples, multiple environment features corresponding to each environment sample are generated, including:

A plurality of environmental features are generated according to the quantity features, volume features, distance features, offset features and NLOS (None-Line-of-Sight) features of a plurality of scatterers in each environmental sample.

It should be noted that the quantitative characteristics describe the total number of scatterers existing in the current wireless propagation environment;

The volume feature Vol describes the ratio of all scatterers to the overall area volume in the current wireless propagation environment and is calculated using the following formula:

Wherein, M represents the number of scatterers in the current wireless propagation environment; V _area represents the volume of the entire area of the wireless propagation environment;

The distance feature Dist describes the distance component from all scatterers to Tx in the current wireless propagation environment and is calculated using the following formula:

The offset feature describes the offset of all scatterers relative to the line between Tx and Rx in the current wireless propagation environment, that is, the sum of the distances from the center points of all scatterers to the line between Tx and Rx;

The NLOS feature describes the degree of obstruction of the LOS (Line-of-Sight) path by the scatterer in the current wireless propagation environment. It can be divided into two types: obstruction and non-obstruction. If there is no obstruction, the feature representing the obstruction is directly set to 0; if there is obstruction, the degree of obstruction of the LOS path by the scatterer is used to represent the obstruction factor of each scatterer. Specifically, when the height of the scatterer is higher than the height of Tx, the ratio of the distance from the center point of the scatterer to the line connecting Tx and Rx to half of the diagonal of the scatterer surface is calculated. The largest obstruction factor in the current wireless propagation environment is selected as the NLOS feature of the wireless propagation environment.

Optionally, in step 102, after generating the road loss simulation data and multiple environmental features corresponding to each environmental sample according to the multiple environmental samples, the road loss prediction method of the embodiment of the present invention further includes:

Construct a mapping data set between environmental characteristics and road damage simulation data;

Divide the mapping dataset into a training dataset and a test dataset;

Perform training according to the training data set to generate a first model;

The first model is tested using the test data set.

Here, mapping data sets between environmental characteristics and path loss simulation data at two communication frequencies, the first communication frequency (6GHz) and the second communication frequency (28GHz), are constructed respectively, and training data sets and test data sets are extracted from the mapping data sets for training and testing of the first model, where the environmental characteristics serve as the input of the first model and the path loss simulation data serve as the output of the first model.

Optionally, step 104, changing the environmental features input into the first model, analyzing the influence of different environmental features on the road loss prediction result, and obtaining the importance analysis result, includes:

Changing the environmental characteristics input into the first model to obtain a road loss prediction result;

The model interpretation method is used to analyze the influence of different environmental characteristics on the road loss prediction results and obtain the importance analysis results.

Here, the environmental characteristics are changed and input into the first model for training. The SHARP method is used to observe the influence of each environmental characteristic on the output path loss prediction result when it is changed, and the importance analysis result is obtained.

Optionally, the path loss prediction method of the embodiment of the present invention further includes:

According to the CDF (Cumulative Distribution Function) between each environmental feature and the road loss simulation data, the correlation between each environmental feature and the road loss simulation data is obtained;

According to the correlation, the target road loss prediction results corresponding to the target environmental characteristics are verified.

It should be noted that by plotting the CDF between each environmental feature and the road loss simulation data, the correlation between each environmental feature and the road loss simulation data is obtained. Specifically, the environmental characteristics and road loss simulation data corresponding to the environmental samples including 5 scatterers under the second communication frequency are selected to observe the influence trend of each environmental feature on the road loss simulation data, thereby obtaining the correlation.

Then, verify whether the target environmental characteristics and the corresponding target road loss prediction results meet the correlation, which helps to determine the accuracy of the target road loss prediction results.

Optionally, the path loss prediction method of the embodiment of the present invention further includes:

The probability decay coefficient is used to reduce the participation of the environmental features with the highest importance analysis results;

Select target environmental features based on the participation level of each environmental feature.

It should be noted that according to the importance analysis results corresponding to the environmental features, the participation is balanced to weaken the phenomenon that the importance of NLOS features is too high due to discontinuous feature values. Therefore, a probability attenuation coefficient ρ∈(0,1) is introduced, that is, when the NLOS feature is selected in the process of random selection of environmental features by the decision tree in RF, the NLOS feature is replaced with the unselected environmental features with a probability of ρ, thereby reducing the participation of environmental features with high importance. Then, according to the participation of each environmental feature, the target environmental feature is determined in the process of random selection of environmental features by the decision tree, that is, among multiple environmental features, the environmental features with high importance are eliminated to obtain the target environmental feature.

Furthermore, the pre-trained RF model is trained according to the target environmental features to obtain a road loss prediction model, that is, the improved RF model shown in FIG2 , which includes k decision trees. S is a data set containing N environmental features and M environmental samples. The data set S is divided into k subsets, namely {S ₁ ,S ₂ …S _k }. For each decision tree, a data set of n target environmental features (determined according to the probability attenuation coefficient) and m environmental samples is randomly selected. Finally, the results {y ₁ ,y ₂ …y _k } output by each decision tree are averaged as the final target road loss prediction result.

Optionally, step 105, according to the importance analysis result, selects the target environment feature to train the first model to generate the target road loss prediction result, including:

The first model is trained using target environment characteristics and road loss simulation data corresponding to the target environment characteristics to generate a target road loss prediction result.

As shown in FIG3 , an embodiment of the present invention further provides a road loss prediction device, including:

A first generating module 301 is used to generate a plurality of environment samples according to scene information of a wireless propagation environment;

The second generating module 302 is used to generate the road loss simulation data and multiple environmental features corresponding to each environmental sample according to the multiple environmental samples;

A first obtaining module 303 is used to perform training according to environmental characteristics and road loss simulation data to obtain a first model;

The second obtaining module 304 is used to change the environmental features input into the first model, analyze the influence of different environmental features on the road loss prediction result, and obtain the importance analysis result;

The third generating module 305 is used to select target environment features to train the first model according to the importance analysis result, and generate a target road loss prediction result.

The embodiment of the present invention generates multiple environmental samples according to the scene information of the wireless propagation environment; generates path loss simulation data and multiple environmental features corresponding to each environmental sample according to the multiple environmental samples; performs training according to the environmental features and the path loss simulation data to obtain a first model; changes the environmental features input into the first model, analyzes the degree of influence of different environmental features on the path loss prediction result, and obtains an importance analysis result; selects target environmental features according to the importance analysis result to train the first model, generates target path loss prediction results, directly maps the characteristic information of the specific wireless propagation environment with the path loss simulation data, and constructs a model with high precision and low computational complexity, so as to efficiently and accurately predict the path loss data for specific application scenarios.

Optionally, the road loss prediction device of the embodiment of the present invention further includes:

The analysis module is used to analyze the accuracy and computational complexity of the target path loss prediction results.

Optionally, the analysis module is specifically used for:

Calculate the root mean square error of the target road loss prediction result;

Get the training time and prediction time of the first model;

Analyze the accuracy of target road loss prediction results based on the root mean square error;

According to the training time and prediction time, the computational complexity of the target path loss prediction results is analyzed.

Optionally, the first generating module 301 is specifically used for:

Generate multiple environment samples with preset space size according to scene information;

Each environmental sample includes the physical positions and sizes of a plurality of scatterers, and different environmental samples include at least one item of the number, size and physical position of the scatterers that is different.

Optionally, the second generating module 302 is specifically configured to:

By setting the communication parameters and the physical locations of Tx and Rx in the wireless propagation environment, channel simulation is performed on multiple environmental samples to generate path loss simulation data.

Optionally, the second generating module 302 is specifically configured to:

A plurality of environmental features are generated according to the quantity features, volume features, distance features, offset features and non-line-of-sight features of a plurality of scatterers in each environmental sample.

Optionally, the road loss prediction device of the embodiment of the present invention further includes:

A construction module, used to construct a mapping data set between environmental features and road loss simulation data;

A partitioning module, used for partitioning the mapping data set into a training data set and a test data set;

A training module, used for training according to a training data set to generate a first model;

The testing module is used to test the first model using a testing data set.

Optionally, the second obtaining module 304 is specifically used for:

Changing the environmental characteristics input into the first model to obtain a road loss prediction result;

The model interpretation method is used to analyze the influence of different environmental characteristics on the road loss prediction results and obtain the importance analysis results.

Optionally, the road loss prediction device of the embodiment of the present invention further includes:

An acquisition module, used for acquiring the correlation between each environmental feature and the road loss simulation data according to the CDF between each environmental feature and the road loss simulation data;

The verification module is used to verify the target road loss prediction result corresponding to the target environmental characteristics according to the correlation.

Optionally, the road loss prediction device of the embodiment of the present invention further includes:

A reduction module is used to reduce the participation of the environmental feature with the highest importance analysis result by adopting a probability attenuation coefficient;

The selection module is used to select the target environmental feature according to the participation degree of each environmental feature.

Optionally, the third generating module 305 is specifically used for:

The first model is trained using target environment characteristics and road loss simulation data corresponding to the target environment characteristics to generate a target road loss prediction result.

It should be noted that the embodiment of the road loss prediction device of the present invention is a device corresponding to the embodiment of the above-mentioned road loss prediction method. All implementation means in the above-mentioned road loss prediction method embodiment are applicable to the embodiment of the road loss prediction device and can achieve the same technical effect.

An embodiment of the present invention further provides an electronic device, comprising: a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the path loss prediction method as described above when executing the computer program.

An embodiment of the present invention further provides a readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the path loss prediction method as described above are implemented.

The processor is the processor of the electronic device in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

In addition, the present invention may repeat reference numerals and/or letters in different examples. This repetition is for the purpose of simplicity and clarity, and does not itself indicate the relationship between the various embodiments and/or settings discussed.

It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusions.

The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims (12)

1. The path loss prediction method is characterized by comprising the following steps of:

generating a plurality of environment samples according to scene information of the wireless propagation environment;

generating path loss simulation data and a plurality of environmental features corresponding to each environmental sample according to the plurality of environmental samples;

Training according to the environmental characteristics and the path loss simulation data to obtain a first model;

changing the environmental characteristics input into the first model, analyzing the influence degree of different environmental characteristics on the path loss prediction result, and obtaining an importance analysis result;

according to the importance analysis result, selecting a target environment characteristic to train the first model, and generating a target path loss prediction result;

Generating path loss simulation data corresponding to each environmental sample according to a plurality of environmental samples, wherein the path loss simulation data comprises:

Channel simulation is respectively carried out on a plurality of environment samples by setting communication parameters and physical positions of a transmitter and a receiver in the wireless propagation environment, so as to generate the path loss simulation data;

generating a plurality of environmental features corresponding to each environmental sample from the plurality of environmental samples, including:

Generating a plurality of environmental features according to the number features, the volume features, the distance features, the offset features and the non-line-of-sight features of a plurality of scatterers in each environmental sample;

according to the importance analysis result, selecting a target environment characteristic to train the first model, and generating a target path loss prediction result, wherein the method comprises the following steps:

The first model is improved according to the importance analysis result, and a path loss prediction model is obtained;

and inputting the target environmental characteristics into the path loss prediction model, and outputting the target path loss prediction result.

2. The method of path loss prediction according to claim 1, further comprising:

and analyzing the accuracy and the computational complexity of the target path loss prediction result.

3. The method of claim 2, wherein analyzing the accuracy and computational complexity of the target path loss prediction result comprises:

calculating the root mean square error of the target path loss prediction result;

acquiring training time and prediction time of the first model;

Analyzing the accuracy of the target path loss prediction result according to the root mean square error;

And analyzing the calculation complexity of the target path loss prediction result according to the training time length and the prediction time length.

4. The method of claim 1, wherein generating a plurality of environmental samples from scene information of a wireless propagation environment comprises:

generating a plurality of environment samples with preset space size according to the scene information;

Wherein each environmental sample includes physical locations and sizes of a plurality of scatterers, and different environmental samples include at least one of a number, size, and physical location of the scatterers that are different.

5. The method of claim 1, wherein after generating the path loss simulation data and the plurality of environmental features corresponding to each environmental sample from the plurality of environmental samples, the method further comprises:

Constructing a mapping data set between the environmental characteristics and the path loss simulation data;

Dividing the mapping data set into a training data set and a test data set;

Training according to the training data set to generate the first model;

and testing the first model by adopting the test data set.

6. The route prediction method according to claim 1, wherein changing the environmental characteristics inputted into the first model, analyzing the degree of influence of different environmental characteristics on the route prediction result, and obtaining the importance analysis result, comprises:

changing the environmental characteristics input into the first model to obtain a path loss prediction result;

and analyzing the influence degree of different environmental characteristics on the path loss prediction result by adopting a model interpretation method to obtain the importance analysis result.

7. The method of path loss prediction according to claim 1, further comprising:

acquiring the correlation between each environmental feature and the path loss simulation data according to the accumulated distribution function between each environmental feature and the path loss simulation data;

And verifying the target path loss prediction result corresponding to the target environmental characteristic according to the correlation.

8. The method of path loss prediction according to claim 1, further comprising:

the probability attenuation coefficient is adopted, so that the participation degree of the environmental characteristic with the highest importance analysis result is reduced;

and selecting the target environmental characteristics according to the participation degree of each environmental characteristic.

9. The method of claim 1, wherein selecting a target environmental feature to train the first model according to the importance analysis result, and generating a target path loss prediction result comprises:

and training the first model by adopting the target environment characteristics and the path loss simulation data corresponding to the target environment characteristics to generate the target path loss prediction result.

10. A path loss prediction apparatus, comprising:

the first generation module is used for generating a plurality of environment samples according to the scene information of the wireless propagation environment;

The second generation module is used for generating path loss simulation data and a plurality of environmental characteristics corresponding to each environmental sample according to the plurality of environmental samples;

The first obtaining module is used for training according to the environmental characteristics and the path loss simulation data to obtain a first model;

the second obtaining module is used for changing the environmental characteristics input into the first model, analyzing the influence degree of different environmental characteristics on the path loss prediction result and obtaining an importance analysis result;

The third generation module is used for selecting target environmental characteristics to train the first model according to the importance analysis result, and generating a target path loss prediction result;

The second generating module is specifically configured to:

Channel simulation is respectively carried out on a plurality of environment samples by setting communication parameters and physical positions of a transmitter and a receiver in the wireless propagation environment, so as to generate the path loss simulation data;

Generating a plurality of environmental features according to the number features, the volume features, the distance features, the offset features and the non-line-of-sight features of a plurality of scatterers in each environmental sample;

The third generating module is specifically configured to:

The first model is improved according to the importance analysis result, and a path loss prediction model is obtained;

and inputting the target environmental characteristics into the path loss prediction model, and outputting the target path loss prediction result.

11. An electronic device, comprising: transceiver, memory, processor and computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the path loss prediction method according to any of claims 1 to 9 when the computer program is executed.

12. A readable storage medium having stored thereon a computer program, which when executed by a processor, implements the steps of the path loss prediction method according to any one of claims 1 to 9.