CN101146313B - A correction method of radio transmission model - Google Patents

Abstract

The invention discloses a correction method for wireless transmission modes. The method comprises (A) collecting the test data in at least one test zone, grouping the test data of each test zone, and determining transmission environment characteristic parameters of each test data group; (B) sorting the test data groups to allow the test data groups similar in transmission environment characteristic parameters in the same one group; and (C) correcting the same wireless transmission mode by using the test data groups sorted in the same group. The invention can increase the correction efficiencyof the wireless transmission modes and improve the portability of the wireless transmission modes.

Description

A kind of bearing calibration of radio transmission model

Technical field

The present invention relates to the cellular radio Communication technology, relate in particular to a kind of method that the radio transmission model of wireless communication system is proofreaied and correct.

Background technology

At present, radio transmission model is the important evidence of predicted path loss in the wireless communication network planning, and the accuracy of propagation model has very big influence for the quality of the network planning.Reached today of considerable scale in the wireless network construction, a lot of zones have no longer needed to have obtained the test data of correcting wireless propagation model by setting up transmitter, because this bearing calibration is not only consuming time but also effort.

In order to substitute above-mentioned bearing calibration, Chinese patent application numbers 200310112566.5, open day be on November 17th, 2004, name of patent application in the patent application document of " a kind of radio network optimization method of cdma system ", a kind of method that the pilot signal strength of opening the back website is proofreaied and correct radio transmission model of gathering of utilizing has been proposed.This method utilizes existing drive test data to carry out the correction of radio transmission model, can use manpower and material resources sparingly; Because each sector can be proofreaied and correct and be obtained different radio transmission models, therefore improved precision of prediction to greatest extent simultaneously.

But, the focusing on of technical scheme in the above-mentioned patent application isolated different sector pilot signals intensity and utilized its correcting wireless propagation model, though this technical scheme has farthest been considered the otherness between different test zone communication environments, but do not consider its common ground, the test data that each test zone is gathered can only be proofreaied and correct a radio transmission model, if for example network has the website of 30 three sectors, need test data (totally 90 groups of test datas) to proofread and correct the radio transmission model that obtains 90 test zones so according to each sector, workload is very big, even proofread and correct by optimizing data, its workload is also not little.Therefore, the correction efficient of this prior art scheme is very low.And because the test data that each test zone is gathered can only be proofreaied and correct a radio transmission model, it is very poor therefore to cause proofreading and correct the radio transmission model transplantability that obtains.

Summary of the invention

In view of this, technical problem to be solved by this invention is to provide a kind of bearing calibration of radio transmission model, proofreaies and correct efficient to improve, and strengthens the portability of radio transmission model.

In order to realize the foregoing invention purpose, main technical schemes of the present invention is:

A kind of bearing calibration of radio transmission model, this method comprises:

Test data in A, above test zone of collection, the test data of each test zone is formed one group, determines the communication environments characteristic parameter of described each test data set;

B, described test data set is classified, the test data set that will possess similar communication environments characteristic is included into same class, and the test data of described test zone comprises the pilot signal strength of test zone at least;

C, proofread and correct same radio transmission model with being included into of a sort test data set;

The described test data set that will possess similar communication environments characteristic is included into same class:

B1, with the communication environments characteristic parameter Sector of described each test data set respectively as a sample;

B2, utilize proximity rules heuristic from described sample, to determine the number L of classification and the cluster centre Z of each classification _k, wherein k is more than or equal to 1, smaller or equal to L;

B3, L the cluster centre Z that step b2 is determined _kAs initial cluster center, adopt the K-averaging method that described sample is carried out cluster, obtain L classification; The test data set of each sample correspondence is included in the classification under this sample.

Preferably, steps A is described determines that the communication environments method of characteristic parameters of the corresponding test data set of each test zone specifically comprises:

A1, obtain the electronic three-dimensional map of all test zones, obtain the type of ground objects sum M of all test zones;

A2, add up total number of test points N of this test zone _i, subscript i represents the sequence identifier of this test zone;

A3, determine each test point in this test zone corresponding type of ground objects in this test zone electronic three-dimensional map;

A4, add up the number of test points N on each type of ground objects in this test zone _Ic, the c in the subscript represents c kind type of ground objects;

A5, determine the communication environments characteristic parameter Sector of this test data set _iFor:

Sector _i＝(Clu _i1，Clu _i2，…Clu _iM)

Wherein, Clu is the number of test points N of a certain type of ground objects _IcWith total number of test points N _iPercentage, the subscript of Clu represents that the sequence identifier of this type of ground objects, M represent the type of ground objects sum in all test zones.

Preferably, step b2 specifically comprises:

B21, selection sample Sector ₁As a cluster centre Z ₁, and set a non-negative threshold value T;

B22, initialization cluster centre Z _kSubscript k=1, the subscript i=2 of sample, total classification number L=1;

B23, determine Sector _iWith cluster centre Z _kBetween distance D _IkIf, D _IkSmaller or equal to T, then with Sector _iBe assigned to Z _kIn the classification for the center, and execution in step b25; Otherwise execution in step b24;

B24, judge whether that all known cluster centres all calculate and finish, if, with Sector _iBe made as new cluster centre, assignment L=L+1, execution in step b25; Otherwise, assignment k=k+1, and return step b23;

B25, judge whether that all samples have all calculated and finish, finish, then finish this step b2, otherwise assignment i=i+1, assignment k=1 return step b23 if calculate.

Preferably, among the step b23, described Sector _iWith Z _kBetween distance D _IkFor:

$D_{\mathrm{ik}}=\left(\sqrt{\underset{j=1}{\overset{M}{\mathrm{\Σ}}}{({\mathrm{Clu}}_{\mathrm{ij}}-{\mathrm{Clu}}_{\mathrm{kj}})}^2}\right).$

Preferably, described step b3 specifically comprises:

B31, for described each sample, from the cluster centre of described L classification, find the cluster centre nearest apart from this sample, with this sample dispensing to being in the classification at center with this nearest cluster centre;

B32, in described each classification, determine the average of interior all samples of this classification, with the new cluster centre of this average as this classification;

B33, judge whether the new cluster centre of each classification is identical with original cluster centre, if identical, then cluster is finished, the test data set of each sample correspondence is included in the classification under this sample; Otherwise, replace this original cluster centre of classifying with the new cluster centre of each classification, return step b31.

Preferably, among the described step b31, a certain sample and distances of clustering centers are:

$D_{\mathrm{ik}}=\left(\sqrt{\underset{j=1}{\overset{M}{\mathrm{\Σ}}}{({\mathrm{Clu}}_{\mathrm{ij}}-{\mathrm{Clu}}_{\mathrm{kj}})}^2}\right).$

Preferably, the test data of described test zone comprises the pilot signal strength of test zone at least.

Because the present invention carries out grouping and classifying according to the similitude of communication environments characteristic to test data, utilize of a sort test data set to proofread and correct same radio transmission model, therefore with respect to prior art, the present invention has reduced the workload of correcting wireless propagation model, has improved correction efficient.Simultaneously, owing to taken into full account the similitude of the communication environments feature of test data, so the portability of the radio transmission model that goes out of institute of the present invention verification is stronger, and a radio transmission model goes for like a plurality of environment facies in the test zone.

In addition, the present invention has also adopted the clustering method of science, test data characteristic in making on the same group by iteration is similar as much as possible, and the test data characteristic between is not on the same group separated as much as possible, and the classification that realizes automation is operated, therefore compare the method for artificial classification, both improved the accuracy of sorting out grouping, improved the efficient of test data packets again.

Description of drawings

Fig. 1 is the flow chart of the method for the invention;

Fig. 2 is the flow chart of communication environments characteristic parameter of the test data set correspondence of each test zone of extraction of the present invention;

Fig. 3 is a flow chart of the test data set of different test zones being classified automatically according to the communication environments parameter of the present invention;

Fig. 4 is the proximity rules heuristic of utilization of the present invention is determined the cluster centre of the number of classification and each classification from described test data set a flow chart.

Embodiment

Below by specific embodiments and the drawings the present invention is described in further details.

Fig. 1 is the flow chart of the method for the invention.Referring to Fig. 1, method of the present invention comprises:

Step 11, the test data of gathering an above test zone, the test data of each test zone is formed one group.

Described test zone for example can be a test zone, also can be the sector.Described test data can be to optimize test data, wherein comprises pilot signal strength (Rx Pilot Power) at least, can isolate the pilot signal strength of different test zones by calculating.Concrete grammar comprises the steps 111 to step 113:

Step 111, the forward direction travelling carriage in collecting test zone receives gross power (Rx Power) and the pilot tone signal to noise ratio intensity (Ec/Io) that comes from different test zones that receives at least.

Step 112, for the test position of each test point, the pilot signal strength Rx Pilot Power (dBm) of the different test zones that receive can be calculated by following formula (1):

(Rx Pilot Power) _i＝(Ec/Io) _i+Rx Power (1)

In the formula (1), i represents the label of different test zones.

Step 113, will save as different files respectively from the pilot signal strength of different test zones, each file is used to preserve the test data of a test zone, the test data that is to say a test zone is formed a test data set, at least can comprise in the test data set that a test data is a pilot signal strength, described label i also can the corresponding test data set of representing the i test zone.Determine that test data file adds up to G.

Step 12, from the test data set of each test zone of receiving, extract communication environments characteristic parameter separately.

Step 13, according to the communication environments characteristic parameter test data set of different test zones is classified automatically, the test data set that will possess similar communication environments characteristic is included into same class.

Step 14, the sorted test data set of foundation are finished model tuning, promptly proofread and correct same radio transmission model with being included into of a sort test data set.For example, be specifically as follows: sorted test data set is imported to be had in the planning simulation software of automatic correcting wireless propagation model function, and software can obtain the proper model parameter from dynamic(al) correction according to test data.The technology that described planning simulation software can utilize existing a plurality of test data to proofread and correct a radio transmission model is proofreaied and correct, and repeats no more herein.

Fig. 2 is the flow chart of the communication environments characteristic parameter of the test data set correspondence of described each test zone of extraction of step 12 of the present invention.Referring to Fig. 2, the idiographic flow of described extraction communication environments characteristic parameter comprises:

Step 21, obtain the electronic three-dimensional map of all test zones.This electronic three-dimensional map should comprise landform and terrestrial object information; And obtain type of ground objects sum M in all test zones.

Step 22, obtain the communication environments characteristic parameter that the test data set from i test zone comprises, specifically may further comprise the steps 221 to step 224:

Step 221, add up the total number of test points N in this test zone _i

Step 222, travel through all numbers of test points, determine each test point corresponding type of ground objects on electronic three-dimensional map according to longitude and latitude.

Test point sum N on step 223, the different types of ground objects of statistics _Ic, calculate the percentage Clu that test point sum on the different types of ground objects accounts for total number of test points of this test zone _IcSpecifically referring to formula (2)

${\mathrm{Clu}}_{\mathrm{ic}}=\frac{N_{\mathrm{ic}}}{N_i}\×100\%---\left(2\right)$

In the formula (2), described N _IcIt is i test data set (i.e. the test data of i test zone) test point sum on c kind type of ground objects; N _iIt is total number of test points of i test zone.

Step 224, determine the communication environments characteristic parameter of the test data set of i test zone, the following formula of computing formula (3):

${\mathrm{Sector}}_i=({\mathrm{<figure-callout\; id="1"\; label="Clu\; i"\; filenames="S2007101760204E00011.png,S2007101760204E00021.png"\; state="\{\{state\}\}">Clu</figure-callout>}}_i,{\mathrm{Clu}}_{i2},...{\mathrm{Clu}}_{\mathrm{iM}})=(\frac{N_{i1}}{N_i}\&times;100\%,\frac{N_{i2}}{N_i}\&times;100\%,...\frac{N_{\mathrm{iM}}}{N_i}\&times;100\%)---\left(3\right)$

In the above-mentioned formula (3), described M is the type of ground objects sum in all test zones.

Step 23, judge whether all test data set have all been calculated and finish that if do not have, then assignment i=i+1 continues to utilize above-mentioned steps 221 to step 224 to calculate the communication environments characteristic parameter of next test data set, otherwise process ends.

Fig. 3 is a flow chart of the test data set of different test zones being classified automatically according to the communication environments parameter of the present invention.Referring to Fig. 3, this flow process comprises:

Step 31, with the communication environments characteristic parameter Sector of a described G test data set ₁, Sector ₂..., Sector _GAs sample, utilize proximity rules heuristic from described sample, to determine the number L of classification and the cluster centre Z of each classification _k, wherein k is more than or equal to 1, smaller or equal to L.

This step 31 is used proximity rules, and to sound out ratio juris as follows:

Be provided with n sample: X1, X2 ... Xn might as well make arbitrary sample as cluster centre Z1, and chooses arbitrary non-negative threshold value T, for simplicity, we select X1=Z1, calculate the distance D 21 of X2 to Z1 then, if D21＞T then sets up a new cluster centre Z2, and X2=Z2, if D21＜T, think that then X2 is being in the territory at center with Z1, i.e. X1, X2 belongs to a class together.Calculate X3 then respectively to Z1, the distance of Z2 obtains D31, D32, if D31＞T, D32＞T then sets up a new cluster centre Z3, and X3=Z3, otherwise X3 is divided in the territory of nearest cluster centre.With similar method all samples are calculated distance, compare threshold, decision ownership.When algorithm finishes, just can obtain C class, C cluster centre, and the sample number in each class.

Fig. 4 is the proximity rules heuristic of utilization of the present invention is determined the cluster centre of the number of classification and each classification from described test data set a flow chart.Referring to Fig. 4, this flow process may further comprise the steps 311 to step 315:

Step 311, selection Sector ₁As cluster centre Z ₁, and set a non-negative threshold value T.

Step 312, initialization cluster centre subscript k=1, test data set subscript i=2, total classification number L=1.

Step 313, obtain cluster centre Z _k, calculate Sector _iWith cluster centre Z _kBetween distance D _Ik

Sector _iWith Z _kBetween distance D _IkThe following formula of computing formula (4):

$D_{\mathrm{ik}}=|{\mathrm{Sector}}_i-Z_k|$

$=\left(\sqrt{\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}{({\mathrm{Clu}}_{\mathrm{ij}}-{\mathrm{Clu}}_{\mathrm{kj}})}^2}\right)---\left(4\right)$

In the formula (4), described i represents different test data set; K represents different cluster centres; J represents different types of ground objects; Sector _i, Z _kIt is respectively the communication environments characteristic parameter of i test data set and k cluster centre.

Step 314～step 315, judgement D _IkWhether smaller or equal to T, if, so with Sector _iBe assigned to Z _kIn the classification for the center, and change step 318; Otherwise, carry out next step.

Step 316～step 317, judge whether that all known cluster centres all calculate and finish, if, with Sector _iBe made as new cluster centre, the classification number L=L+1 that assignment is total carries out next step; Otherwise assignment k=k+1 returns step 313.

Step 318～step 319, judge whether that all samples have all calculated and finish, if then obtain the cluster centre Z of preliminary classification number L and each classification _k(1≤k≤L), flow process finishes, otherwise assignment i=i+1, k=1 return step 313.

In following steps 32 to step 35, L the cluster centre Z that step 31 is determined _kAs initial cluster center, adopt the K-averaging method that described sample is carried out cluster, obtain L classification; The test data set of each sample correspondence is included in the classification under this sample.

Described K-average ratio juris is: at first select k object as initial cluster center arbitrarily from n data object; And, then, respectively they are distributed to the cluster the most similar, i.e. the cluster of cluster centre representative to it according to the similarity (being the distance between the communication environments feature in the present invention) of they and these cluster centres for other object of be left; And then calculate the average of all objects in this cluster, with the new cluster centre of this average as this cluster; Constantly repeat this process till the canonical measure function begins convergence.Generally all adopt mean square deviation as the canonical measure function.A described k cluster has following characteristics: each cluster itself is compact as much as possible, and separates as much as possible between each cluster.

In the present invention, sampling K-averaging method described sample is carried out cluster detailed process referring to following steps 32 to step 35.

Step 32, traversal Sector ₁, Sector ₂..., Sector _G, find the nearest cluster centre Z of distance sample point for each sample _k, with sample Sector _iBe assigned to Z _kIn the classification for the center.This step specifically may further comprise the steps 321 to step 324:

Step 321, initialization sample subscript i=1.

Step 322, calculating Sector _iWith all cluster centre Z _k(the distance D between 1≤k≤L) _Ik, and find minimum range D _IkMinCorresponding k value is specifically referring to formula (5).

$D_{\mathrm{ikMin}}=\mathrm{Min}|{\mathrm{Sector}}_i-Z_k|$

$=\mathrm{Min}\left(\sqrt{\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}{({\mathrm{Clu}}_{\mathrm{ij}}-{\mathrm{Clu}}_{\mathrm{kj}})}^2}\right)---\left(5\right)$

In the formula (5), i represents different test data set; K represents different cluster centres; J represents different types of ground objects, Sector _i, Z _kIt is respectively the communication environments characteristic parameter of i test data set and k cluster centre.

Step 323, with Sector _iBe assigned to Z _kIn the classification for the center.

Step 324, judge whether that all samples all calculate and finish, if, execution in step 33; Otherwise assignment i=i+1 returns step 322.

Step 33, recomputate described L the classification new cluster centre Z _k'; Specifically may further comprise the steps 331 to step 333:

Step 331, initialization k=1.

Step 332, obtain with Z _kBe the number Q of sample in the class at center, and recomputate such center; New cluster centre Z _k' be with Z _kBe the average of all samples in the class at center, the following formula of computing formula (6):

${Z_k}^{\&prime;}=\frac{\underset{p=1}{\overset{Q}{\mathrm{\&Sigma;}}}{\mathrm{Sector}}_p}{Q}=(\frac{\underset{p=1}{\overset{Q}{\mathrm{\&Sigma;}}}{\mathrm{Clu}}_{p1}}{Q},\frac{\underset{p=1}{\overset{Q}{\mathrm{\&Sigma;}}}{\mathrm{Clu}}_{p2}}{Q},...,\frac{\underset{p=1}{\overset{Q}{\mathrm{\&Sigma;}}}{\mathrm{Clu}}_{\mathrm{pM}}}{Q})---\left(6\right)$

In the formula (6), described Clu _PjBe with Z _kCommunication environments characteristic parameter component for sample in the class at center.

Step 333, judge whether L cluster centre all recomputates and finish, finish execution in step 34 if calculate; Otherwise assignment k=k+1 returns step 332.

L the cluster centre Z that step 34～step 35, judgement newly obtain _k' the cluster centre Z that whether obtains with last iteration _kIdentical, if identical, then stop iteration, cluster is finished, and obtains L classification, and the test data set of each sample correspondence is included into classification under this sample, and flow process finishes; Otherwise, assignment Z _k=Z _k', return step 32.

Below in conjunction with concrete parameter, be example so that the test data correcting wireless propagation model is carried out in the a-quadrant, further specify above-mentioned method.

Step 51, suppose to have in the a-quadrant 40 websites, amount to 120 test zones.Optimization of collection test data in the a-quadrant at first.

Step 52, separation obtain the pilot signal strength Rx Pilot Power from these 120 test zones, and save as different files respectively.

Step 53, obtain the electronic three-dimensional map of a-quadrant, obtain the type of ground objects of this electronic chart, they are respectively the open ground (openinurban), greenery patches (greenland), forest (forest), residential area (residential) in open ground (open), the city, general city (meanurban), dense city (denseurban) and industrial area (industrial) 8 classes altogether.

The communication environments characteristic parameter that step 54, described 120 test datas of calculating comprise, as shown in table 1 is the communication environments characteristic parameter of different test datas:

Sector	open	Opening- urban	greenLand	forest	Residen- tial	Mean- urban	Dense- urban	industrial
									Sector 1	0.00％	30.20％	5.25％	0.00％	0.00％	30.60％	33.95％	0.00％
Sector 2	0.00％	29.60％	0.00％	0.00％	5.43％	26.50％	29.15％	9.32％
									…	…	…	…	…	…	…	…	…
Sector 119	21.65％	12.35％	23.32％	5.30％	20.80％	0.00％	0.00％	16.58％
									Sector 120	18.45％	14.50％	17.67％	7.50％	23.10％	0.00％	0.00％	18.78％

Table 1

Below with the test data Sector in the table 1 ₁Be example, computational methods are described, the computational methods of all the other test datas are identical therewith;

541), Sector ₁Test data sum on 8 kinds of atural object open, openinurban, greenland, forest, residential, meanurban, denseurban, industrial is respectively: 0,766,133,0,0,776,861,0.Set Clu ₁₁, Clu ₁₂..., Clu ₁₈Represent that successively test point sum on these 8 kinds of atural objects accounts for the percentage of data point sum in the test data 1, then:

Clu ₁₂＝[766/(766+133+776+861)]×100％＝30.2％

Clu ₁₃＝[133/(766+133+776+861)]×100％＝5.25％

Clu ₁₆＝[776/(766+133+776+861)]×100％＝30.6％

Clu ₁₇＝[861/(766+133+776+861)]×100％＝33.95％

Clu ₁₁＝Clu ₁₄＝Clu ₁₅＝Clu ₁₈＝0％

542), obtain Sector ₁The communication environments characteristic parameter be:

Sector ₁＝(0％，30.2％，5.25％，0％，0％，30.6％，33.95％，0％)

Step 55, with the communication environments characteristic parameter of described 120 test datas as sample, set threshold T=10%, calculate 5 initial testing data qualifications according to proximity rules heuristic, as shown in table 2 is 5 initial test datas classification:

The cluster numbering	Cluster centre	Number of samples	Sample number
				Zone1	Z1＝Sector 1	13	Sector 1、Sector 3、…
Zone2	Z1＝Sector 2	25	Sector 2、Sector 4、…
				Zone3	Z1＝Sector 24	36	Sector 24、Sector 26、…
Zone4	Z1＝Sector 78	34	Sector 78、Sector 82、…
				Zone5	Z1＝Sector 89	12	Sector 89、Sector 96、…

Table 2

The concrete steps that obtain above-mentioned 5 initial testing data qualifications are as follows:

551) select Sector ₁As cluster centre Z ₁, calculate Sector ₂With Z ₁Between distance D ₂₁

${\mathrm{<figure-callout\; id="21"\; label="D"\; filenames="S2007101760204E00021.png"\; state="\{\{state\}\}">D</figure-callout>}}_{21}=\left(\sqrt{{(30.20-29.60)}^2+{(5.25-0)}^2+{(0-5.43)}^2+{(30.6-26.5)}^2+{(33.95-29.15)}^2+{(0-9.32)}^2}\right)\%$

$\&ap;13.6\%$

552) because D ₂₁＞10%, so with Sector ₂Be set at new cluster centre Z ₂

553) calculate Sector according to identical method ₃With Z ₁, Z ₂Between distance, if with certain distances of clustering centers less than 10%, then with Sector ₃Belong in the class at this cluster centre place; If distance is all greater than threshold value 10%, then with Sector ₃Be set at new cluster centre.

554) the rest may be inferred, calculates until all test datas to finish, and obtains 5 initial testing data qualifications at last.

Step 56,120 samples of traversal find the nearest cluster centre Z of distance sample point to each sample _k, with sample Sector _iBe assigned to Z _kIn the cluster for the center.

Step 57, recomputate new cluster centre Z _k', be example explanation computational methods with class Zone1 below, the computational methods of other cluster centre are identical therewith;

${Z_1}^{\text{'}}=\frac{\underset{i=1}{\overset{13}{\mathrm{\&Sigma;}}}{\mathrm{<figure-callout\; id="13"\; label="Sector\; i"\; filenames="S2007101760204E00011.png"\; state="\{\{state\}\}">Sector</figure-callout>}}_i}{13}$

Wherein, Sector _iBe 13 all samples that comprise among the Zone1, the 13rd, the sum of sample among the Zone1.

Step 58, judge the cluster centre Z that all newly obtain _k' whether with before cluster centre Z _kIdentical, if different, then will calculate Z _k' replace former Z as new cluster centre _k, repeating step 56,57,58 no longer changes until the cluster centre that newly obtains; If all identical, then change step 510.

Step 510, obtain the classification of final test data, as shown in table 3:

The cluster numbering	Cluster centre	Number of samples	Sample number
				Zone1	The mean value of Z1=18 sample	18	Sector1、Sector3、…
Zone2	The mean value of Z1=25 sample	25	Sector2、Sector4、…
				Zone3	The mean value of Z1=31 sample	31	Sector24、Sector29、…
Zone4	The mean value of Z1=30 sample	30	Sector78、Sector85、…
				Zone5	The mean value of Z1=16 sample	16	Sector89、Sector97、…

Table 3

Step 511, described 5 class testing data are imported in the planning simulation software successively, utilize software automodel calibration function can obtain 5 groups of propagation models.

The above; only for the preferable embodiment of the present invention, but protection scope of the present invention is not limited thereto, and anyly is familiar with the people of this technology in the disclosed technical scope of the present invention; the variation that can expect easily or replacement all should be encompassed within protection scope of the present invention.

Claims (6)

1. the bearing calibration of a radio transmission model is characterized in that, this method comprises:

Test data in A, above test zone of collection, the test data of each test zone is formed one group, determines the communication environments characteristic parameter of described each test data set;

B, described test data set is classified, the test data set that will possess similar communication environments characteristic is included into same class, and the test data of described test zone comprises the pilot signal strength of test zone at least;

C, proofread and correct same radio transmission model with being included into of a sort test data set;

The described test data set that will possess similar communication environments characteristic is included into same class:

B1, with the communication environments characteristic parameter Sector of described each test data set respectively as a sample;

B2, utilize proximity rules heuristic from described sample, to determine the number L of classification and the cluster centre Z of each classification _k, wherein k is more than or equal to 1, smaller or equal to L;

B3, L cluster centre Zk that step b2 is determined are as initial cluster center, and employing K-averaging method is carried out cluster to described sample, obtains L classification; The test data set of each sample correspondence is included in the classification under this sample.

2. the bearing calibration of radio transmission model according to claim 1 is characterized in that, steps A is described determines that the communication environments method of characteristic parameters of the corresponding test data set of each test zone specifically comprises:

A1, obtain the electronic three-dimensional map of all test zones, obtain the type of ground objects sum M of all test zones;

A2, add up total number of test points N of this test zone _i, subscript i represents the sequence identifier of this test zone;

A3, determine each test point in this test zone corresponding type of ground objects in this test zone electronic three-dimensional map;

A4, add up the number of test points N on each type of ground objects in this test zone _Ic, the c in the subscript represents c kind type of ground objects;

A5, determine the communication environments characteristic parameter Sector of this test data set _iFor:

Sector _i＝(Clu _i1，Clu _i2，...Clu _iM)

Wherein, Clu is the number of test points N of a certain type of ground objects _IcWith total number of test points N _iPercentage, the subscript of Clu represents that the sequence identifier of this type of ground objects, M represent the type of ground objects sum in all test zones.

3. the bearing calibration of radio transmission model according to claim 1 is characterized in that, step b2 specifically comprises:

B21, selection sample Sector ₁As a cluster centre Z ₁, and set a non-negative threshold value T;

B22, initialization cluster centre Z _kSubscript k=1, the subscript i=2 of sample, total classification number L=1;

B23, determine Sector _iWith cluster centre Z _kBetween distance D _IkIf, D _IkSmaller or equal to T, then with Sector _iBe assigned to Z _kIn the classification for the center, and execution in step b25; Otherwise execution in step b24;

B24, judge whether that all known cluster centres all calculate and finish, if, with Sector _iBe made as new cluster centre, assignment L=L+1, execution in step b25; Otherwise, assignment k=k+1, and return step b23;

B25, judge whether that all samples have all calculated and finish, finish, then finish this step b2, otherwise assignment i=i+1, assignment k=1 return step b23 if calculate.

4. the bearing calibration of radio transmission model according to claim 3 is characterized in that, among the step b23, and described Sector _iWith Z _kBetween distance D _IkFor:

$D_{\mathrm{ik}}=\left(\sqrt{\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}{({\mathrm{Clu}}_{\mathrm{ij}}-{\mathrm{Clu}}_{\mathrm{kj}})}^2}\right).$

5. the bearing calibration of radio transmission model according to claim 1 is characterized in that, described step b3 specifically comprises:

B31, for described each sample, from the cluster centre of described L classification, find the cluster centre nearest apart from this sample, with this sample dispensing to being in the classification at center with this nearest cluster centre;

B32, in described each classification, determine the average of interior all samples of this classification, with the new cluster centre of this average as this classification;

B33, judge whether the new cluster centre of each classification is identical with original cluster centre, if identical, then cluster is finished, the test data set of each sample correspondence is included in the classification under this sample; Otherwise, replace this original cluster centre of classifying with the new cluster centre of each classification, return step b31.

6. the bearing calibration of radio transmission model according to claim 5 is characterized in that, among the described step b31, a certain sample and distances of clustering centers are: $D_{\mathrm{ik}}=\left(\sqrt{\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}{({\mathrm{Clu}}_{\mathrm{ij}}-{\mathrm{Clu}}_{\mathrm{kj}})}^2}\right).$

Images