CN105430664B

CN105430664B - It is a kind of to be fitted the method and apparatus that path loss is propagated in prediction based on classification

Info

Publication number: CN105430664B
Application number: CN201510726700.3A
Authority: CN
Inventors: 李小龙; 何峰; 闵冬; 高源�; 王灿
Original assignee: Shanghai Huawei Technologies Co Ltd
Current assignee: Shanghai Huawei Technologies Co Ltd
Priority date: 2015-10-30
Filing date: 2015-10-30
Publication date: 2019-05-28
Anticipated expiration: 2035-10-30
Also published as: CN105430664A

Abstract

It is a kind of to be fitted the method and apparatus that path loss is propagated in prediction based on classification, to improve the efficiency of creation propagation model, the accuracy that path loss is propagated in prediction is improved, is facilitated according to the corresponding propagation model of propagation characteristic Adaptive matching.In some feasible embodiments, method includes: to calculate electromagnetic wave propagation feature between transmitted from transmitter to receiver according to measured data；Measured data is clustered using at least one propagation characteristic, and constructs random forest grader；Non- ginseng fitting is carried out to every a kind of measured data that cluster obtains and generates corresponding propagation model；The propagation path loss between transmitter and receiver is calculated using the corresponding propagation model of random forest grader Adaptive matching according to the propagation characteristic between transmitted from transmitter to receiver.

Description

Method and device for predicting propagation path loss based on classification fitting

Technical Field

The invention relates to the technical field of communication, in particular to a method and a device for predicting propagation path loss based on classification fitting.

Background

The coverage evaluation of the wireless network through a planning tool is one of important means for planning and optimizing the high-quality wireless network. The simulation accuracy of basic indexes (such as level values, interference and speed) of various generations of communication technologies (such as 2G/3G/4G) is the key influencing the accuracy of coverage evaluation. The most central part of the simulation of each basic index is the calculation of the wireless electromagnetic wave propagation path loss between a transmitter and a receiver. The industry mainly depends on a Propagation Model (Propagation Model) to estimate Propagation path loss, the Propagation Model is a mathematical Model obtained through a large amount of data or a calculation method, and the generation process of the Model is called as fitting.

Propagation models are divided into two categories according to implementation: deterministic propagation models, empirical (statistical) propagation models. The deterministic propagation model is represented by a ray tracing technology, and the empirical (statistical) propagation model is mainly based on specific empirical formulas such as HATA, Okumura and LEE. Due to the diversity and complexity of planning scenes, both deterministic propagation models and empirical (statistical) propagation models need to be corrected according to actual scene data so that the propagation models can be used for predicting the network coverage.

Practice shows that the adopted ray tracing technology has the defects of large calculation amount and low efficiency for the deterministic propagation model; moreover, the propagation model obtained by the ray tracing technology generally adopts a propagation model in a planning region, and the propagation model cannot be adaptively adapted to the corresponding propagation model according to propagation characteristics, so that the propagation model is not high in applicability. The method has the defects of low simulation accuracy, low data effectiveness utilization rate of propagation model correction and the like for an empirical (statistical) propagation model.

Disclosure of Invention

The embodiment of the invention provides a method and a device for predicting propagation path loss based on classification fitting, which are used for improving the efficiency of creating a propagation model, improving the accuracy of predicting the propagation path loss and facilitating the self-adaptive matching of a corresponding propagation model according to propagation characteristics.

The first aspect of the present invention provides a method for predicting propagation path loss based on classification fitting, including:

calculating the propagation characteristics of electromagnetic waves between a transmitter and a receiver according to the measured data; clustering the measured data by using at least one propagation characteristic, and constructing a random forest classifier; carrying out non-parametric fitting on each type of measured data obtained by clustering to generate a corresponding propagation model; and according to the propagation characteristics from the transmitter to the receiver, calculating the propagation path loss between the transmitter and the receiver by utilizing the random forest classifier to adaptively match the corresponding propagation model.

In a first possible implementation manner, before the calculating propagation characteristics of the electromagnetic wave between the transmitter and the receiver according to the measured data, the method further includes: and cleaning and filtering the acquired measured data by adopting one or two of a conditional filtering method and an arithmetic filtering method.

With reference to the first aspect or the first possible implementation manner of the first aspect, in a second possible implementation manner, the propagation feature includes one or more of the following features: average inter-station distance, ratio of land features in coverage area of a cell, effective transmitter height, shielding influence of land features on a path from a transmitter to a receiver, receiving distance, average building height, density of buildings around the receiver, relative azimuth angle of a horizontal plane, relative declination angle of a vertical plane, and diffraction loss.

With reference to the first aspect or the first or second possible implementation manner of the first aspect, in a third possible implementation manner, the clustering measured data by using at least one propagation feature includes: selecting at least one propagation characteristic from the propagation characteristics as a classification fitting condition according to the correlation between the propagation characteristics and the propagation path loss; and clustering the measured data by using the selected at least one propagation characteristic.

With reference to the first aspect or any one of the first to third possible implementation manners of the first aspect, in a fourth possible implementation manner, before adaptively matching, by using the random forest classifier, a corresponding propagation model according to propagation characteristics between a transmitter and a receiver, the method further includes: and correcting the propagation model which does not meet the theoretical trend according to the propagation characteristics.

A second aspect of the present invention provides an apparatus for predicting propagation path loss based on classification fitting, including: the characteristic calculation module is used for calculating the propagation characteristics of the electromagnetic waves between the transmitter and the receiver according to the measured data; the cluster processing module is used for clustering the measured data by utilizing at least one propagation characteristic and constructing a random forest classifier; the model generation module is used for carrying out non-parametric fitting on each type of measured data obtained by clustering to generate a corresponding propagation model; and the matching calculation module is used for calculating the propagation path loss between the transmitter and the receiver by utilizing the random forest classifier to adaptively match the corresponding propagation model according to the propagation characteristics between the transmitter and the receiver.

In a first possible implementation manner, the apparatus further includes: and the cleaning and filtering module is used for cleaning and filtering the acquired measured data by adopting one or two of a conditional filtering method and an arithmetic filtering method.

With reference to the second aspect or the first possible implementation manner of the second aspect, in a second possible implementation manner, the propagation feature includes one or more of the following features: average inter-station distance, ratio of land features in coverage area of a cell, effective transmitter height, shielding influence of land features on a path from a transmitter to a receiver, receiving distance, average building height, density of buildings around the receiver, relative azimuth angle of a horizontal plane, relative declination angle of a vertical plane, and diffraction loss.

With reference to the second aspect or the first or second possible implementation manner of the second aspect, in a third possible implementation manner, the apparatus further includes: the characteristic selection module is used for selecting at least one propagation characteristic from the propagation characteristics as a classification fitting condition according to the correlation between the propagation characteristics and the propagation path loss; the clustering module is specifically configured to cluster the measured data using the at least one propagation feature selected by the feature selection module.

With reference to the second aspect or any one of the first to third possible implementation manners of the second aspect, in a fourth possible implementation manner, the model generating module is further configured to correct a propagation model that does not satisfy a theoretical trend according to the propagation characteristics.

A third aspect of the invention provides a computer device comprising a processor, a memory, a bus and a communication interface; the memory is configured to store computer executable instructions, and the processor is connected to the memory through the bus, and when the computer device is running, the processor executes the computer executable instructions stored in the memory, so as to cause the computer device to perform the method for predicting propagation path loss based on class fitting according to the first aspect of the present invention.

As can be seen from the above, in some possible embodiments of the present invention, the propagation characteristics are calculated from the measured data; clustering by using the propagation characteristics, and constructing a random forest classifier; performing non-parametric fitting on each type obtained by clustering to generate a corresponding propagation model; according to the propagation characteristics from the transmitter to the receiver, the technical characteristics of the propagation path loss between the transmitter and the receiver are calculated by utilizing the random forest classifier to adaptively match the corresponding propagation model, and the following technical effects are achieved:

1. the propagation characteristics are utilized to cluster the measured data, so that the utilization rate of the measured data is improved, and the effectiveness of a large amount of measured data in propagation path loss prediction is higher;

2. based on the clustering result, a non-parametric fitting method is adopted to perform classification fitting on the measured data, so that the calculated amount is reduced, and the efficiency is improved;

3. the propagation model obtained by carrying out classification fitting on a large amount of actually measured data based on propagation characteristic clustering and non-parametric fitting has higher prediction precision, so that the prediction result is more accurate and more effective;

4. the constructed random forest classifier can adaptively match a corresponding propagation model between a transmitter and a receiver according to propagation characteristics, and the applicability of the propagation model is improved.

5. The method of the embodiment of the invention improves the speed, the precision and the accuracy of the propagation path loss prediction, solves or reduces various defects of the prior art, and can ensure the accuracy and the rationality of the wireless network planning.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the embodiments and the drawings used in the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.

FIG. 1 is a schematic diagram of a communication system in an embodiment of the invention;

FIG. 2 is a flow chart of a method for predicting the propagation path loss of an electromagnetic wave;

FIG. 3 is a flow chart of a method of predicting propagation path loss based on classification fitting according to an embodiment of the present invention;

FIG. 4 is a graph of the effect of the algorithmically filtered propagation data;

FIG. 5 is a flow chart of the calculation of the average inter-site distance;

FIG. 6 is a schematic diagram of the matching of ground feature types;

FIG. 7a is a schematic diagram of an antenna covering a vertical cross-section;

FIG. 7b is a schematic view of a horizontal plane corresponding to FIG. 7 a;

FIG. 7c is a schematic view of the angle between OC and true north;

FIG. 8 is a schematic diagram of the effective transmitter height;

FIG. 9a is a schematic view of an NLos scene;

FIG. 9b is a schematic view of a first Fresnel zone;

FIG. 10 is a flow chart of the calculation of the average height of the building;

FIG. 11 is a two-dimensional projection effect plot of a first Fresnel zone;

FIG. 12 is a schematic diagram of the horizontal lobe relative azimuth;

FIG. 13 is a schematic diagram of the vertical lobe relative azimuth;

FIG. 14 is a schematic illustration of the effects of a non-parametric fit;

FIG. 15 is a graph comparing the fitting effect of the non-parametric fit GAM algorithm and the least squares algorithm;

FIG. 16 is a graph showing the trend of the fit effect of a model at a local point;

FIG. 17 is a graph showing the comparison of the fitting effect of the classification fitting model with the deterministic and empirical models;

FIG. 18 is a schematic structural diagram of an apparatus for predicting propagation path loss based on classification fitting according to an embodiment of the present invention;

fig. 19 is a schematic structural diagram of a computer device according to an embodiment of the present invention.

Detailed Description

In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

The technical scheme of the embodiment of the invention is applied to a communication system comprising a transmitter and a receiver, and the communication system can belong to any one of 2G, 3G, 4G and other communication systems. Wherein, 2G is the English abbreviation of the second-Generation mobile phone communication technology specification (English full name: 2-Generation Wireless technology), 3G is the English abbreviation of the third-Generation mobile communication technology specification (English full name: 3rd-Generation), and 4G is the English abbreviation of the 4 th-Generation mobile communication technology.

As shown in fig. 1, the communication system 100 according to the embodiment of the present invention may include, for example, a transmitter 110 and a receiver 120, wherein the transmitter 110 may be, for example, various base stations (eNode B), including a general base station, a micro base station, an indoor base station, a Home base station (Home eNode B), and the like; the receiver 120 may be, for example, a mobile station (mobile station), a mobile phone, or a base station. Wherein the transmitter 110 may also be referred to as a station and the receiver 120 may also be referred to as a receiving point.

The technical scheme of the embodiment of the invention is used for realizing the prediction of the electromagnetic wave propagation path loss between the transmitter and the receiver. The prediction of the propagation path loss of the electromagnetic wave is essentially the prediction of the severity of the influence of the propagation characteristics of the transmitter to the receiver on the transmission of the electromagnetic wave.

As shown in fig. 2, it is a flowchart of a method for predicting the propagation path loss of an electromagnetic wave in the conventional technology, and includes the following processes:

setp 1: fitting/correcting a propagation model suitable for the planning area according to the measured data;

optionally, before this step, a step of filtering and cleaning the measured data may be further included.

Setp 2: matching a proper propagation model from the transmitter to the receiver according to the propagation environment characteristics;

setp 3: the propagation path loss between the transmitter to the receiver is calculated from the propagation model and a predicted path loss matrix can be constructed.

The technical scheme of the embodiment of the invention improves the steps. The technical solution of the embodiments of the present invention will be described in detail by the following specific examples.

(example one)

The first embodiment of the invention provides a method for predicting propagation path loss based on classification fitting. The method is a prediction calculation method of propagation characteristics between a transmitter and a receiver, and the method utilizes propagation characteristics to perform clustering, obtains a propagation model for each type by adopting a non-parametric fitting method, constructs a random forest classifier, adaptively matches each propagation model to each transmitter and receiver pair by utilizing the random classifier, and obtains a propagation path loss result of refined prediction by utilizing the matched propagation model to perform prediction.

Referring to fig. 3, a specific process of the method according to an embodiment of the present invention may include:

300. and cleaning and filtering the obtained measured data.

First, it should be noted that this step is an optional step. In the subsequent steps, all the obtained measured data can be adopted for calculation, and the measured data cleaned and filtered in the step can also be adopted for calculation.

The actual measurement data refers to various types of data actually measured in the actual operation process of the communication system or data which is obtained through other ways and is helpful for predicting propagation path loss, for example, the data includes the position and height of a transmitter, the transmission intensity of electromagnetic waves, the position and height of a receiver, the reception intensity of electromagnetic waves, various engineering parameter data, the distance between stations (distance between stations), electronic map data, and the like, which are not listed one by one herein.

The cleaning and filtering means that invalid, abnormal or large-error data which do not accord with the rule are filtered by using a certain rule, and only the data which accord with the rule are reserved.

Two cleaning and filtering modes are listed, one is a conditional filtering method, and the other is an arithmetic filtering method, and the embodiment of the invention can adopt one or two of the two modes to clean and filter the acquired measured data. However, it should be noted that the embodiment of the present invention is not limited to a specific cleaning and filtering manner, and in some embodiments, the cleaning and filtering manner may also be performed in a manner not listed herein. These two cleaning and filtering methods will be described in detail later.

310. And calculating the propagation characteristics of the electromagnetic wave between the transmitter and the receiver according to the measured data.

The energy attenuation degree of the wireless electromagnetic wave in the space depends on the propagation environment from the transmitter to the receiver, and the embodiment of the invention extracts the characteristics which influence the energy attenuation of the electromagnetic wave in the propagation environment and are used for describing the propagation environment from the transmitter to the receiver, wherein the extracted characteristics are called propagation characteristics. The propagation characteristics may include one or more of the following characteristics: average inter-station distance, ratio of land features in coverage area of a cell, effective transmitter height, shielding influence of land features on a path from a transmitter to a receiver, receiving distance, average building height, density of buildings around the receiver, relative azimuth angle of a horizontal plane, relative declination angle of a vertical plane, and diffraction loss. The specific calculation method of each propagation characteristic is described later.

In this step, propagation characteristics of electromagnetic waves between any pair of transmitters and receivers within a range indicated by the measured data are calculated from the measured data. The any pair of transmitter and receiver may be, for example, a first pair of transmitter and receiver, and specifically includes a first transmitter and a first receiver, where the first transmitter is any transmitter in the range, and the first receiver is any receiver in the range.

320. And selecting at least one propagation characteristic from the propagation characteristics as a classification fitting condition according to the correlation between the propagation characteristics and the propagation path loss.

First, it should be noted that this step is an optional step. In the subsequent step, all the propagation characteristics obtained by calculation can be used as a classification fitting condition, at least one propagation characteristic selected in the step can be used as a classification fitting condition, and particularly, a plurality of propagation characteristics can be randomly selected from all the propagation characteristics to be used as the classification fitting condition. The correlation between the propagation characteristic and the propagation path loss is a degree of influence of the propagation characteristic on the transmission of the electromagnetic wave, and a calculation method thereof will be described later.

330. And clustering the measured data by using at least one propagation characteristic, and constructing a random forest classifier.

The at least one propagation feature in this step refers to at least one of the plurality of propagation features calculated in step 110, and may be, for example, at least one propagation feature selected as a classification fitting condition in step 120; alternatively, all of the plurality of propagation characteristics calculated in step 110 may be used.

In this step, the accuracy of generating the propagation model in the subsequent steps can be improved by clustering the propagation characteristics. In the step, a random forest classifier is constructed by using the propagation characteristics, so that the propagation model of the corresponding category can be accurately found through the propagation characteristics in the subsequent prediction.

340. And carrying out non-parametric fitting on each type of measured data obtained by clustering to generate a corresponding propagation model.

Different from the linear fitting method of the traditional least square method generally adopted in the prior art, the embodiment of the invention adopts a non-parametric fitting method, and each type obtained by clustering is respectively fitted so as to generate the corresponding propagation model. The gain of the non-parametric fitting method is higher, and the effect is better.

Optionally, this step may further include: and correcting the propagation model which does not meet the theoretical trend according to the propagation characteristics. The part of the propagation model obtained by fitting may not meet the theoretical trend, and the part of the propagation model can be corrected according to the propagation characteristics in the step so as to meet the requirements.

350. And according to the propagation characteristics from the transmitter to the receiver, calculating the propagation path loss between the transmitter and the receiver by utilizing the random forest classifier to adaptively match the corresponding propagation model.

For each pair of transmitter and receiver, in this step, the corresponding propagation model can be adaptively matched by using the random classifier constructed in the previous step according to the propagation characteristics between the pair of transmitter and receiver, and then the propagation path loss between the pair of transmitter and receiver can be calculated by using the matched propagation model, so as to obtain a more accurate and more effective prediction result.

For example, in this step, the corresponding propagation model may be adaptively matched to the second pair of transmitter and receiver by using a random forest classifier according to the propagation characteristics between the second pair of transmitter and receiver, where the second transmitter may be any transmitter in the range, and the second receiver may be any receiver in the range.

As can be seen from the above, in some possible embodiments of the present invention, a method for predicting propagation path loss based on classification fitting is disclosed, which calculates propagation characteristics according to measured data; clustering by using the propagation characteristics, and constructing a random forest classifier; performing non-parametric fitting on each type obtained by clustering to generate a corresponding propagation model; according to the propagation characteristics from the transmitter to the receiver, the technical characteristics of the propagation path loss between the transmitter and the receiver are calculated by utilizing the random forest classifier to adaptively match the corresponding propagation model, and the following technical effects are achieved:

From the above, the method of the embodiment of the invention comprises the following steps: cleaning and filtering (optional) measured data, calculating electromagnetic wave propagation characteristics, selecting propagation characteristics (optional), clustering and constructing a random forest classifier, generating or correcting a propagation model in a non-parametric fitting manner, and calculating a self-adaptive matching propagation model and a propagation path loss. In order to better understand the technical solutions provided by the embodiments of the present invention, the following respectively describes each of the above steps in more detail. And finally, the experimental effect of the technical scheme of the embodiment of the invention is also provided.

Firstly, cleaning and filtering the measured data (optional step)

In the data acquisition process of an actual network, some invalid values and missing values appear in measured data due to factors such as the precision of test equipment, a Global Positioning System (GPS) information blind spot, route deviation, working parameters which are not updated in time and the like, so that the precision of a propagation model which is fit by subsequent classification can be greatly influenced, and the embodiment of the invention provides two modes of data filtering methods: and (4) performing conditional filtering and arithmetic adjustment filtering, and cleaning and filtering the actually measured data.

Conditional data filtering

The measured data is filtered through experience in engineering application, the data filtering conditions are solidified, automatic processing of a software tool can be realized during specific application, and the conditions are not in sequence. For example, the data filtering conditions are as follows:

(1) invalid work parameter data filtering

And matching the imported DT (Drive Test)/CW (Continuous Wave Test)/MR (Measurement Report) data according to the working parameters, and performing import failure processing on the measured data which cannot be matched with the working parameters.

(2) Range of level values

And deleting the data of the imported measured data outside the range of the receiving threshold (the receiving threshold value defined by the 2G/3G/4G communication protocol is recommended to be used).

(3) Cell measurement point filtering

In the introduced measured data, the measured data of the measured cell is deleted when the measured data is smaller than a certain range, and the recommended value of one embodiment of the invention is 200, and the deleted data is deleted when the recommended value is smaller than 200.

(4) Distance filtering

Firstly, the average station spacing (hereinafter referred to as Inter-Site Distance, abbreviated as ISD) of the planned network is calculated, and the ISD calculation method is shown as the first characteristic.

Step 1: calculating minimum Distance (Min Distance)

If the antenna is a directional antenna, the Min Distance calculation method comprises the following steps:

min Dis tan ═ γ × ISD, a suggested value of γ is 0.07;

if the antenna is an omnidirectional antenna, the Min Distance calculation method comprises the following steps:

min Dis tan ═ γ × ISD, a recommended value of γ is 0.06.

Step 2: calculating Max Distance

If the antenna is a directional antenna, the Max Distance calculation method comprises the following steps:

max distance ═ γ × ISD, a suggested value of γ is 1.33;

max distance ═ γ × ISD, a suggested value of γ is 1.16.

Step 3: distance filtering

Deleting the measured data which do not satisfy the Distance range of [ Min Distance, Max Distance ].

(5) Combining azimuth and antenna lobe width filtering

And automatically deleting the actual measurement data of the drive test points which fall outside the range of the horizontal lobe of the antenna of the transmitter.

(6) Filtering of ground objects

Combining with an electronic map, automatically deleting data points with the ground feature proportion lower than a certain proportion, and suggesting values: 4 percent.

(II) arithmetic data filtering

(1) Data cleaning method based on feature fitting error

And carrying out non-parametric fitting on the characteristics of the actually measured data, and filtering the data with fitting errors larger than a certain threshold.

(2) Abnormal value data cleaning method based on robust Mahalanobis distance

By observing the distance from each measured data point to the center of the sample, if the distance of some measured data points is too large and is larger than a certain discrete degree, the measured data points are judged to be abnormal values, and the abnormal values are deleted.

The data which do not conform to the electromagnetic wave propagation rule can be automatically identified by the algorithm filtering, and correspondingly, discrete points are obtained, and the effect of the propagation data filtered by the algorithm is shown in figure 4:

the light-colored dots at the edges in fig. 4 represent abnormal data identified by the algorithm, and the black dots in the middle represent data after cleaning filtering. The algorithm filtering can effectively ensure the effectiveness in actually measured propagation data and avoid the interference of abnormal data on the accuracy of the model.

Secondly, calculating the propagation characteristics of the electromagnetic waves

The energy attenuation degree of the wireless electromagnetic wave in the space depends on the propagation environment from the transmitter to the receiver, and the energy attenuation law of the electromagnetic wave is similar when the electromagnetic wave propagates in the similar environment. Therefore, the invention extracts the characteristics influencing the attenuation of the energy of the electromagnetic wave in the propagation environment and is used for describing and characterizing the propagation environment from the transmitter to the receiver.

Propagation characteristics one: average inter-station spacing

When a network is established, the type of the scene to which the target planning area belongs may be determined, for example: the number of sites for constructing the network is estimated according to scenes such as Dense Urban (Dense Urban), Urban (Urban), Suburban (Suburban), Rural (road) and coverage area probability, so that the average Site distances (international-Site Distance, abbreviated as ISD) of different target planning scene areas are inconsistent. In turn, the scene type of the target planning region can be reflected by the average inter-site distance of the sites, and different propagation models should be used for different site densities. The invention takes the index as the clustering characteristic for model self-adaptive selection.

The average inter-station distance is a Cell-level characteristic index, and is classified X and defined as Cell_ISDConsider an antenna zoom-out scenario. And calculating the average distance between the sites (namely base stations) of the topological adjacent cells as the average inter-site distance of the cells by determining the nearest topological adjacent cells.

As shown in fig. 5, the average inter-site distance is calculated as follows:

501. acquiring the longitude and latitude of a current cell site and SiteListAll (a complete site list) in the radius of a cell propagation model;

502. selecting the nearest N sites in the SiteListAll as computing sites (SiteListFilter);

503. calculating the average station spacing between SitelsFilters to obtain the Cell_ISD。

Calculating more than N sites within a calculation range (configuration parameters, a suggested value: 4000m) according to N, and calculating less than N sites according to the actual number of sites; if no calculating station exists in the calculating range, the cell ISD value default value is the cell calculating radius; wherein, N is a positive integer and is a preset value.

Cell_ISDThe calculation formula is as follows:

wherein,

n represents the number of cells in the SiteListFilter, including the calculation of the cells;

distance (i) represents the horizontal distance between any cell in the SiteListFilter.

(II) propagation characteristics II: proportion of land features in cell coverage

And various ground object types are matched and fixed according to the ground objects, such as: land (bridge), plant (cultivation), building (building) and water (water), wherein different land feature types have different influences on wireless electromagnetic wave propagation, corresponding propagation environments of cells with similar land features in a site coverage range are similar, and the propagation characteristics can be used as clustering index characteristics for self-adaptive propagation model selection.

(1) Surface feature type determination

The surface feature type information can be obtained from Geographic information system Data (GIS Data for short), the surface feature names of different maps are defined to be inconsistent, surface feature type matching needs to be performed manually, the invention defines 4 surface feature types, namely land (including bridge), vegetation, building and water, for matching, and the matching result in one embodiment is shown in fig. 6.

(2) Community terrain statistics calculation

According to the antenna height, the downtilt angle, and the vertical/horizontal lobe width in the cell parameter information, the coverage of the near and far points covered by the cell can be determined, as shown in fig. 7 a.

Fig. 7a is a schematic diagram of the antenna coverage on a vertical section, and for ease of understanding and modeling calculations, a schematic diagram of a horizontal plane corresponding to fig. 7a is shown in fig. 7 b.

O (x) in FIGS. 7a and 7b_o,y_o) Denotes the site antenna position, OA denotes the coverage near point distance, OB denotes the coverage far point distance, and θ denotes the horizontal lobe width. The shaded portion represents the coverage area, and the embodiment of the invention also calculates the ground feature proportion of the coverage area represented by the shaded portion.

The calculation steps are as follows:

step 1: determining a near point distance OA and a far point distance OB;

as is known, if the antenna height h (m) is obtained from the working parameter information, including the antenna hanging height, the building height (note: including no ground object altitude), the electrical downtilt + mechanical downtilt T (°), and the vertical lobe width V (°), the near point distance OA is calculated as:

wherein if OA is not less than 2 ISD, OA is 2 ISD;

the distance to distant point OB calculation method is as follows:

if it is notThen: OB ═ 2 × ISD;

if it is notThen:

the ISD represents the average station spacing around the calculated cell, and the calculation method is shown as a first propagation characteristic.

Step 2: judgment of C (x)_c,y_c) Whether or not it falls within the shadow range

Known as C (x)_c,y_c) Point and O (x)_o,y_o) The horizontal lobe width θ (°) and the antenna direction angle F (°) are determined as follows:

if OA is OB, the proportion of each ground object is set to 0 without calculating the cell;

otherwise, the OC distance satisfies: OA is not more than OC, OB and OC form an included angle with the north direction [ F-theta/2, F + theta/2 ], the position of the C point is in a shadow part, and the ground object information of the point is obtained and recorded.

Referring to fig. 7c, the calculation method of the included angle between OC and north direction is as follows:

memory vectorThen vectorThe direction angle of (d) is calculated by the following method: wherein X<0

Step 3: calculating the ground object information proportion of the shadow part

Traversing all the cells, and calculating the grid number M of the shaded part_allTo determine the number N of the ground object grids_clutter(N_land: indicating the number of land feature grids, N_bridge: representing the number of bridge ground object grids, N_vegetation: representing the number of the vegeatation ground object grids, N_building: represents the number of building ground object grids, N_water: representing the number of water ground object grids), using N_clutter/N_allAnd (5) calculating the proportion of each type of ground object.

(III) propagation characteristics III: effective transmitter height

Transmitter actual Height (including antenna hanging Height)_antennaHeight of building_buildingHeight of ground object above sea level_clutter) Relative to the actual Height of the receiver (including the Height of the ground object)_clutterHeight of human body_body) Height for effective transmitter_tx-rxThe schematic diagram is shown in fig. 8.

Step 1: actual height of transmitter

Transmitter has antenna hanging Height_antennaHeight of building_buildingHeight of ground object above sea level_clutterTherefore, transmitter Height_txThe calculation formula is as follows:

Height_tx＝Height_antenna+Height_building+Height_clutter

the above formula unit: m (meters).

Step 2: actual height of receiver

Height of ground object existing in receiver_clutterHeight of human body_bodyHence the receiver actual Height_rxThe calculation formula is as follows:

Height_rx＝Height_clutter+Height_body

the above formula unit: m (meters).

Step 3: calculating effective transmitter height

Effective transmitter Height_tx-rxThe calculation formula is as follows:

Height_tx-rx＝Height_tx-Height_rx

the above formula unit: m (meters).

(IV) propagation characteristics IV: ground object shielding effect on transmitter to receiving path

Los (English full name: Line of sight, Chinese meaning: Line of sight); this means that there is no shielding on the electromagnetic wave transmission path and free space propagation is possible. Nlos (English line of sight, Chinese meaning: non line of sight) means that a shielding object enters an electromagnetic wave transmission channel, and diffraction and reflection often occur to deflect a signal, so that the time of the signal reaching a receiving antenna is slightly later than that of a direct signal. Since these deflected signals are out of phase with the direct signal, they reduce their power or cancel them out completely. Under the Nlos scene, the shielding severity of shielding projection of the building in the first Fresnel zone has change on the electromagnetic wave propagation, and the shielding proportion of the maximum shielding building in the Nlos is used as a fitting characteristic index. Los scene, the shielding ratio is 0.

(1) First Fresnel zone

According to the huygens-chenille principle, during the transmission of electromagnetic waves, every point on the wave front is a wave source of spherical waves for secondary radiation. A line is connected between the transmitting wave source and the receiver, the line is taken as an axis, the transmitting wave source and the receiver are taken as an ellipse of a focus, and the ellipse is rotated by 360 degrees according to the axis to obtain an ellipsoid which is a Fresnel area. The size of the fresnel zone is related to the frequency of the electron waves and the distance between the transmitters of the receiver.

In free space, electromagnetic energy from a transmitter to a receiver point is mainly propagated through a first Fresnel zone, and whether the transmitter and the receiver end belong to Los or NLos is judged by considering whether buildings are shielded in the first Fresnel zone. For modeling calculation, the invention takes a vertical section for calculation. If no building appears in the first Fresnel zone, Los exists between the transmitter and the receiver; otherwise, it is NLos. As shown in particular in fig. 9 a. As can be seen in fig. 9a, there is not necessarily free space propagation, even if the receiver can be seen directly from the transmitter.

(2) Radius of first Fresnel zone

As shown in FIG. 9b, Q is the source of the transmitted wave, P is the receiver position, S₁The plane is perpendicular to the line QP, C₁Is S₁Cross-sectional circle of face and Fresnel zone, F₁Is the radius of a cross-sectional circle, d₁、d₂Q point and P point to source C, respectively₁The distance between the centers of the circles, and the area where the circle is located is the first Fresnel zone. C₁Radius of circle F₂The calculation formula is as follows:

wherein, the lambda is the wavelength,f is the frequency (Hz), v denotes the speed of light 3 x 10⁸m。

(3) Los/Nlos determination

The ground feature information Data can be acquired from GIS Data, and the electronic map needs to contain grid-level Clutter (messy) information and Building information. The steps of judging the Los/Nlos from the transmitter to the receiver are as follows:

step 1: obtaining ground feature information and calculating step length

Inputting parameters: map accuracy

Taking the map precision as the calculation step length n_calcstep(m) as emissionThe method comprises the steps of sampling the ground feature information between a machine and a receiver, obtaining a map precision value by default, and outputting n_calcstep(m)。

Step 2: computing transmitter-to-receiver connections

According to transmitter position coordinates (X)_tx,Y_tx) And acquiring the position coordinates of the main antenna in a multi-antenna scene. Transmitter Height_tx(m), receiver position coordinates (X)_rx,Y_rx) Height of receiver_rx(m), then the transmitter is connected with the receiver by a horizontal distance d_h(m) can be represented as:

the transmitter-to-receiver link spatial slope may be expressed as:

k_{tx_rx}＝(Height_tx-Height_rx)/d_h

the transmitter and receiver connections may be represented as:

Y＝k_{tx_rx}*X+b

wherein b is a constant.

Step 3: calculating the coordinate offset of the connecting line

According to the horizontal distance d between the transmitter and the receiver_hStep n of calculating ground feature_calcstep(m), the total sampling frequency of the ground features is d_h/n_calcstepSending out ergodic sampling point from transmitter, sampling point Clutter_iPosition coordinates relative to transmitter X_txOffset amountCan be represented by (i, 1. ltoreq. i. ltoreq. d_h/n_calcstep)：

Relative to transmitter Y_txOffset amountCan be expressed as:

then Clutter_iThe position coordinates areAccording to Clutter_iThe position coordinates can acquire the grid point feature information (feature height:building height:) If there is no building data, it will

Step 4: calculating the first Fresnel zone radius Fⁱ

According to Clutter_iPosition coordinates and height of the obstruction, Cluter_iThe distance from the center of the first fresnel zone to the transmitter may be:

Clutter_ithe first fresnel zone center-to-receiver distance may be expressed as:

then Clutter_iRadius of the first Fresnel zone FⁱComprises the following steps:

wherein,

the lambda is the wavelength of the light beam,f is frequency (MHz), v denotes the speed of light 3 x 10⁸m。

Step 5: calculate Clutter_iRadius of first Fresnel zone and spatial slope of transmitter

Wherein, Height_txRepresenting the actual height of the transmitter.

Step 6: calculate Clutter_iHeight of shield to transmitter spatial slope

Step 7: judge Clutter_iWhether or not to block the propagation path from the transmitter to the receiver:

if it is notThen Clutter_iHaving a shielding for the propagation path, Clutter_iFirst Fresnel zone diameter 2 x F in vertical sectionⁱThe calculation method comprises the following steps:

step 8: distance on vertical section of shielded part

The calculation method comprises the following steps:

if the number of the first and second antennas is greater than the predetermined number,then

If not, then,then

Otherwise, Clutter_iNo shielding of transmission path from transmitter to receiver, shielding ratio

The round is finished, and the next Clutter is traversed_i+1。

Step 9: after all ground objects are traversed, the shielding ratio ShelterRate from the receiver to the transmitter_ClutterComprises the following steps:

ShelterRate_Clutteras a receiver-side feature.

(V) propagation characteristics five: receiving distance

Horizontal distance between transmitter and receiver, let transmitter coordinate Tx (X)_tx，Y_tx) Receiver coordinate Rx (X)_rx，Y_rx) Then the receiving distance d from the receiver to the transmitter:

unit: m (meters).

(VI) propagation characteristics six: average height of building

The signal can be diffracted to the receiver end through the roof and the wall corner of the building, and the average building height in the first Fresnel zone influences the diffraction effect.

Firstly, judging Los/Nlos between a transmitter and a receiver, wherein the method is shown in the fourth characteristic.

If yes, the average height of the building is considered to have no influence on the propagation path from the transmitter to the receiver, and the default parameter is 0.

Otherwise, the influence of the average height of the building on the propagation environment from the transmitter to the receiver needs to be considered. The influence of the signal after diffraction and reflection caused by the shielding of the shielding object is related to the average height of the building, and the scheme considers the average height of the building projected by the first Fresnel zone into a two-dimensional plane.

As shown in fig. 10, the calculation flow includes:

(1) traversing a computational grid

Traversing the calculation Grid to obtain Grid according to the calculation range of the cell transmitter Tx_listObtaining a computing Grid_i。

(2) Determining transmitters Tx to Grid_iNlos/Los relationship of

The judgment method is shown in the fourth characteristic.

(3) Computing Tx to Grid_iA first Fresnel region two-dimensional projection area;

tx to Grid in reality_iThe first Fresnel zone is a space ellipsoid, in the scheme, in order to facilitate modeling and calculation, a section vertical to a vertical section is taken along the axis of the ellipsoid to obtain an ellipse, and the building covered by projecting the ellipse with the same size onto a two-dimensional surface can be used for Tx to Grid_iPropagation has an effect. The effect is shown in fig. 11 when viewed from the top to the bottom of the three-dimensional space.

As shown in fig. 11, the grid building information through which the outer rectangles of the first fresnel regions of the transmitters Tx to Rx project ellipses are calculated. The calculation steps are as follows:

step 1: the ellipse minor axis 2b, major axis 2a, and focal distance 2c are calculated.

Known as Tx (X)_tx，Y_tx) To Rx (X)_rx，Y_rx) The distance between the two points is the ellipse focal point distance 2c, and the calculation method is as follows:

the minor axis 2b distance is the maximum radius of the first fresnel zone from the transmitter to the receiver, and therefore the calculation method is as follows:

due to d₁＝d₂C, the above formula is simplified, and the semiminor axis b is:

According to the elliptical nature, b²+c²＝a²The following can be obtained:

step 2: calculating coordinates of intersection points of the long axis and the rectangle

Let Tx-side intersection coordinates be Tx '(X'_tx，Y′_tx) Rx (X 'for Rx-side intersection coordinates'_rx，Y′_rx) Representing that the distance from Tx' to Tx can be represented as (a-c), then the Tx to Rx is represented as a vector:

the Tx side extends to the rectangle intersection coordinate Tx '(X'_tx，Y′_tx) Can be expressed as:

rx side extension and rectangular intersection point coordinate Rx '(X'_rx，Y′_rx) Can be expressed as:

step 3: calculating the offset coordinates of the two sides of the long axial short axis

According to the deviation of coordinate points on two sides of the long axis to the directions on two sides of the short axis vertical to the long axis, four vertex coordinates of a rectangle can be obtained, and the deviation step length to the direction of the short axis is the map precision n_calcstep(m) the number of shifts n of the major axis coordinate to both sides_offset：

Let the major axis intersections Tx 'to Rx' be represented by vectors

Perpendicular toIs offset by a direction vector ofThen:

or

The two directions of offset are represented above, and the coordinate calculation method after offset according to the offset step length in each direction is as follows:

switch (offset direction) arch

For(i＝0；i<n_offset；i++){

The offset coordinate iTx 'of the transmitter Tx' end is calculated:

receiver Rx 'end offset coordinates iRx' are calculated:

}

For(i＝0；i<n_offset；i++){

the offset coordinate iTx 'of the transmitter Tx' end is calculated:

receiver Rx 'end offset coordinates iRx' are calculated:

}

this step outputs the coordinates after each offset (iX'_tx,iY′_tx) And (iX'_rx,iY′_rx)。

Step 4: determining the grid index of the connecting line according to the coordinates of the two points and the calculation step length

Step (iX) 'determined'_tx,iY′_tx) And (iX'_rx,iY′_rx) Calculating step length by using map precision n_calcstepThe calculation method can obtain the grid index through which the two-point connecting line passes by referring to the propagation characteristic four, and the calculation method can obtain the grid index through which the two-point connecting line passesAndcalculating the offset coordinate in the offset direction to obtain all the grid indexes to form a list iClutterList_index。

The obtained iClutestrist_indexGrid index deduplication to get ClutterList_index，ClutterList_indexThe inner grid is a grid in the two-dimensional projection area in the first Fresnel area.

(4) Calculating the average height AvgHeight of buildings in the projection area_building

Traverse ClutterList_indexObtaining the grid point ground object Clutter_iJudging Clutter_iWhether the ground object type is Building or not, and if so, acquiring the height of the BuildingHeight of ground objectTraverse the next Clutter_i+1(ii) a Otherwise, go to the next Clutter_i+1。

AvgHeight_buildingThe calculating method of (2):

(seventh) propagation characteristics seventh: receiving surrounding building density

The density of buildings around the receiver affects the electromagnetic wave propagation path loss, i.e. simulates the street effect. The proportion of buildings in a certain grid around the receiving point is obtained. The calculation method comprises the following steps:

according to the configured calculation range (CalculateAreaOfRateBuilding), each CalculateAreaOfRateBuilding layer grid in the north, west, south and east directions is used as the calculation range, and the calculation method comprises the following steps:

NumberBin_building: representing the number of grids belonging to the building within the calculation range;

NumberBin: indicating the number of grids in the calculation range.

(eight) propagation characteristics eight: relative azimuth angle of horizontal plane

In the process of propagation of electromagnetic waves, attenuation of antenna lobes inhibits the loss of propagation links of the electromagnetic waves, so that a receiving point has different inhibiting effects relative to the direction of a main lobe of an antenna, and a horizontal lobe relative azimuth angle (horizon angle) serves as a classification and fitting characteristic. Horizon angle calculation:

as shown in fig. 12, is a plot of horizontal lobe versus azimuth HorizontalAngle.

In fig. 12, OA is the main lobe direction of the cell, UE is the receiving point, θ is horizon angle, and the calculation method is as follows:

step 1, acquiring an included angle theta between a horizontal lobe of a cell and a due north direction_cell

Is obtained from the ginseng.

Step 2, solving the connection between the receiving point UE and the cell position O and the true northIncluded angle of direction theta_cell：

Let O point coordinate O (x)_o,y_o) UE Point coordinates UE (x)_ue,y_ue)，θ_cellCalculating the formula:

step 3. angle theta of horizon Angle is calculated as:

θ＝|θ_ue-θ_cell|

if θ ≧ 180 °, then:

θ＝|360-θ|

(nine) propagation characteristics nine: vertical plane relatively declination angle

The vertical angle formed by the transmitter and the receiving point also affects the antenna attenuation. The vertical lobe relative downward inclination angle (VerticalAngle) is used as a classification and fitting characteristic. VerticalAngle calculation:

as shown in fig. 13, is a vertical lobe versus azimuth vertical angle diagram.

In fig. 13, OA is a vertical main lobe direction of a cell, UE is a receiving point, θ is a relative downtilt angle (vertical angle) of a vertical lobe, and θ' is an angle between the vertical main lobe direction of an antenna vertical plane and a horizontal plane; h_txThe transmitter height comprises antenna hanging height, building height and altitude; h is_ueThe receiver height comprises a human body height (default 1.5m) and an altitude; the method for calculating the relative downward inclination angle (VerticalAngle) theta of the vertical lobe comprises the following steps:

step 1: calculating theta'

And (3) enabling the electrical downtilt + mechanical downtilt to be T (degree), and obtaining the vertical lobe width V (degree), T (degree) and V (degree) from the working parameters, wherein the calculation method of theta' is as follows:

θ′＝T+v/2

step 2: calculating the horizontal distance d

Let O point coordinate O (x)_o,y_o) UE Point coordinates UE (x)_ue,y_ue) Then the horizontal distance d from the transmitter to the receiving point is:

step 3: the vertical lobe relative downtilt angle (VerticalAngle) θ is calculated as:

if h is_ue≥H_tx，

If not, then,

(ten) propagation characteristics ten: diffraction loss

The diffraction enables the radio waves to pass through the obstacle, forming a field strength behind the obstacle, i.e. a diffraction field strength. Huygen's law holds that this is because each point in front of an obstacle can be used as a new wave source to generate spherical secondary waves, and the field formed by the secondary waves behind the obstacle is a diffraction wave field.

A single-peak diffraction refers to a situation where there is only one edge-shaped obstacle, such as a mountain, in the first fresnel region of the transceiver antenna.

Multimodal diffraction refers to the presence of two or more edge-shaped obstacles in the first fresnel zone between the transmitting and receiving antennas, and the calculation of multimodal diffraction losses is much more complicated than that of unimodal diffraction losses, which are commonly used in the four algorithms of butlington, Epstein and Peterson, dygout, Atlas.

The invention calculates the diffraction loss from the transmitter to the receiver as one of the fitting characteristics.

Thirdly, selecting the propagation characteristics (optional)

Because different propagation environments are possibly related to only a plurality of specific propagation characteristics, the propagation characteristics need to be screened according to actual propagation conditions, and the propagation characteristics with higher correlation are selected as classification and fitting characteristic factors, so that the invention provides a characteristic selection method which comprises the following steps:

step 1: calculating correlation between propagation characteristics and propagation path loss

Step B provides an electromagnetic wave characteristic calculation method, calculates the correlation of all propagation path losses from each measured data transmitter to the receiver, and makes the propagation characteristic A sequence be

Feature_a＝{Feature_a(1),Feature_a(2),Feature_a(3),.....,Feature_a(n)}，

Let the propagation path loss Pathloss be

pathloss_b＝{pathloss_b(1),pathloss_b(2),pathloss_b(3),.....,pathloss_b(n)}，

The correlation coefficient γ between propagation characteristics is calculated as follows:

the characteristics with low correlation coefficient are filtered according to the correlation coefficient threshold, the strength of the correlation coefficient is judged according to the following table 1, and the characteristics with more than weak correlation are recommended to be reserved as classification fitting characteristic indexes.

TABLE 1

Correlation coefficient

Strength of correlation

0.8-1.0	Very strong correlation
		0.6-0.8	Strong correlation
0.4-0.6	Moderate degree of correlation
		0.2-0.4	Weak correlation
0.0-0.2	Very weak or no correlation

Step 2: computing pairwise propagated feature direct correlation

Using one of the two propagation characteristics if there is a strong correlation between themThe propagation characteristics are subjected to cluster fitting, so that the efficiency can be improved, and meanwhile, the redundancy is removed to improve the cluster fitting effect. Let propagation Feature A series Feature_a＝{Feature_a(1),Feature_a(2),Feature_a(3),.....,Feature_a(n) }, spread Feature B series Feature_b＝{Feature_b(1),Feature_b(2),Feature_b(3),.....,Feature_b(n), the correlation coefficient γ between propagation characteristics A, B is calculated as follows:

the invention recommends to reserve one of two propagation characteristics A, B with strong correlation as a classification fitting characteristic and delete the propagation characteristics with lower correlation between the two propagation characteristics and the propagation path loss.

Step 3: constructing feature matrices

Associating the transmitter-side characteristics with the receiver-side characteristics to form a characteristic matrix, the structure of which is as follows:

TABLE 2

Three, clustering and random forest classifier construction

Feature clustering

After the feature matrix is obtained, the propagation features need to be clustered, and the clustering modes provided by the invention include two clustering modes, one is clustering based on fitting errors, and the other is an industry classical classification algorithm K-means.

(1) Clustering based on fitting error

The clustering method based on the fitting error has the principle that the characteristic data with low propagation characteristic error and a certain threshold are clustered into one class by using a non-parametric fitting mathematical method, and the characteristic data with large deviation is clustered into another class. The method comprises the following specific steps:

step 1: traversing propagation features in a feature matrix

Let the obtained propagation characteristics A be

Feature_a＝{Feature_a(1),Feature_a(2),Feature_a(3),.....,Feature_a(n)}；

Step 2: non-parametric fitting of features to propagation path loss

And performing non-parametric fitting on the propagation characteristic A and the propagation path loss Pathloss, wherein a GAM algorithm is recommended to be adopted as a fitting algorithm. And calculating a fitting error according to the following calculation formula:

meanerror＝(m-p)/n

the invention proposes fitting error values of-0.15, 0.15.

Step 3: propagation signatures

The propagation signature matrix is labeled 1 for fit errors within the proposed range and 2 for propagation signature data outside the range. Therefore, if there are N classes corresponding to the features, there will be 2N classes combinations, i.e. the number of clusters.

(2) K-means clustering method

The K-means algorithm is an indirect clustering method based on similarity measurement among samples, and belongs to an unsupervised learning method. It is widely used in the art and the present invention recommends this approach.

(II) constructing a classifier

The classifier is constructed by utilizing the propagation characteristics, the purpose is to ensure that the propagation fitting model of the corresponding category can be accurately found through the propagation characteristics in the prediction process, the lower the error fraction of the classifier is, the higher the accuracy of finding the correct classification model is, and the classifier is constructed by adopting the industry classical random forest algorithm.

Four, non-parametric fitting generation or correction of propagation models

(I) non-parametric fitting model

The invention discloses a Generalized Additive Model (GAM) algorithm which is recommended by using a propagation characteristic fitting model as multi-dimensional non-parametric regression fitting. The GAM algorithm is invoked using the propagation characteristics and the fitting effect is shown in FIG. 14.

Through verification, the invention finds that the GAM non-parametric fitting brings 1-3 standard deviation gains compared with the linear fitting effect of the traditional least square method, and the standard deviation is fitted under the condition of same transmission characteristic data, as shown in figure 15. It should be noted that, in fig. 15, the verification data is not subjected to data filtering or the like, and is the fitting result of the most original data of the existing network.

(II) model trend correction

It should be noted that this step is an optional step. In practical application, if a certain type of propagation model is found not to conform to the theoretical trend, trend correction can be carried out on the type of propagation model.

The present invention uses the actual measurement data collected by the network as the input data, and if the collected propagation data is insufficient or the propagation feature extraction range is insufficient, the result may be obtained as shown in fig. 16.

The result shown in fig. 16 is a fitting result of a certain model in certain local point data, and a black line indicates a collected propagation path loss and a red line indicates a propagation path loss prediction line that is not collected. It can be seen that after 1000m, the prediction of the model will appear inaccurate.

The optimization method comprises the following steps:

step 1: constructing a generic predictive model

And performing linear fitting on all the propagation characteristics and the propagation path loss to obtain the universal lm model.

Step 2: constructing generic model data

In the calculation range, if the situation that no measured data exists from the transmitter to the receiver occurs, the propagation path loss data is constructed by adopting a general lm model.

Step 3: performing model fitting

Model fitting is carried out on the constructed data and actually acquired measured data, and a GAM non-parametric fitting algorithm is recommended by the method.

Fifthly, self-adaptive matching propagation model and propagation path loss calculation

By using the random forest classifier constructed in the previous step, the propagation characteristics in the data to be predicted are adaptively matched to the corresponding propagation model by using a random forest algorithm, propagation path loss prediction is performed, and a propagation path loss matrix structure can be constructed, for example, as shown in table 3 below.

TABLE 3

Transmitter and receiver

Reception point 1

Reception point 2

Reception point 3

Reception point 4

……

Receiving point n

Trx₁

pathloss

……

pathloss

Trx₂

pathloss

……

pathloss

Trx₃

pathloss

……

pathloss

……

Sixth, effect of experiment

The final technical effect of the invention is compared with the effects of a traditional empirical model (SPM) and a deterministic model (Volcano), and the comparison result of the propagation path loss prediction accuracy is as follows under the same data input condition:

(1) comparison of fitting standard deviations

The fit effect can characterize the fit between the model and the propagation path loss, and whether the model is available for the area planning is generally determined by correcting the standard deviation. The technical effect of the present invention is demonstrated by comparing the fitting Standard deviation of the deterministic Model (Volcano Model) and the empirical Model (SPM: Standard Propagation Model), as shown in FIG. 17.

(2) Percent prediction error comparison

Besides good fitting effect, the propagation model needs to have better effect compared with actual propagation path loss in practical application. The comparison method comprises the steps that a part of data is used as correction data, a part of data is used as acceptance data, and the error proportion in the acceptance data is counted to prove the technical effect of the invention. Please refer to the effect of hangzhou DT as shown in table 4 and the effect of guangzhou MR as shown in table 5.

(3) Comparison of efficiency

By verification, with the use of the hang state local point under the same planning scene condition (4000 ten thousand sets of transmitters and receivers), the classification fitting model consumes time: 1 hour and 13 minutes, the Volcano model takes 4 hours and 11 minutes, and the efficiency of the classification fitting model is 2 times higher than that of the deterministic model.

In summary, the embodiment of the present invention provides a method for classified fitting prediction of propagation path loss, which comprehensively considers propagation characteristics and efficiently and accurately predicts propagation path loss by using a non-parametric fitting method. The technical scheme of the embodiment of the invention can be used as a network planning basic module and is suitable for a 2G/3G/4G network system, and the input data source comprises CW/DT/MR data.

(example two)

In order to better implement the above-mentioned aspects of the embodiments of the present invention, the following also provides related devices for implementing the above-mentioned aspects cooperatively.

Referring to fig. 18, an embodiment of an apparatus 1800 for predicting propagation path loss based on classification fitting according to the present invention may include:

a characteristic calculating module 1801, configured to calculate, according to the measured data, propagation characteristics of electromagnetic waves between the transmitter and the receiver;

a clustering module 1802, configured to cluster the measured data by using at least one propagation characteristic, and construct a random forest classifier;

a model generating module 1803, configured to perform non-parametric fitting on each type of measured data obtained by clustering to generate a corresponding propagation model;

and the matching calculation module 1804 is used for calculating the propagation path loss between the transmitter and the receiver by using the random forest classifier to adaptively match the corresponding propagation model according to the propagation characteristics between the transmitter and the receiver.

In some embodiments of the present invention, the apparatus 1800 further comprises:

and a cleaning and filtering module 1805, configured to clean and filter the acquired measured data by using one or two of a conditional filtering method and an arithmetic filtering method.

In some embodiments of the invention, the propagation characteristics comprise one or more of the following characteristics: average inter-station distance, ratio of land features in coverage area of a cell, effective transmitter height, shielding influence of land features on a path from a transmitter to a receiver, receiving distance, average building height, density of buildings around the receiver, relative azimuth angle of a horizontal plane, relative declination angle of a vertical plane, and diffraction loss.

a feature selection module 1806, configured to select at least one propagation feature from the propagation features as a classification fitting condition according to a correlation between the propagation features and the propagation path loss;

the clustering module 1802 is specifically configured to cluster the measured data by using the at least one propagation feature selected by the feature selection module as the classification fitting condition.

In some embodiments of the present invention, the model generating module 1803 is further configured to modify a propagation model that does not satisfy a theoretical trend according to the propagation characteristics.

It can be understood that the functions of each functional module of the apparatus for predicting propagation path loss based on classification fitting according to the embodiments of the present invention may be specifically implemented according to the method in the foregoing method embodiments, and the specific implementation process may refer to the related description in the foregoing method embodiments, and will not be described herein again.

As can be seen from the above, in some possible embodiments of the present invention, a device for predicting propagation path loss based on classification fitting is disclosed, which calculates propagation characteristics according to measured data; clustering by using the propagation characteristics, and constructing a random forest classifier; performing non-parametric fitting on each type obtained by clustering to generate a corresponding propagation model; according to the propagation characteristics from the transmitter to the receiver, the technical characteristics of the propagation path loss between the transmitter and the receiver are calculated by utilizing the random forest classifier to adaptively match the corresponding propagation model, and the following technical effects are achieved:

(example III)

An embodiment of the present invention further provides a computer storage medium, where a program may be stored, and when the program is executed, the program includes some or all of the steps of the method for predicting propagation path loss based on classification fitting described in the above method embodiment.

(example four)

Referring to fig. 19, an embodiment of the present invention further provides a computer device 1900, which may include: processor 1910, memory 1920, communication interface 1930, bus 1940; the memory 1920 is configured to store computer executable instructions, the processor 1910 is connected to the memory 1920 through the bus 1940, and when the computer device 1900 is running, the processor 1910 executes the computer executable instructions stored in the memory 1920 to make the computer device 1900 execute part or all of the steps of the method for predicting propagation path loss based on class fitting according to the embodiment of the present invention.

The computer equipment clusters the measured data by using the propagation characteristics, so that the utilization rate of the measured data is improved, and the effectiveness of a large amount of measured data in propagation path loss prediction is higher; based on the clustering result, a non-parametric fitting method is adopted to perform classification fitting on the measured data, so that the calculated amount is reduced, and the efficiency is improved; the propagation model obtained by carrying out classification fitting on a large amount of actually measured data based on propagation characteristic clustering and non-parametric fitting has higher prediction precision, so that the prediction result is more accurate and more effective; the constructed random forest classifier can adaptively match a corresponding propagation model between a transmitter and a receiver according to propagation characteristics, so that the applicability of the propagation model is improved; the speed, the precision and the accuracy of the propagation path loss prediction are improved, various defects in the prior art are overcome or reduced, and the accuracy and the rationality of wireless network planning can be guaranteed.

In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to the related descriptions of other embodiments.

It should be noted that, for simplicity of description, the above-mentioned method embodiments are described as a series of acts or combination of acts, but those skilled in the art will recognize that the present invention is not limited by the order of acts, as some steps may occur in other orders or concurrently in accordance with the invention. Further, those skilled in the art should also appreciate that the embodiments described in the specification are preferred embodiments and that the acts and modules referred to are not necessarily required by the invention.

It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.

In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, a division of a unit is merely a logical division, and an actual implementation may have another division, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.

Units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.

Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by associated hardware instructed by a program, which may be stored in a computer-readable storage medium, and the storage medium may include: ROM, RAM, magnetic or optical disks, and the like.

The method and the device for predicting propagation path loss based on classification fitting provided by the embodiment of the invention are described in detail, a specific example is applied in the text to explain the principle and the implementation of the invention, and the description of the embodiment is only used for helping to understand the method and the core idea of the invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.

Claims

1. A method for predicting propagation path loss based on classification fitting is characterized by comprising the following steps:

calculating the propagation characteristics of electromagnetic waves between a transmitter and a receiver according to the measured data;

clustering the measured data by using at least one propagation characteristic, and constructing a random forest classifier;

carrying out non-parametric fitting on each type of measured data obtained by clustering to generate a corresponding propagation model;

and according to the propagation characteristics from the transmitter to the receiver, calculating the propagation path loss between the transmitter and the receiver by utilizing the random forest classifier to adaptively match the corresponding propagation model.

2. The method of claim 1, wherein prior to calculating propagation characteristics of the electromagnetic wave from the transmitter to the receiver from the measured data, further comprising:

and cleaning and filtering the acquired measured data by adopting one or two of a conditional filtering method and an arithmetic filtering method.

3. The method of claim 1,

the propagation characteristics include one or more of the following characteristics: average inter-station distance, ratio of land features in coverage area of a cell, effective transmitter height, shielding influence of land features on a path from a transmitter to a receiver, receiving distance, average building height, density of buildings around the receiver, relative azimuth angle of a horizontal plane, relative declination angle of a vertical plane, and diffraction loss.

4. The method of claim 1, wherein clustering the measured data using the at least one propagation feature further comprises:

selecting at least one propagation characteristic from the propagation characteristics as a classification fitting condition according to the correlation between the propagation characteristics and the propagation path loss;

the clustering measured data using at least one propagation feature comprises:

clustering the measured data using the at least one propagation feature selected as a classification fit condition.

5. The method as claimed in any one of claims 1-4, wherein before adaptively matching the corresponding propagation model using the random forest classifier according to the propagation characteristics from the transmitter to the receiver, the method further comprises: and correcting the propagation model which does not meet the theoretical trend according to the propagation characteristics.

6. An apparatus for predicting propagation path loss based on classification fitting, comprising:

the characteristic calculation module is used for calculating the propagation characteristics of the electromagnetic waves between the transmitter and the receiver according to the measured data;

the cluster processing module is used for clustering the measured data by utilizing at least one propagation characteristic and constructing a random forest classifier;

the model generation module is used for carrying out non-parametric fitting on each type of measured data obtained by clustering to generate a corresponding propagation model;

and the matching calculation module is used for calculating the propagation path loss between the transmitter and the receiver by utilizing the random forest classifier to adaptively match the corresponding propagation model according to the propagation characteristics between the transmitter and the receiver.

7. The apparatus of claim 6, further comprising:

and the cleaning and filtering module is used for cleaning and filtering the acquired measured data by adopting one or two of a conditional filtering method and an arithmetic filtering method.

8. The apparatus of claim 6,

9. The apparatus of claim 6, further comprising:

the characteristic selection module is used for selecting at least one propagation characteristic from the propagation characteristics as a classification fitting condition according to the correlation between the propagation characteristics and the propagation path loss;

the clustering module is specifically configured to cluster the measured data by using the at least one propagation feature selected as the classification fitting condition by the feature selection module.

10. The apparatus according to any one of claims 6 to 9,

and the model generation module is also used for correcting the propagation model which does not meet the theoretical trend according to the propagation characteristics.

11. A computer device, comprising a processor, a memory, a bus, and a communication interface; the memory is configured to store computer-executable instructions, the processor is coupled to the memory via the bus, and when the computer device is running, the processor executes the computer-executable instructions stored in the memory to cause the computer device to perform the method of predicting propagation path loss based on class fitting according to any one of claims 1 to 5.