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

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

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

Technical field

The present invention relates to fields of communication technology, and in particular to a kind of that the method and dress that path loss is propagated in prediction are fitted based on classification It sets.

Background technique

Coverage evaluating is carried out to wireless network by Planning Tool, is wireless network planning and the optimization for carrying out high quality One of important means.It is respectively quasi- for the emulation of the communication technology (such as 2G/3G/4G) base values (such as level value, interference, rate) True property is the key that influence coverage evaluating accuracy.The part of the emulation core the most of each base values is exactly transmitter to connecing Radio magnetic wave propagates the calculating of path loss between receipts machine.Industry relies primarily on propagation model (Propagation Model) to biography It broadcasts path loss to be estimated, propagation model is the mathematical model obtained by mass data or calculation method, the generation of model Journey is fitting.

Propagation model is divided into two classes: deterministic type propagation model, experience (statistics) propagation model according to implementation.It determines Type propagation model is using ray tracing technique as representative, and experience (statistics) propagation model is with the specific experience such as HATA, Okumura, LEE Based on formula.Due to plan scene diversity, complexity, no matter deterministic type propagation model or experience (statistics) propagating mode Type is all to need to carry out propagation model revision according to actual scene data to be used to predict network coverage situation.

Practice discovery, for deterministic type propagation model, the ray tracing technique of use has computationally intensive, and low efficiency lacks It falls into；Also, the propagation model that Tracing Technology obtains is usually a planning region using a propagation model, can not be certainly Respective propagation model is adaptively adapted to according to propagation characteristic, causes propagation model application not high.Experience (statistics) is propagated The defects such as model has accuracy of simulation low, and propagation model revision is low to data validity utilization rate.

Summary of the invention

The embodiment of the present invention provides a kind of method and apparatus for being fitted prediction propagation path loss based on classification, is passed with improving creation The efficiency of model is broadcast, the accuracy that path loss is propagated in prediction is improved, helps to be propagated accordingly according to propagation characteristic Adaptive matching Model.

First aspect present invention provides a kind of method for being fitted prediction propagation path loss based on classification, comprising:

Electromagnetic wave propagation feature between transmitted from transmitter to receiver is calculated according to measured data；It is propagated using at least one special Sign clusters measured data, and constructs random forest grader；Non- ginseng is carried out to every a kind of measured data that cluster obtains Fitting generates corresponding propagation model；According to the propagation characteristic between transmitted from transmitter to receiver, classified using the random forest The corresponding propagation model of device Adaptive matching calculates the propagation path loss between transmitter and receiver.

In the first possible implementation, described that electromagnetic wave between transmitted from transmitter to receiver is calculated according to measured data Propagation characteristic before, further includes: use one or both of conditional filter method and algorithms filter method, to acquisition Measured data carry out cleaning filtering.

With reference to first aspect or the first possible implementation of first aspect, in second of possible implementation In, the propagation characteristic includes one of following characteristics or a variety of: average station spacing, cell coverage area atural object ratio have Transmitter height is imitated, atural object blocks influence on transmitted from transmitter to receiver path, receives distance, building average height, receiver Adjacent Buildings density, horizontal plane relative bearing, vertical plane is with respect to tilt angled down, diffraction loss.

With reference to first aspect or the first or second of possible implementation of first aspect, possible at the third In implementation, described to carry out cluster to measured data using at least one propagation characteristic include: according to propagation characteristic and to propagate Correlation between path loss selects at least one propagation characteristic as classification fitting condition from the propagation characteristic；Utilize choosing At least one propagation characteristic selected clusters measured data.

With reference to first aspect or first aspect the first any one of to the third possible implementation, In four kinds of possible implementations, the propagation characteristic according between transmitted from transmitter to receiver utilizes the random forest point Before the corresponding propagation model of class device Adaptive matching, further includes: according to the propagation characteristic to the biography for being unsatisfactory for theoretical trend Model is broadcast to be modified.

Second aspect of the present invention provides a kind of device that prediction propagation path loss is fitted based on classification, comprising: feature calculation mould Block, for calculating electromagnetic wave propagation feature between transmitted from transmitter to receiver according to measured data；Clustering processing module, for benefit Measured data is clustered at least one propagation characteristic, and constructs random forest grader；Model generation module, for pair It clusters obtained every a kind of measured data and carries out the corresponding propagation model of non-ginseng fitting generation；Matching primitives module is used for basis Propagation characteristic between transmitted from transmitter to receiver, using the corresponding propagation model of random forest grader Adaptive matching, Calculate the propagation path loss between transmitter and receiver.

In the first possible implementation, described device further include: cleaning filtering module, for using conditional mistake One or both of filtering method and algorithms filter method carry out cleaning filtering to the measured data of acquisition.

In conjunction with the possible implementation of the first of second aspect or second aspect, in second of possible implementation In, the propagation characteristic includes one of following characteristics or a variety of: average station spacing, cell coverage area atural object ratio have Transmitter height is imitated, atural object blocks influence on transmitted from transmitter to receiver path, receives distance, building average height, receiver Adjacent Buildings density, horizontal plane relative bearing, vertical plane is with respect to tilt angled down, diffraction loss.

It is possible at the third in conjunction with the first or second of possible implementation of second aspect or second aspect In implementation, described device further include: feature selection module, for according to related between propagation characteristic and propagation path loss Property, select at least one propagation characteristic as classification fitting condition from the propagation characteristic；The clustering processing module, specifically At least one propagation characteristic for being selected using the feature selection module clusters measured data.

In conjunction with any one of the first of second aspect or second aspect to the third possible implementation, In four kinds of possible implementations, the model generation module is also used to according to the propagation characteristic to being unsatisfactory for theoretical trend Propagation model be modified.

Third aspect present invention provides a kind of computer equipment, and the computer equipment includes processor, memory, bus And communication interface；The memory passes through described total for storing computer executed instructions, the processor and the memory Line connection, when computer equipment operation, the computer execution that the processor executes the memory storage refers to It enables, so that the computer equipment executes the side for propagating path loss based on classification fitting prediction as described in the first aspect of the invention Method.

Therefore in some feasible embodiments of the invention, propagation characteristic is calculated using according to measured data； It is clustered using propagation characteristic, and constructs random forest grader；Every a kind of non-ginseng fitting of progress that cluster obtains is generated Corresponding propagation model；It is adaptive using the random forest grader according to the propagation characteristic between transmitted from transmitter to receiver Corresponding propagation model is matched, the technical characteristic of the propagation path loss between transmitter and receiver is calculated, achieves following technology Effect:

1, measured data is clustered using propagation characteristic, improves the utilization rate to measured data, so that a large amount of real Validity of the measured data in propagation path loss prediction is higher；

2, it is based on cluster result, classification fitting is carried out to measured data using non-ginseng approximating method, reduces calculation amount, mentions High efficiency；

3, the propagation model that classification fitting obtains is carried out to a large amount of measured datas based on propagation characteristic cluster and non-ginseng fitting, With higher precision of prediction, prediction result can be made more acurrate more effective；

4, the random forest grader constructed can be according to propagation characteristic Adaptive matching between the transmitter and receiver Corresponding propagation model improves the application of propagation model.

5, present invention method improves speed, precision and the accuracy for propagating path loss prediction, solves or reduces The number of drawbacks of the prior art can ensure the accuracy and reasonability of wireless network planning.

Detailed description of the invention

Technical solution in order to illustrate the embodiments of the present invention more clearly, below will be to institute in embodiment and description of the prior art Attached drawing to be used is needed to be briefly described, it should be apparent that, the accompanying drawings in the following description is only some implementations of the invention Example, for those of ordinary skill in the art, without creative efforts, can also obtain according to these attached drawings Obtain other attached drawings.

Fig. 1 is the schematic diagram of communication system in the embodiment of the present invention；

Fig. 2 is a kind of flow chart of Electromagnetic Wave Propagation path loss prediction technique；

Fig. 3 is the flow chart that the embodiment of the present invention is fitted that the method for path loss is propagated in prediction based on classification；

Fig. 4 is the effect picture of the propagation data by algorithms filtering；

Fig. 5 is the calculation flow chart of average station spacing；

Fig. 6 is the matching schematic diagram of type of ground objects；

Fig. 7 a is the schematic diagram on antenna covering vertical section；

Fig. 7 b is the schematic diagram of horizontal plane corresponding with Fig. 7 a；

Fig. 7 c is the schematic diagram of OC Yu direct north angle；

Fig. 8 is the schematic diagram of effective transmitter height；

Fig. 9 a is NLos schematic diagram of a scenario；

Fig. 9 b is the first luxuriant and rich with fragrance Neil area schematic diagram；

Figure 10 is the calculation flow chart of building average height；

Figure 11 is the first two-dimensional projection, luxuriant and rich with fragrance Neil area effect picture；

Figure 12 is horizontal lobe relative bearing schematic diagram；

Figure 13 is vertical lobe relative bearing schematic diagram；

Figure 14 is non-ginseng fitting effect schematic diagram；

Figure 15 is the fitting effect comparison diagram of non-ginseng fitting GAM algorithm and least square method algorithm；

Figure 16 is the fitting effect tendency chart of certain some model of office point；

Figure 17 is classification model of fit and deterministic type, empirical models fitting contrast effect figure；

Figure 18 is the structural schematic diagram that the embodiment of the present invention is fitted that the device of path loss is propagated in prediction based on classification；

Figure 19 is a kind of structural schematic diagram of computer equipment of the embodiment of the present invention.

Specific embodiment

In order to enable those skilled in the art to better understand the solution of the present invention, below in conjunction in the embodiment of the present invention Attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is only The embodiment of a part of the invention, instead of all the embodiments.Based on the embodiments of the present invention, ordinary skill people The model that the present invention protects all should belong in member's every other embodiment obtained without making creative work It encloses.

Technical solution of the embodiment of the present invention, applied to the communication system including transmitter and receiver, which can Belong to any one of 2G, the communication systems such as 3G, 4G.Wherein, 2G is second generation mobile communication technical specification (full name in English: 2- Generation wireless telephone technology) English abbreviation, 3G be 3rd generation mobile communication technology rule The English abbreviation of lattice (full name in English: 3rd-Generation), 4G are fourth generation mobile communication technology (full name in English: the 4th Generation mobile communication technology) English abbreviation.

As shown in Figure 1, communication system 100 described in the embodiment of the present invention for example may include transmitter 110 and receive Machine 120, wherein transmitter 110 for example can be various base stations (eNode B), including ordinary base station, micro-base station, indoor base station, Home eNodeB (Home eNode B) etc.；Receiver 120 for example can be mobile station (mobile station), mobile phone etc., or Person is also possible to base station.Wherein, transmitter 110 is referred to as website, and receiver 120 is referred to as receiving point.

Technical solution of the embodiment of the present invention for realizing Electromagnetic Wave Propagation path loss between transmitted from transmitter to receiver prediction.Electricity Electromagnetic wave propagation path loss the essence of prediction is that the propagation characteristic of transmitted from transmitter to receiver influences the pre- of severity to the transmission of electromagnetic wave It surveys.

As shown in Fig. 2, being a kind of flow chart of Electromagnetic Wave Propagation path loss prediction technique in routine techniques, including following mistake Journey:

Setp1: it is suitble to the propagation model of planning region according to measured data fitting/correction；

Optionally, before this step, it can also include the steps that one is filtered cleaning to measured data.

Setp2: being that transmitted from transmitter to receiver matches suitable propagation model according to communication environments feature；

Setp3: the propagation path loss between transmitted from transmitter to receiver is calculated according to propagation model, and it is possible to construct prediction Path loss matrix.

Technical solution of the embodiment of the present invention improves above-mentioned steps.Below by specific embodiment, to this hair Bright embodiment technical solution is described in detail.

(embodiment one)

The embodiment of the present invention one is to provide a kind of method for being fitted prediction propagation path loss based on classification.This method is one The prediction calculation method of propagation characteristic between kind transmitted from transmitter to receiver, this method is clustered using propagation characteristic, to each Class obtains propagation model using non-ginseng approximating method, and constructs random forest grader, is passed each using probabilistic classifier It broadcasts model adaptation and is matched to each transmitter and receiver to upper, predicted using matched propagation model, essence can be obtained Refine the propagation path loss result of prediction.

Referring to FIG. 3, the detailed process of present invention method can include:

300, cleaning filtering is carried out to the measured data of acquisition.

Firstly, it is necessary to explanation, this step is optional step.It can be using the whole actual measurement numbers obtained in subsequent step According to being calculated, filtered measured data can also be cleaned using this step and be calculated.

Wherein, described measured data refers to that actual measurement obtains various types of in communication system real operational process Data or the data for facilitating prediction and propagating path loss obtained by other approach, position and height for example including transmitter Degree, electromagnetic radiation intensity, the position of receiver and height, electromagnetic wave receiving intensity, various work parameter evidences, distance between sites (are stood The distance between point), electronic map data, etc. will not enumerate herein.

Described cleaning filtering refers to, will be invalid using certain rule, or abnormal or error is biggish, waits not Legal data filter out, and only retain legal data.

Two kinds of cleaning filter types are enumerated herein, one is conditional filter method, one is algorithms filter method, The embodiment of the present invention can be used one or both of both modes and carry out cleaning filtering to the measured data of acquisition.But it needs Illustrate, the embodiment of the present invention is not intended to limit specific cleaning filter type, can also use herein not in some embodiments The mode enumerated carries out cleaning filtering.Both cleaning filter types will be described in detail below.

310, electromagnetic wave propagation feature between transmitted from transmitter to receiver is calculated according to measured data.

Energy attenuation degree of the radio magnetic wave in space depends on the communication environments around transmitted from transmitter to receiver, this hair Bright embodiment will come out in the feature extraction of influence electromagnetic wave energy decaying in communication environments, for describing characterization transmitter to reception The feature of the communication environments of machine, the extraction is known as propagation characteristic.The propagation characteristic may include one of following characteristics or more Plant: be averaged station spacing, cell coverage area atural object ratio, effective transmitter height, and atural object hides on transmitted from transmitter to receiver path Gear influences, and receives distance, building average height, receiver Adjacent Buildings density, horizontal plane relative bearing, vertical plane phase To tilt angled down, diffraction loss.The calculation method of each specific propagation characteristic illustrates below.

It should be noted that according to measured data, in the range of calculating measured data expression, any pair of hair in this step Penetrate electromagnetic wave propagation feature between machine and receiver.Any pair of transmitter and receiver for example can be first pair of hair Machine and receiver are penetrated, the first transmitter and the first receiver are specifically included, wherein the first transmitter is hair either one or two of in range Machine is penetrated, the first receiver is any one receiver in range.

320, according to the correlation between propagation characteristic and propagation path loss, at least one pass is selected from the propagation characteristic Feature is broadcast as classification fitting condition.

Firstly, it is necessary to explanation, this step is optional step.It can use the whole biographies being calculated in subsequent step Feature is broadcast as classification fitting condition, at least one propagation characteristic that this step selects also is can use and is fitted item as classification Part particularly can also randomly select several propagation characteristic as classification fitting condition from whole propagation characteristics.Wherein, Propagation characteristic and propagate path loss between correlation refer to, transmission influence degree of the propagation characteristic to electromagnetic wave, calculation method Illustrate below.

330, measured data is clustered using at least one propagation characteristic, and constructs random forest grader.

Described at least one propagation characteristic refers in a variety of propagation characteristics being calculated in step 110 in this step At least one, for example, it may be being selected as at least one propagation characteristic of classification fitting condition in step 120；Alternatively, can also To be the whole in a variety of propagation characteristics being calculated in step 110.

In this step, propagation characteristic cluster can be proposed into the accuracy that rear subsequent step generates propagation model.In this step, Random forest grader also is constructed using propagation characteristic, it is therefore an objective to guarantee accurately find by propagation characteristic in subsequent prediction The propagation model of respective classes.

340, non-ginseng fitting is carried out to every a kind of measured data that cluster obtains and generates corresponding propagation model.

Unlike the linear fit method of the traditional least square method generallyd use in the prior art, the present invention is implemented In example, using non-ginseng approximating method, moreover, being that the every one kind obtained for cluster is fitted respectively, to generate corresponding Propagation model.The gain of non-ginseng approximating method is higher, and effect is more preferable.

It optionally, can also include: according to the propagation characteristic in this step to the propagation model for being unsatisfactory for theoretical trend It is modified.Theoretical trend may be unsatisfactory for by being fitted obtained part propagation model, can also be according to propagation characteristic in this step To the part, propagation model is modified, and is allowed to meet the requirements.

350, according to the propagation characteristic between transmitted from transmitter to receiver, the random forest grader Adaptive matching is utilized Corresponding propagation model calculates the propagation path loss between transmitter and receiver.

It, can be according to this to the propagation between transmitter and receiver in this step for every a pair of of transmitter and receiver Feature, the corresponding propagation model of probabilistic classifier Adaptive matching constructed using previous step, then, so that it may with being matched Propagation model, to calculate this to the propagation path loss between transmitter and receiver, thus obtain more precisely, more effectively predict As a result.

It for example, can be according to the propagation characteristic between second pair of transmitter and receiver, using random in this step Forest classified device, the corresponding propagation model of Adaptive matching to second pair of transmitter and receiver, wherein the second transmitter can be with It is any one transmitter in range, the second receiver can be any one receiver in range.

Therefore in some feasible embodiments of the invention, discloses a kind of be fitted based on classification and predict to pass The method for broadcasting path loss calculates propagation characteristic using according to measured data；It is clustered using propagation characteristic, and constructs random forest Classifier；Corresponding propagation model is generated to every a kind of non-ginseng fitting of progress that cluster obtains；According to transmitted from transmitter to receiver it Between propagation characteristic calculate transmitter and reception using the corresponding propagation model of random forest grader Adaptive matching The technical characteristic of propagation path loss between machine, achieves following technical effect:

1, measured data is clustered using propagation characteristic, improves the utilization rate to measured data, so that a large amount of real Validity of the measured data in propagation path loss prediction is higher；

2, it is based on cluster result, classification fitting is carried out to measured data using non-ginseng approximating method, reduces calculation amount, mentions High efficiency；

3, the propagation model that classification fitting obtains is carried out to a large amount of measured datas based on propagation characteristic cluster and non-ginseng fitting, With higher precision of prediction, prediction result can be made more acurrate more effective；

4, the random forest grader constructed can be according to propagation characteristic Adaptive matching between the transmitter and receiver Corresponding propagation model improves the application of propagation model.

5, present invention method improves speed, precision and the accuracy for propagating path loss prediction, solves or reduces The number of drawbacks of the prior art can ensure the accuracy and reasonability of wireless network planning.

From the above mentioned, present invention method includes the following steps: the cleaning filtering (optional) of measured data, calculates electricity Electromagnetic wave propagation feature selects propagation characteristic (optional), and cluster and building random forest grader, non-ginseng fitting generates or amendment passes Model is broadcast, Adaptive matching propagation model and propagation path loss calculate.The skill that embodiment provides to facilitate the understanding of the present invention Art scheme is separately below described in detail above-mentioned each step.Also, finally also providing skill of the embodiment of the present invention The experiment effect of art scheme.

One, (optional step) is filtered in the cleaning of measured data

Real network acquisition data procedures in can because test equipment precision, global positioning system (full name in English: Global Positioning System, English abbreviation: GPS) information blind spot, path deviations, work ginseng not in time update etc. factors It will cause in measured data and some invalid values and missing values occur, so that meeting extreme influence subsequent classification fits the propagating mode come The precision of type, the embodiment of the present invention provides the data filtering method of two ways: conditional filtering, algorithms adjustment filtering, right Measured data carries out cleaning filtering.

(1) conditional data filtering

Measured data is filtered by the experience on engineer application, curing data filter condition, when concrete application Software tool may be implemented into automatically process, each condition is without sequencing.For example, data filtering condition is as follows:

(1) invalid work joins data filtering

According to work ginseng to importing DT (Drive Test, drive test)/CW (Continuous Wave, continuous wave test)/MR (Measurement Report, measurement report) data are matched, and the measured data for that can not be matched to work ginseng imports Failure handling.

(2) level value range

Measured data after importing is in threshold level (it is recommended that the threshold level value for using 2G/3G/4G communication protocol definition) Data outside range are deleted.

(3) cell measurement point filters

It imports in measured data, the measured data for being measured cell is deleted less than a certain range of, the present invention one The recommended value of embodiment is 200, being deleted less than 200.

(4) distance filtering

It first calculates planning network and is averaged station spacing (full name in English: Inter-Site Distance, English abbreviation: ISD), ISD calculation method is shown in feature one.

Step 1: it calculates minimum range (Min Distance)

If it is directional aerial, Min Distance calculation method are as follows:

The recommended value of Min Dis tan ce=γ × ISD, γ are 0.07；

If it is omnidirectional antenna, Min Distance calculation method:

The recommended value of Min Dis tan ce=γ × ISD, γ are 0.06.

Step 2: maximum distance Max Distance is calculated

If it is directional aerial, Max Distance calculation method:

The recommended value of Max Dis tan ce=γ × ISD, γ are 1.33；

If it is omnidirectional antenna, Min Distance calculation method:

The recommended value of Max Dis tan ce=γ × ISD, γ are 1.16.

Step 3: distance filtering

The measured data that will be unsatisfactory for [Min Distance, Max Distance] distance range is deleted.

(5) azimuth and antenna lobe width filter are combined

It is automatically deleted the drive test point measured data fallen outside the horizontal lobe range of transmitter antenna.

(6) atural object filters

In conjunction with electronic map, it is automatically deleted and falls on the ground object ratio lower than a certain proportion of data point, it is proposed that value: 4%.

(2) algorithms data filtering

(1) data cleaning method based on feature fitting error

The feature of measured data is subjected to non-ginseng fitting, filtering error of fitting is greater than the data of certain thresholding.

(2) based on the exceptional value data cleaning method of steady mahalanobis distance

By the distance of each measured data point of observation to center of a sample, if the distance of certain measured datas point is excessive, greatly In certain dispersion degree, then judgement is exceptional value, and deletes these exceptional values.

Algorithms filtering energy automatic identification does not meet the data of Electromagnetic Wave Propagation rule, and corresponding more discrete point, The propagation data effect filtered by algorithms is as shown in Figure 4:

The point of edge light color indicates that the abnormal data identified by algorithm, the point of middle section black indicate in Fig. 4 Clean filtered data.Algorithms filtering can effectively ensure to survey the validity in propagation data, avoid abnormal data pair The interference of model accuracy.

Two, Electromagnetic Wave Propagation feature is calculated

Energy attenuation degree of the radio magnetic wave in space depends on the communication environments around transmitted from transmitter to receiver, electromagnetism Wave is propagated in similar environment, and the energy attenuation rule shown is also similar.Therefore, the present invention will influence electricity in communication environments The feature extraction of magnetic wave energy decaying comes out, for describing the communication environments of characterization transmitted from transmitter to receiver.

(1) propagation characteristic one: average station spacing

When building network, often can by judging the affiliated scene type in goal programming region, such as: dense city The scenes such as (Dense Urban), city (Urban), city suburbs (Suburban), rural area (Rural) and overlay area are general Rate, to estimate the website quantity of building network, therefore, the website of different target planning scene areas is averaged, and (English is complete for station spacing Claim: Inter-Site Distance, English abbreviation: ISD) it is inconsistent.In turn, can by the average station spacing of website come Reflect the affiliated scene type in goal programming region, different site densities should use different propagation models.The present invention is by this index As cluster feature, chosen for model adaptation.

Average station spacing is cell-level characteristic index, is classification X, is defined as Cell_ISD, consider antenna pulling away scene.Pass through It determines topological adjacent area recently, calculates the average distance between topological adjacent area website (i.e. base station), the average station spacing as cell.

As shown in figure 5, steps are as follows for the calculating of average station spacing:

501, (all stations SiteListAll in current cell site longitude and latitude and cell propagation model radius are obtained Point list)；

502, N number of website nearest in SiteListAll is chosen as calculating website (SiteListFilter)；

503, average station spacing between SiteListFilter is calculated, Cell is obtained_ISD。

Be more than that N number of website is calculated according to N in computer capacity (configuration parameter, it is proposed that value: 4000m), insufficient N number of website according to Real site number calculates；If being that cell calculates half by cell ISD value default value without calculating website in computer capacity Diameter；Wherein, N is positive integer, is a preset value.

Cell_ISDCalculation formula is as follows:

Wherein,

N indicates cell number in SiteListFilter, comprising calculating cell；

Distance (i) indicates the horizontal distance in SiteListFilter between arbitrary cells.

(2) propagation characteristic two: cell coverage area atural object ratio

Fixed a variety of types of ground objects are matched to according to atural object, such as: land (land, (bridge) containing bridge), plant (vegetation), building (building), water (water), the influence that different types of ground objects propagate radio magnetic wave have Difference, the similar cell of atural object in website coverage area correspond to communication environments and also answer similar, which can be used as cluster Index feature is chosen for adaptive propagation model.

(1) type of ground objects determines

Type of ground objects information can be from GIS data (full name in English: Geographic Information System, English abbreviation: GIS Data) in obtain, different maps is inconsistent to the name definition of atural object, needs exist for artificial Carry out type of ground objects matching, the present invention defines 4 kinds of types of ground objects, i.e. land (contain bridge), vegetation, building, Water is matched, and the matching result in a kind of embodiment is as shown in Figure 6.

(2) cell atural object statistics calculates

Join antenna height, angle of declination, vertical/horizontal lobe width in information according to cell work, can determine MPS process Close, far point coverage area, as shown in Figure 7a.

Fig. 7 a is the schematic diagram on antenna covering vertical section, corresponding with Fig. 7 a in order to facilitate understanding with Modeling Calculation The schematic diagram of horizontal plane is as shown in Figure 7b.

O (x in Fig. 7 a and 7b_o,y_o) indicate sites antenna position, OA indicate covering anomalistic distance, OB indicate covering far point away from From θ indicates horizontal lobe width.Dash area indicates overlay area, and the embodiment of the present invention also calculates dash area expression The atural object ratio of overlay area.

Steps are as follows for calculating:

Step 1: anomalistic distance OA, far point distance OB are determined；

It is known to obtain antenna height H (m) from work ginseng information, high, depth of building, which is hung, comprising antenna (notes: not including Atural object height above sea level), electrical down-tilting+mechanical tilt T (°), vertical lobe width V (°), then anomalistic distance OA calculation method are as follows:

Wherein, if OA >=2*ISD, OA=2*ISD；

Far point distance OB calculation method is as follows:

IfThen: OB=2*ISD；

IfThen:

Wherein, ISD indicates that this calculates average station spacing, calculation method around cell and sees propagation characteristic one.

Step 2: judge C (x_c,y_c) whether fall in shaded region

Known C (x_c,y_c) put and O (x_o,y_o) point, horizontal lobe width θ (°), antenna directional angle F (°), judge it is as follows:

If OA=OB, this cell is not calculated, and every kind of atural object ratio is set as 0；

Otherwise, OC distance meets: in [θ/2 F- θ/2, F+], C point position is fallen for OA≤OC≤OB&&OC and direct north angle In dash area, the terrestrial object information is obtained, and record.

Wherein, Fig. 7 c is please referred to, OC and direct north angle calcu-lation method are as follows:

Remember vectorThen vectorDeflection calculate by the following method: wherein X < 0

Step 3: computational shadowgraph part atural object information scales

All cells are traversed, dash area grid number M is calculated_all, determine atural object grid number N_clutter(N_land: it indicates Land atural object grid number, N_bridge: indicate bridge atural object grid number, N_vegetation: indicate vegetation atural object grid number, N_building: indicate building atural object grid number, N_water: indicate water atural object grid number), utilize N_clutter/N_allIt finds out every Class atural object proportion.

(3) propagation characteristic three: effective transmitter height

Transmitter actual height (hangs high Height comprising antenna_antenna, depth of building Height_building, atural object height above sea level Height Height_clutter) relative to receiver actual height (include object height Height_clutter, human height Height_body) it is effective transmitter height Height_tx-rx, schematic diagram is as shown in Figure 8.

Step 1: transmitter actual height

There are antennas to hang high Height for transmitter_antenna, depth of building Height_building, atural object height above sea level Height_clutter, therefore transmitter height Height_txCalculation formula is as follows:

Height_tx=Height_antenna+Height_building+Height_clutter

Above formula unit: m (rice).

Step 2: receiver actual height

There are object height Height for receiver_clutter, human height Height_body, therefore receiver actual height Height_rxCalculation formula is as follows:

Height_rx=Height_clutter+Height_body

Above formula unit: m (rice).

Step 3: effective transmitter height is calculated

Effective transmitter height Height_tx-rxCalculation formula is as follows:

Height_tx-rx=Height_tx-Height_rx

Above formula unit: m (rice).

(4) propagation characteristic four: atural object blocks influence on transmitter to RX path

Los (full name in English: Line of sight, Chinese meaning: sighting distance),；Expression does not have on electromagnetic wave transmission path Shelter can make free-space propagation.Nlos (full name in English: None line of sight, Chinese meaning: non line of sight), Indicate that shelter enters electromagnetic transmission channel, diffraction, reflection occurs so that signal deflects in this often, causes its arrival The time of receiving antenna is slightly later than direct signal.Since the signal and direct signal of these deflections have phase difference, so they Its power can be reduced or be completely counterbalanced by.Under Nlos scene, blocking for building is projected in the first luxuriant and rich with fragrance Neil area Serious shielding degree changes Electromagnetic Wave Propagation influence, and the masking ratio for blocking building maximum in Nlos is special as fitting Levy index.Los scene, masking ratio 0.

(1) first luxuriant and rich with fragrance Neil area

According to Huygens's Fei Nier principle, electromagnetic wave in transmission process, the every bit in wave surface be all one into The wave source of the spherical wave of row secondary radiation.Emit wave source to a line is connected between receiver, using this line as axle center, emits wave source It is the ellipse of focus with receiver, rotates 360 ° to obtain spheroid to be the area Fei Nier according to axle center.Luxuriant and rich with fragrance Neil area size with The distance between electronics wave frequency rate, receiver transmitter correlation.

In free space, the electromagnetic energy of transmitted from transmitter to receiver point mainly passes through the first luxuriant and rich with fragrance Neil area and is propagated, and examines Whether have the blocking of building, to determine that transmitter and receiver end belong to Los or NLos if considering in the first luxuriant and rich with fragrance Neil area. In order to carry out Modeling Calculation, the present invention takes vertical cross-section to calculate.If there is not building in the first luxuriant and rich with fragrance Neil area, It is Los between transmitter and receiver；Conversely, being then NLos.Specifically as illustrated in fig. 9.As it can be seen that it even if can be direct by Fig. 9 a Receiver is seen from transmitter, is also not necessarily free-space propagation.

(2) first luxuriant and rich with fragrance Neil area radiuses

As shown in figure 9b, Q is transmitting wave source, and P is receiver location, S₁Face is the vertically face with QP straight line, C₁For S₁Face with The section circle in luxuriant and rich with fragrance Neil area, F₁For the radius of section circle, d₁、d₂It is Q point and P point respectively to source C₁The distance in the center of circle, this circle institute Region be the first luxuriant and rich with fragrance Neil area.C₁Radius of circle F₂Calculation formula are as follows:

Wherein, λ is wavelength,F is frequency (Hz), and v indicates light velocity 3*10⁸m。

(3) Los/Nlos judges

Terrestrial object information data can be obtained from GIS Data, be needed in electronic map (miscellaneous comprising lattice level Clutter Disorderly) information, Building information.Judge that steps are as follows by transmitted from transmitter to receiver Los/Nlos:

Step 1: terrestrial object information material calculation is obtained

Input parameter: the accuracy of map

Using the accuracy of map as material calculation n_calcstep(m), it samples as terrestrial object information between transmitted from transmitter to receiver Input value, default obtain accuracy of map value, export n_calcstep(m)。

Step 2: transmitted from transmitter to receiver line is calculated

According to transmitter site coordinate (X_tx,Y_tx), when multiple antennas scene, obtain primary antenna position coordinates.Transmitter height Height_tx(m), receiver location coordinate (X_rx,Y_rx), receiver height Height_rx(m), then transmitter and receiver line water Flat distance d_h(m) it may be expressed as:

Transmitter may be expressed as: with receiver line space slope

k_{tx_rx}=(Height_tx-Height_rx)/d_h

Transmitter may be expressed as: with receiver line

Y=k_{tx_rx}*X+b

Wherein, b is constant.

Step 3: line coordinate shift amount is calculated

According to horizontal distance d between transmitter and receiver_h, atural object material calculation n_calcstep(m), then the total sampling of atural object time Number is d_h/n_calcstep, sample point, sample point Clutter are traversed from transmitter point_iPosition coordinates are relative to transmitter X_txPartially Shifting amountIt is represented by, (i, 1≤i≤d_h/n_calcstep):

Relative to transmitter Y_txOffsetIt may be expressed as:

Then Clutter_iPosition coordinates areAccording to Clutter_iPosition coordinates can obtain Take the grid point terrestrial object information (object height:Depth of building:), there is no building number According to when, will

Step 4: the first luxuriant and rich with fragrance Neil area radius F is calculatedⁱ

According to Clutter_iPosition coordinates and shelter height, then Clutter_iThe first luxuriant and rich with fragrance Neil area center of circle is to transmitter Distance can be with are as follows:

Clutter_iThe first luxuriant and rich with fragrance Neil area center of circle can be indicated to receiver distance are as follows:

Then Clutter_iFirst luxuriant and rich with fragrance Neil area radius FⁱAre as follows:

Wherein,

λ is wavelength,F is frequency (MHz), and v indicates light velocity 3*10⁸m。

Step 5: Clutter is calculated_iSlope between first luxuriant and rich with fragrance Neil area's radius and transmitter space

Wherein, Height_txIndicate transmitter actual height.

Step 6: Clutter is calculated_iShelter height is to slope between transmitter space

Step 7: judge Clutter_iWhether the propagation path of transmitted from transmitter to receiver is blocked:

IfThen Clutter_iPropagation path is blocked, Clutter_iFirst on vertical cross-section Luxuriant and rich with fragrance Neil area diameter 2*Fⁱ, calculation method:

Step 8: distance on the part vertical cross-section that is blocked

Calculation method:

IfThen

Otherwise,Then

Otherwise, Clutter_iIt is unobstructed to the propagation path of transmitted from transmitter to receiver, masking ratio

Epicycle terminates, and traverses next Clutter_i+1。

Step 9: traversal completes all atural objects, then the masking ratio ShelterRate of receiver to transmitter_ClutterAre as follows:

ShelterRate_ClutterAs receiver end feature.

(5) propagation characteristic five: distance is received

Horizontal distance between transmitter and receiver enables transmitter coordinate Tx (X_tx, Y_tx), receiver coordinate Rx (X_rx, Y_rx), then the reception distance d of receiver to transmitter:

Unit: m (rice).

(6) propagation characteristic six: building average height

Signal can be diffracted into receiver end, the average building in the first luxuriant and rich with fragrance Neil area by building roof, corner Height has an impact diffracting effect.

First, it is determined that Los/Nlos between transmitter and receiver, method is shown in feature four.

If it is Los, then it is assumed that building average height does not influence the propagation path of transmitted from transmitter to receiver, default Parameter is 0.

Otherwise, need to consider influence of the building average height to transmitted from transmitter to receiver communication environments.Signal by Influence is related to building average height after shelter blocks generation diffraction, reflection, and this programme will consider the first luxuriant and rich with fragrance Neil area throwing Building average height in shadow to two-dimensional surface.

As shown in Figure 10, calculation process includes:

(1) computation grid is traversed

According to cell transmitter Tx computer capacity, traverses computation grid and obtain Grid_list, obtain computation grid Grid_i。

(2) judge transmitter Tx to Grid_iNlos/Los relationship

Judgment method is shown in feature four.

(3) Tx to Grid is calculated_iThe first two-dimensional projection, luxuriant and rich with fragrance Neil area region；

Tx to Grid in reality_iThe first luxuriant and rich with fragrance Neil area be spatial ellipsoid body, scheme is right for the ease of modeling and calculating Spheroid carries out vertically obtaining an ellipse with the section of vertical section along axle center, and oval same size is projected to two-dimensional surface The upper building covered can be to Tx to Grid_iPropagation has an impact.From top to bottom from solid space, effect such as Figure 11 institute Show.

As shown in figure 11, the first luxuriant and rich with fragrance Neil area of transmitter Tx to Rx projects the grid that elliptical extraneous rectangle is passed through and builds Object information is built to be intended to be calculated.Steps are as follows for calculating:

Step 1: ellipse short shaft 2b, long axis 2a, focal length 2c are calculated.

Known Tx (X_tx, Y_tx) arrive Rx (X_rx, Y_rx) between distance, be elliptic focus distance 2c, calculation method is as follows:

Short axle 2b distance is the luxuriant and rich with fragrance Neil of transmitted from transmitter to receiver first area maximum radius, therefore calculation method is as follows:

Due to d₁=d₂=c, by above-mentioned formula abbreviation, semi-minor axis b are as follows:

Wherein, λ is wavelength,F is frequency (Hz), and v indicates light velocity 3*10⁸m。

According to ellipse properties, b²+c²=a², it can obtain:

Step 2: calculate long axis on rectangle intersecting point coordinate

It enables, the intersecting point coordinate Tx ' (X ' of the side Tx_tx, Y '_tx), the intersecting point coordinate Rx ' (X ' of the side Rx_rx, Y '_rx) indicate, Tx ' Distance to Tx is represented by (a-c), then Tx to Rx is indicated with vector:

Then the side Tx extends and rectangle intersecting point coordinate Tx ' (X '_tx, Y '_tx) may be expressed as:

The side Rx extends and rectangle intersecting point coordinate Rx ' (X '_rx, Y '_rx) may be expressed as:

Step 3: long axial short axle two sides offset coordinates are calculated

It is deviated according to long axis two sides coordinate points to two side direction of short axle perpendicular to long axis, four of rectangle can be obtained Apex coordinate is accuracy of map n to short-axis direction offset step_calcstep(m), long axial coordinate is offset to both sides frequency n_offset:

Long axis intersection point Tx ' to Rx ' is enabled to be indicated with vector

Perpendicular toOffset direction vector beThen:

Or

The both direction of offset indicated above, the Coordinate calculation method after being deviated in each direction according to offset step are as follows:

Switch (offset direction)

For (i=0；i<n_offset；i++){

Offset coordinates iTx ' the calculating of the end transmitter Tx ':

Offset coordinates iRx ' the calculating of the end receiver Rx ':

}

For (i=0；i<n_offset；i++){

Offset coordinates iTx ' the calculating of the end transmitter Tx ':

Offset coordinates iRx ' the calculating of the end receiver Rx ':

}

}

Coordinate (iX ' after each offset of this step output_tx,iY′_tx) and (iX '_rx,iY′_rx)。

Step 4: determine line by grid index according to two o'clock coordinate, material calculation

Upper step has determined that (iX '_tx,iY′_tx) and (iX '_rx,iY′_rx), material calculation uses accuracy of map n_calcstep, meter Calculation method can get the grid index that two o'clock line is passed through referring to propagation characteristic four, willWith The offset coordinates calculating of offset direction gets all by grid index, formation list iClutterList_index。

The iClutterList that will acquire_indexGrid index duplicate removal obtains ClutterList_index, ClutterList_indexInterior grid is grid in two-dimensional projection region in the first luxuriant and rich with fragrance Neil area.

(4) building average height AvgHeight in view field is calculated_building

Traverse ClutterList_index, obtain grid point atural object Clutter_i, judge Clutter_iType of ground objects whether be Building, if so, obtaining depth of buildingObject heightIt traverses next Clutter_i+1；Otherwise, next Clutter is traversed_i+1。

AvgHeight_buildingCalculation method:

(7) propagation characteristic seven: Adjacent Buildings density is received

The Effects of Density Electromagnetic Wave Propagation path loss of receiver Adjacent Buildings, i.e. simulation street effect.It obtains and receives The ratio of building in the certain grid of point surrounding.Calculation method are as follows:

According to the computer capacity (CalculateAreaofRateBuilding) of configuration, due north, due west, due south, positive east To each CalculateAreaofRateBuilding layers of grid as computer capacity, calculation method are as follows:

NumberBin_building: indicate the grid number for belonging to building in computer capacity；

NumberBin: the grid number in computer capacity is indicated.

(8) propagation characteristic eight: horizontal plane relative bearing

For electromagnetic wave in communication process, the decaying of antenna lobe inhibits electromagnetic wave propagation link load, therefore receiving point Direction relative to antenna main lobe has different inhibitory effects, horizontal lobe relative bearing (HorizontalAngle) As classification, fit characteristic.HorizontalAngle is calculated:

It as shown in figure 12, is horizontal lobe relative bearing HorizontalAngle schematic diagram.

OA is cell main lobe direction in Figure 12, and UE is receiving point, θ HorizontalAngle, and calculation method is as follows:

Step 1: obtaining cell level lobe and direct north angle is θ_cell

It is obtained from work ginseng.

Step 2: ask receiving point UE to subdistrict position O line and direct north angle theta_cell:

Enable O point coordinate O (x_o,y_o), UE point coordinate UE (x_ue,y_ue), θ_cellCalculation formula:

The angle theta calculation formula of Step 3:HorizontalAngle:

θ=| θ_ue-θ_cell|

If θ >=180 °, then:

θ=| 360- θ |

(9) propagation characteristic nine: vertical plane is with respect to tilt angled down

Transmitter is same as the vertical angle that receiving point is formed to influence antenna damping.Vertical lobe is with respect to angle of declination (VerticalAngle) as classification, fit characteristic.VerticalAngle is calculated:

It as shown in figure 13, is vertical lobe relative bearing VerticalAngle schematic diagram.

OA is the vertical main lobe direction of cell in Figure 13, and UE is receiving point, and θ is vertical lobe with respect to angle of declination (VerticalAngle), θ ' is antennas orthogonal face main lobe direction and horizontal sextant angle；H_txFor transmitter height, comprising antenna hang it is high, Depth of building, height above sea level；h_ueFor receiver height, include human height (default 1.5m), height above sea level；Then vertical lobe The calculation method step of opposite angle of declination (VerticalAngle) θ:

Step 1: θ ' is calculated

Electrical down-tilting+mechanical tilt T (°), vertical lobe width V (°) are enabled, T (°), V (°) are obtained from work ginseng, then the meter of θ ' Calculation method are as follows:

θ '=T+v/2

Step 2: horizontal distance d is calculated

Enable O point coordinate O (x_o,y_o), UE point coordinate UE (x_ue,y_ue), then the horizontal distance d of transmitter to receiving point are as follows:

Step 3: then vertical lobe is calculated with respect to angle of declination (VerticalAngle) θ are as follows:

If h_ue≥H_tx,

Otherwise,

(10) propagation characteristic ten: diffraction loss

Diffraction enables radio wave to pass through barrier, forms field strength namely diffraction field strength at the rear of barrier. Huygen law is thought, this is because each point in front of barrier, which can be used as new wave source, generates spherical surface secondary wave, it is secondary Grade wave is exactly diffracted wave field in the field that the rear of barrier is formed.

Unimodal diffraction refers to only having the case where knife-edge obstacle in the first Fresnel zone of dual-mode antenna, such as Mountain peak.

Multimodal diffraction refers to that there are two or more tooth shapes in the first Fresnel zone between receipts transmitting antenna The calculating of the case where barrier, multimodal diffraction loss are increasingly complex more than the calculating of unimodal diffraction loss, more there are commonly Bullington (Bu Lingdun) algorithm, Epstein and Peterson (Yi Pusi butyl- Peter is gloomy) algorithm, Deygout (wear height It is special) algorithm, Atlas (Etta Lars) four kinds of algorithms of algorithm.

The present invention calculates the diffraction loss of transmitted from transmitter to receiver, as one of fit characteristic.

Three, propagation characteristic (optional) is selected

Since different communication environments are possible to only related to specific several propagation characteristics, it is therefore desirable to according to practical biography It broadcasts situation to screen propagation characteristic, selects the higher propagation characteristic of correlation as classification, the fit characteristic factor, the present invention A kind of feature selection approach is provided:

Step 1: it calculates propagation characteristic and propagates correlation between path loss

Electromagnetic wave characteristics calculation method has been provided in step B, calculates all of every measured data transmitted from transmitter to receiver The correlation for propagating path loss, enables the propagation characteristic A ordered series of numbers be

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

Path loss Pathloss ordered series of numbers is propagated in order

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

Then the calculation formula of relative coefficient γ is as follows between propagation characteristic:

The low feature of phase relation is filtered out according to related coefficient thresholding, determines that phase relation is strong and weak according to following table 1, the present invention Recommend the feature for retaining weak correlation or more as classification fit characteristic index.

Table 1

Phase relation coefficient

Phase relation intensity

0.8-1.0	Extremely strong correlation
		0.6-0.8	Strong correlation
0.4-0.6	Moderate correlation
		0.2-0.4	Weak correlation
0.0-0.2	It is extremely weak related or without correlation

Step 2: propagation characteristic directly related property two-by-two is calculated

If strong correlation between two propagation characteristics carries out cluster fitting using one of propagation characteristic, in this way It can not only improve efficiency, while remove redundancy and improving cluster fitting effect.Enable propagation characteristic A ordered series of numbers Feature_a= {Feature_a(1),Feature_a(2),Feature_a(3),.....,Feature_a(n) }, propagation characteristic B ordered series of numbers Feature_b= {Feature_b(1),Feature_b(2),Feature_b(3),.....,Feature_b(n) }, then related between propagation characteristic A, B The calculation formula of property coefficient γ is as follows:

The present invention recommends two propagation characteristics A, B retaining strong correlation one of as classification fit characteristic, deletes two A propagation characteristic and the propagation lower propagation characteristic of path loss correlation.

Step 3: construction feature matrix

By transmitter terminal feature and receiver end feature association, eigenmatrix is formed, structure such as the following table 2:

Table 2

Three, cluster and construct random forest grader

(1) feature clustering

It after obtaining eigenmatrix, needs to cluster propagation characteristic, cluster mode provided in the present invention has two Kind, one is the clusters based on error of fitting, and one is industry classical taxonomy algorithm K-means.

(1) based on the cluster of error of fitting

Clustering method principle based on error of fitting is using non-ginseng fit mathematics method that propagation characteristic error is low and one The characteristic for determining thresholding is gathered for one kind, is another kind of by big the gathering of deviation.Specific step is as follows:

Step 1: propagation characteristic in traversal eigenmatrix

The propagation characteristic A that acquires is enabled to be

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

Step 2: by feature and the non-ginseng fitting of path loss progress is propagated

Propagation characteristic A and propagation path loss Pathloss are carried out non-ginseng to be fitted, the present invention recommends to be used as using GAM algorithm quasi- Hop algorithm.And error of fitting is calculated, calculation formula is as follows:

Meanerror=(m-p)/n

It is proposed that error of fitting value [- 0.15,0.15].

Step 3: propagation characteristic label

It is by label of the error of fitting in suggested range in propagation characteristic matrix, the propagation characteristic data mark outside range It is denoted as 2.Therefore, if there is the category combinations that the corresponding classification of N number of feature will have 2N kind, i.e. cluster number.

(2) K-means clustering procedure

K-means algorithm is a kind of indirect clustering method based on similarity measurement between sample, belongs to unsupervised learning side Method.Industry is widely used, and the present invention recommends to use the method.

(2) classifier is constructed

Classifier is constructed using propagation characteristic, it is therefore an objective to guarantee accurately find phase by propagation characteristic in prediction The propagation model of fit of classification is answered, classifier false segmentation rate is lower, and the accuracy for finding correct disaggregated model is higher, and the present invention uses Industry classics random forests algorithm constructs classifier.

Four, non-ginseng fitting generates or corrects propagation model

(1) non-ginseng model of fit

It is the fitting of multidimensional non parametric regression using propagation characteristic model of fit, the present invention recommends Generalized addictive models (English Full name: Generalized additive model, English abbreviation: GAM) algorithm.GAM algorithm is called using propagation characteristic, is intended It is as shown in figure 14 to close effect.

By verifying, present invention discover that passing through the linear fit effect band of the relatively traditional least square method of the non-ginseng fitting of GAM Carry out the gain of 1-3 standard deviation, under same propagation characteristic data cases, fit standard is poor, as shown in figure 15.It needs to illustrate It is that verify data is operated without data filtering etc. in Figure 15, for existing net most original data fitting result.

(2) model trend is corrected

It should be noted that this step is optional step.In practical application, if it find that certain class propagation model is not inconsistent rationally By trend, trend amendment can be carried out to such propagation model.

The measured data that the present invention is acquired using network is as input data, if collected propagation data is insufficient or passes It is insufficient to broadcast feature extraction range, then it is as shown in figure 16 to be likely to be obtained result.

Shown in Figure 16 the result is that in certain office point data some model fitting result, black line indicates collected propagation Path loss, red line are the not collected propagation path loss line of prediction.As can be seen that the prediction of model will occur not after 1000m Accurate situation.

Optimization method is as follows:

Step 1: construction general predictive model

By all propagation characteristics and path loss progress linear fit is propagated, obtains general lm model.

Step 2: construction universal model data

In computer capacity, there is no the case where measured data if there is the transmitted from transmitter to receiver, using general lm mould Type construction propagates path loss data.

Step 3: models fitting is carried out

The measured data for constructing data and actual acquisition is subjected to models fitting, the present invention recommends the non-ginseng fitting algorithm of GAM.

Five, Adaptive matching propagation model and propagation path loss calculate

Using the random forest grader constructed in preceding step, by the propagation characteristic in data to be predicted using random Forest algorithm, Adaptive matching carry out propagating path loss prediction to respective propagation model, propagate path loss square and it is possible to construct Battle array structure, such as shown in the following table 3.

Table 3

Transmitter

Receiving point 1

Receiving point 2

Receiving point 3

Receiving point 4

……

Receiving point n

Trx1

pathloss

pathloss

pathloss

pathloss

……

pathloss

Trx2

pathloss

pathloss

pathloss

pathloss

……

pathloss

Trx3

pathloss

pathloss

pathloss

pathloss

……

pathloss

……

Six, experiment effect

The final technical effect of the present invention and Conventional wisdom model (SPM), deterministic model (Volcano) are used as effect pair Than it is as follows to propagate path loss forecasting accuracy comparing result under identical data input condition:

(1) fit standard difference compares

Fitting effect can characterization model and propagate path loss between laminating degree, generally determined by calibration standard difference Whether the model can be used for the regional planning.The present invention passes through comparison deterministic model (Volcano model), empirical model The fit standard difference of (SPM:Standard Propagation Model) proves technical effect of the invention, such as Figure 17 institute Show.

(2) prediction percentage error comparison

Propagation model in addition to well more preferable in fitting effect, also to see in practical applications with the comparison of actual propagation path loss Effect.Control methods herein is that a part of data count product acceptance data as product acceptance data as correction data, partial data Middle error rate proves technical effect of the invention.Please refer to the effect of Hangzhou DT as shown in table 4 and as shown in table 5 The effect of Guangzhou MR.

(3) efficiency comparative

By verifying, using Hangzhou office point under equally planning scene condition (40,000,000 groups of transmitters and receiver), point Class model of fit is time-consuming: 13 minutes 1 hour, deterministic model Volcano was 11 minutes 4 hours time-consuming, model of fit efficiency of classifying Higher than 2 times of deterministic model.

To sum up, prediction being fitted the embodiment of the invention provides a kind of classification and propagating path loss method, this method considers to pass comprehensively Feature is broadcast, using non-ginseng approximating method, path loss is propagated in efficiently and accurately prediction.Technical solution of the embodiment of the present invention can be used as network rule Basic module is drawn, 2G/3G/4G network formats are suitable for, input data source includes CW/DT/MR data.

(embodiment two)

For the above scheme of the better implementation embodiment of the present invention, it is also provided below and implements above scheme for cooperating Relevant apparatus.

Figure 18 is please referred to, the embodiment of the present invention provides a kind of device 1800 that prediction propagation path loss is fitted based on classification, can Include:

Feature calculation module 1801, it is special for calculating electromagnetic wave propagation between transmitted from transmitter to receiver according to measured data Sign；

Clustering processing module 1802, for being clustered using at least one propagation characteristic to measured data, and construct with Machine forest classified device；

Model generation module 1803, it is corresponding for carrying out non-ginseng fitting generation to every a kind of measured data that cluster obtains Propagation model；

Matching primitives module 1804, for according to the propagation characteristic between transmitted from transmitter to receiver, using described random gloomy The corresponding propagation model of woods classifier Adaptive matching calculates the propagation path loss between transmitter and receiver.

In some embodiment of the invention, described device 1800 further include:

Filtering module 1805 is cleaned, for using one of conditional filter method and algorithms filter method or two Kind, cleaning filtering is carried out to the measured data of acquisition.

In some embodiment of the invention, the propagation characteristic includes one of following characteristics or a variety of: between average station Away from, cell coverage area atural object ratio, effective transmitter height, atural object blocks influence on transmitted from transmitter to receiver path, receives Distance, building average height, receiver Adjacent Buildings density, horizontal plane relative bearing, vertical plane with respect to tilt angled down, Diffraction loss.

In some embodiment of the invention, described device 1800 further include:

Feature selection module 1806, for according to propagation characteristic and propagating the correlation between path loss, from propagations spy Select at least one propagation characteristic as classification fitting condition in sign；

The clustering processing module 1802, specifically for being selected as classification fitting condition using the feature selection module At least one propagation characteristic measured data is clustered.

In some embodiment of the invention, the model generation module 1803, is also used to according to the propagation characteristic to not The propagation model for meeting theoretical trend is modified.

It is appreciated that each functional module of the device that prediction propagation path loss is fitted based on classification of the embodiment of the present invention Function can be implemented according to the method in above method embodiment, and specific implementation process can refer in above method embodiment Associated description, details are not described herein again.

Therefore in some feasible embodiments of the invention, discloses a kind of be fitted based on classification and predict to pass The device for broadcasting path loss calculates propagation characteristic using according to measured data；It is clustered using propagation characteristic, and constructs random forest Classifier；Corresponding propagation model is generated to every a kind of non-ginseng fitting of progress that cluster obtains；According to transmitted from transmitter to receiver it Between propagation characteristic calculate transmitter and reception using the corresponding propagation model of random forest grader Adaptive matching The technical characteristic of propagation path loss between machine, achieves following technical effect:

1, measured data is clustered using propagation characteristic, improves the utilization rate to measured data, so that a large amount of real Validity of the measured data in propagation path loss prediction is higher；

2, it is based on cluster result, classification fitting is carried out to measured data using non-ginseng approximating method, reduces calculation amount, mentions High efficiency；

3, the propagation model that classification fitting obtains is carried out to a large amount of measured datas based on propagation characteristic cluster and non-ginseng fitting, With higher precision of prediction, prediction result can be made more acurrate more effective；

4, the random forest grader constructed can be according to propagation characteristic Adaptive matching between the transmitter and receiver Corresponding propagation model improves the application of propagation model.

5, present invention method improves speed, precision and the accuracy for propagating path loss prediction, solves or reduces The number of drawbacks of the prior art can ensure the accuracy and reasonability of wireless network planning.

(embodiment three)

The embodiment of the present invention also provides a kind of computer storage medium, which can be stored with program, should Program includes the part that the method for predicting propagation path loss is fitted based on classification or complete recorded in above method embodiment when executing Portion's step.

(example IV)

Figure 19 is please referred to, the embodiment of the present invention also provides a kind of computer equipment 1900, it may include: processor 1910 is deposited Reservoir 1920, communication interface 1930, bus 1940；The memory 1920 is for storing computer executed instructions, the processing Device 1910 is deposited 1920 reservoirs and is connect by the bus 1940 with described, when the computer equipment 1900 operation, the place Reason device 1910 executes the computer executed instructions that the memory 1920 stores, so that the computer equipment 1900 executes Some or all of the method that prediction propagation path loss is fitted based on classification as described in the embodiment of the present invention one step.

The computer equipment clusters measured data using propagation characteristic, improves the utilization rate to measured data, So that validity of a large amount of measured datas in propagation path loss prediction is higher；Based on cluster result, using non-ginseng approximating method pair Measured data carries out classification fitting, reduces calculation amount, improves efficiency；Based on propagation characteristic cluster and the fitting of non-ginseng to a large amount of Measured data carries out the propagation model that classification fitting obtains, and has higher precision of prediction, can make prediction result is more acurrate more to have Effect；The random forest grader of building can Adaptive matching passes accordingly between the transmitter and receiver according to propagation characteristic Model is broadcast, the application of propagation model is improved；Speed, precision and the accuracy for propagating path loss prediction are improved, solves or subtracts Lack the number of drawbacks of the prior art, can ensure the accuracy and reasonability of wireless network planning.

In the above-described embodiments, it all emphasizes particularly on different fields to the description of each embodiment, is not described in some embodiment Part, may refer to the associated description of other embodiments.

It should be noted that for the various method embodiments described above, for simple description, therefore, it is stated as a series of Combination of actions, but those skilled in the art should understand that, the present invention is not limited by described sequence of movement because according to According to the present invention, certain steps can use other sequences or carry out simultaneously.Secondly, those skilled in the art should also know that, The embodiments described in the specification are all preferred embodiments, and not necessarily the present invention must for related actions and modules Must.

It is apparent to those skilled in the art that for convenience and simplicity of description, the system of foregoing description, The specific work process of device and unit, can refer to corresponding processes in the foregoing method embodiment, and details are not described herein.

In several embodiments provided herein, it should be understood that disclosed system, device and method can be with It realizes by another way.For example, the apparatus embodiments described above are merely exemplary, for example, the division of unit, Only a kind of logical function partition, there may be another division manner in actual implementation, such as multiple units or components can be with In conjunction with or be desirably integrated into another system, or some features can be ignored or not executed.Another point, it is shown or discussed Mutual coupling, direct-coupling or communication connection can be through some interfaces, the INDIRECT COUPLING of device or unit or Communication connection can be electrical property, mechanical or other forms.

Unit may or may not be physically separated as illustrated by the separation member, shown as a unit Component may or may not be physical unit, it can and it is in one place, or may be distributed over multiple networks On unit.It can some or all of the units may be selected to achieve the purpose of the solution of this embodiment according to the actual needs.

It, can also be in addition, the functional units in various embodiments of the present invention may be integrated into one processing unit It is that each unit physically exists alone, can also be integrated in one unit with two or more units.Above-mentioned integrated list Member both can take the form of hardware realization, can also realize in the form of software functional units.

It, can if integrated unit is realized in the form of SFU software functional unit and when sold or used as an independent product To be stored in a computer readable storage medium.Based on this understanding, technical solution of the present invention substantially or Say that all or part of the part that contributes to existing technology or the technical solution can embody in the form of software products Out, which is stored in a storage medium, including some instructions are used so that a computer equipment (can be personal computer, server or the network equipment etc.) executes all or part of each embodiment method of the present invention Step.And storage medium above-mentioned include: USB flash disk, it is mobile hard disk, read-only memory (ROM, Read-Only Memory), random Access various Jie that can store program code such as memory (RAM, Random Access Memory), magnetic or disk Matter.

Those of ordinary skill in the art will appreciate that all or part of the steps in the various methods of above-described embodiment is can It is completed with instructing relevant hardware by program, which can be stored in a computer readable storage medium, storage Medium may include: ROM, RAM, disk or CD etc..

Be provided for the embodiments of the invention above it is a kind of based on classification be fitted prediction propagate path loss method and apparatus into It has gone and has been discussed in detail, used herein a specific example illustrates the principle and implementation of the invention, the above implementation The explanation of example is merely used to help understand method and its core concept of the invention；Meanwhile for the general technology people of this field Member, according to the thought of the present invention, there will be changes in the specific implementation manner and application range, in conclusion this explanation Book content should not be construed as limiting the invention.

Claims (11)

1. a kind of be fitted the method that path loss is propagated in prediction based on classification characterized by comprising

Electromagnetic wave propagation feature between transmitted from transmitter to receiver is calculated 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；

According to the propagation characteristic between transmitted from transmitter to receiver, passed accordingly using the random forest grader Adaptive matching Model is broadcast, the propagation path loss between transmitter and receiver is calculated.

2. the method according to claim 1, wherein it is described according to measured data calculate transmitted from transmitter to receiver it Between before electromagnetic wave propagation feature, further includes:

Using one or both of conditional filter method and algorithms filter method, the measured data of acquisition is cleaned Filtering.

3. the method according to claim 1, wherein

The propagation characteristic includes one of following characteristics or a variety of: average station spacing, cell coverage area atural object ratio have Transmitter height is imitated, atural object blocks influence on transmitted from transmitter to receiver path, receives distance, building average height, receiver Adjacent Buildings density, horizontal plane relative bearing, vertical plane is with respect to tilt angled down, diffraction loss.

4. the method according to claim 1, wherein it is described using at least one propagation characteristic to measured data into Before row cluster further include:

According to the correlation between propagation characteristic and propagation path loss, at least one propagation characteristic is selected to make from the propagation characteristic For fitting condition of classifying；

It is described cluster is carried out to measured data using at least one propagation characteristic to include:

Measured data is clustered using at least one propagation characteristic for being selected as classification fitting condition.

5. method according to any one of claims 1-4, which is characterized in that described according between transmitted from transmitter to receiver Propagation characteristic, before the corresponding propagation model of random forest grader Adaptive matching, further includes: according to the biography Feature is broadcast to be modified the propagation model for being unsatisfactory for theoretical trend.

6. a kind of be fitted the device that path loss is propagated in prediction based on classification characterized by comprising

Feature calculation module, for calculating electromagnetic wave propagation feature between transmitted from transmitter to receiver according to measured data；

Clustering processing module for clustering using at least one propagation characteristic to measured data, and constructs random forest point Class device；

Model generation module generates corresponding propagating mode for carrying out non-ginseng fitting to every a kind of measured data that cluster obtains Type；

Matching primitives module, for utilizing the random forest grader according to the propagation characteristic between transmitted from transmitter to receiver The corresponding propagation model of Adaptive matching calculates the propagation path loss between transmitter and receiver.

7. device according to claim 6, which is characterized in that further include:

Filtering module is cleaned, for using one or both of conditional filter method and algorithms filter method, to acquisition Measured data carry out cleaning filtering.

8. device according to claim 6, which is characterized in that

The propagation characteristic includes one of following characteristics or a variety of: average station spacing, cell coverage area atural object ratio have Transmitter height is imitated, atural object blocks influence on transmitted from transmitter to receiver path, receives distance, building average height, receiver Adjacent Buildings density, horizontal plane relative bearing, vertical plane is with respect to tilt angled down, diffraction loss.

9. device according to claim 6, which is characterized in that further include:

Feature selection module, for being selected from the propagation characteristic according to the correlation between propagation characteristic and propagation path loss At least one propagation characteristic is as classification fitting condition；

The clustering processing module, specifically for being selected as described in classification fitting condition extremely using the feature selection module A kind of few propagation characteristic clusters measured data.

10. according to the device any in claim 6-9, which is characterized in that

The model generation module is also used to repair the propagation model for being unsatisfactory for theoretical trend according to the propagation characteristic Just.

11. a kind of computer equipment, which is characterized in that the computer equipment includes that processor, memory, bus and communication connect Mouthful；For storing computer executed instructions, the processor is connect with the memory by the bus memory, when When the computer equipment is run, the processor executes the computer executed instructions of the memory storage, so that institute It states computer equipment and executes the method according to any one of claims 1 to 5 that prediction propagation path loss is fitted based on classification.