CN109738926B - A kind of GNSS multipath effect correcting method based on BP neural network technology - Google Patents
A kind of GNSS multipath effect correcting method based on BP neural network technology Download PDFInfo
- Publication number
- CN109738926B CN109738926B CN201910254859.8A CN201910254859A CN109738926B CN 109738926 B CN109738926 B CN 109738926B CN 201910254859 A CN201910254859 A CN 201910254859A CN 109738926 B CN109738926 B CN 109738926B
- Authority
- CN
- China
- Prior art keywords
- multipath
- data
- multipath errors
- satellite
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 42
- 238000000034 method Methods 0.000 title claims abstract description 40
- 230000000694 effects Effects 0.000 title claims abstract description 39
- 238000005516 engineering process Methods 0.000 title claims abstract description 26
- 238000012544 monitoring process Methods 0.000 claims abstract description 62
- 238000012549 training Methods 0.000 claims abstract description 21
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 238000012986 modification Methods 0.000 claims abstract description 4
- 230000004048 modification Effects 0.000 claims abstract description 4
- 238000012545 processing Methods 0.000 claims description 36
- 238000010606 normalization Methods 0.000 claims description 19
- 210000002569 neuron Anatomy 0.000 claims description 10
- 230000008030 elimination Effects 0.000 claims description 6
- 238000003379 elimination reaction Methods 0.000 claims description 6
- 238000012805 post-processing Methods 0.000 claims description 6
- 239000000284 extract Substances 0.000 claims description 5
- 239000000203 mixture Substances 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 3
- 125000004122 cyclic group Chemical group 0.000 abstract description 4
- 230000007613 environmental effect Effects 0.000 abstract description 3
- 230000000737 periodic effect Effects 0.000 abstract description 3
- 238000000605 extraction Methods 0.000 abstract 1
- 238000004321 preservation Methods 0.000 abstract 1
- 238000012937 correction Methods 0.000 description 6
- 239000011159 matrix material Substances 0.000 description 4
- 238000005259 measurement Methods 0.000 description 3
- 210000005036 nerve Anatomy 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 239000005433 ionosphere Substances 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- DMBHHRLKUKUOEG-UHFFFAOYSA-N diphenylamine Chemical compound C=1C=CC=CC=1NC1=CC=CC=C1 DMBHHRLKUKUOEG-UHFFFAOYSA-N 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 230000002401 inhibitory effect Effects 0.000 description 1
- 230000000116 mitigating effect Effects 0.000 description 1
- 210000004218 nerve net Anatomy 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000010183 spectrum analysis Methods 0.000 description 1
- 230000003313 weakening effect Effects 0.000 description 1
Landscapes
- Position Fixing By Use Of Radio Waves (AREA)
Abstract
The present invention provides a kind of GNSS multipath effect correcting method based on BP neural network technology successively includes Multipath Errors extractions, sample pre-selection, sample preprocessing, sample training, network topology structure determination, data preservation, Data Post and model modification；By being modeled using the signal forward-propagating of BP neural network algorithm and the backpropagation of error to Multipath Errors, multipath effect error caused by motion conditions, elevation of satellite, signal-to-noise ratio and the monitoring point ambient environmental conditions of satellite can effectively be weakened, the cyclic fluctuation in tri- direction result monitoring station X, Y, Z trained simultaneously be improved significantly, and dynamic update can be carried out to model according to newly-increased Multipath Errors data, can utmostly weaken multipath effect periodic error caused by monitoring point.
Description
Technical field
The present invention relates to multipath effect technical fields, particularly, are related to a kind of GNSS based on BP neural network technology
Multipath effect correcting method.
Background technique
GNSS (Global Navigation Satellite System) is the general designation of various navigation system, comprising:
The navigation system such as GPS, BDS, GLONASS, Galileo.With the rapid development of satellite navigation system, GNSS technology is being led
It is got a lot of applications in the multiple fields such as boat, deformation monitoring, positioning, time service.Due to GNSS it is round-the-clock, it is full-automatic, be not necessarily to people
The characteristics of work is intervened is widely used in deformation monitoring.
GNSS obtains the deformation of monitored object in real time in deformation monitoring using short baseline relative positioning technology, due to
Baseline is shorter, and most error has been effectively eliminated by filtering and differential technique.However multipath effect can not pass through
These means effectively eliminate.Multipath effect seriously damages the precision of GNSS measurement, and the losing lock of signal can be also caused when serious, is
A kind of important error source in GNSS measurement.
There are mainly three types of the existing effective ways for weakening Multipath Errors: (1) selecting to block when acquiring data less
Place；(2) algorithm for inhibiting multipath effect is added in GNSS receiver or using the antenna for having choke coil；(3) it adopts
Multipath effect is rejected with post-processing algorithm.The position of general monitoring point is determining in the project, can not be had highly desirable
Spacious environment, so first method and being not suitable for；Second method needs to improve within hardware, needs receiver manufacturer
To optimize；Application in the project generally can weaken Multipath Errors using the third method.
Simplest post-processing algorithm is to select star method, using it is commonplace be to establish multiple path routing model method.Two kinds of sides
Method is being affected, based on this principle smaller to the image of carrier observations to Pseudo-range Observations first by multipath effect
Calculate the multipath effect of different satellites；Star method is wherein selected mainly to defend according to the Multipath Errors Select Error being calculated is lesser
Star participates in resolving；Modelling establishes the multipath effect model of every satellite according to more days Multipath Errors, then according to model
The Pseudo-range Observations of each satellite are corrected.Currently, multipath modeling method mainly pass through fitting of a polynomial,
Kalman filter technology, spectrum analysis, signal-to-noise ratio-Multipath Errors, elevation angle-Multipath Errors realize that multipath changes to model
Just.
Above-mentioned multipath mitigation method has some limitations: selecting star method fairly simple, but this method is being seen
Measure satellite number it is less in the case where, rejecting excessive observation data can be such that calculation accuracy reduces, and can be used only in satellite situation very
Good place；For modelling, although since multipath effect has certain periodicity, its numerical value and satellite motion feelings
Condition, receiver ambient enviroment have a much relations, rule performance it is more complicated, can not by simply filtering, signal-to-noise ratio, height
It spends angle and models accurate modeling, so the above modeling method equally exists very big difficulty for multipath correction, and can be used only in
In post-processing, it can not be used in real time data processing.
It is badly in need of a kind of new technique of GNSS multipath effect correcting method based on BP neural network technology in the industry.
Summary of the invention
It is an object of that present invention to provide a kind of GNSS multipath effect correcting method based on BP neural network technology, with solution
Certainly the complicated difficult of Multipath Errors can not be calculated with accurate modeling and in real time to resolve multipath effect and correct the technology of error and be asked
Topic.
To achieve the above object, a kind of GNSS multipath effect correction based on BP neural network technology provided by the invention
Method is implemented on the basis of BP neural network technical application, and BP neural network algorithm is implied including an input layer, one
Layer and an output layer, 2 times wherein to stand on the basis of input layer number with the common number of satellites in monitoring station, output layer has
Three neurons, respectively the coordinate time sequence in tri- direction of base station and monitoring station X, Y, Z；BP neural network activation primitive is adopted
With Sigmoid function:
Specifically includes the following steps:
(1), Multipath Errors extract；One period base station of continuous acquisition and the almanac data of monitoring station, carrier wave and
1) Pseudo-range Observations handle the data of acquisition as the following formula, the multipath for extracting base station and monitoring station difference satellite misses
Poor MP sequence；MP sequence by several epoch Mp1、Mp2Multipath Errors value composition；Simultaneously to formula 1) M that is calculatedp1、Mp2
Multipath Errors value carries out sample pre-selection, obtains the Multipath Errors MP sequence not comprising ambiguity information；
Formula 1) in f1、f2For the frequency of carrier phase observation data；φ1、φ2For carrier phase observation data；P1, p2 are pseudorange
Observation；Mp1、Mp2For Multipath Errors value；
Sample pre-selection in step (1), specifically: 1. according to formula 1) acquire the M of each satellite current epochp1、Mp2Multichannel
The mean value of diameter error amount；2. will be according to formula 1) M of subsequent each epoch that is calculatedp1、Mp2Multipath Errors value is individually subtracted
1. mean value that step calculates obtains the M not comprising ambiguity informationp1、Mp2Multipath Errors value；To obtain being gone through by several
Member and the M for not including ambiguity informationp1、Mp2The MP sequence data of Multipath Errors value composition.
(2), sample preprocessing: the multichannel that will be calculated in step (1) by the method for the multiple fitting of a polynomial of unitary
Elimination of rough difference in diameter error MP sequence obtains the clean MP sequence data not comprising rough error；
The multiple fitting of a polynomial of unitary is the fitting of unitary cubic polynomial, specifically carries out elimination of rough difference as follows:
(2.1), assume polynomial equation are as follows: y=a0+a1x+a2x2+a3x3；
(2.2), the sum of square of deviations of Multipath Errors are as follows:(2.3), according to R2Minimum principle finds out polynomial
Coefficient: a0、a1、a2、a3；
(2.4), the fitting Multipath Errors value of various time points is found out using multinomial, and the multipath after fitting is missed
Difference presses formula 2) processing, it will be greater than the data rejecting of the middle error of 3 times of Multipath Errors values；
Formula 2) in yjFor coefficient a0、a1、a2、a3The match value at j moment, M after fittingp1j、Mp2jFor the Multipath Errors at j moment
Value, 3V are the middle error of 3 times of Multipath Errors values.
Double deference processing is carried out to the MP sequence data after the excluding gross error of satellite common in monitoring station and base station, is obtained
The Double deference time series data of Multipath Errors, then Double deference time series data is normalized；
Double deference processing includes single poor processing and double difference processing, single poor processing are as follows: will be each common in monitoring station and base station
M after the excluding gross error of satellitep1、Mp2Data press formula 3) processing；
Formula 3) in MSupervise p1i、MSupervise p2i、MBase p2i、MBase p1iI-th satellite in monitoring station and base station being calculated for step (1)
The Multipath Errors of p1, p2 frequency range；Md1i、Md2iFor i-th satellite p1、p2Single difference Multipath Errors value of frequency range；
Double difference processing are as follows: the M that will be obtained after single poor processingd1i、Md2iSingle difference Multipath Errors value and reference satellite it is more
Tracking error value makees difference processing, forms Double deference Multipath Errors value；Reference satellite is selected as base station or monitoring station monitors
The maximum satellite of elevation angle；The maximum satellite of elevation angle for selecting base station or monitoring station to monitor as reference satellite, be by
Bigger in the elevation angle of satellite, the signal strength of satellite is higher, the signal of satellite launch by ionosphere and it is tropospheric influence it is small,
It is more accurate true.
Formula 4) in Md1ij、Md2ijFor the Double deference Multipath Errors of i-th satellite and reference satellite j moment p1, p2 frequency range
Value.
Normalized are as follows: by the data M in Double deference time seriesd1ij、Md2ijBy formula 5) processing, while will normalization
M ' afterwardsd1ij、M′d2ijDouble deference Multipath Errors data save respectively；
Md1ij、Md2ijIt is the Double deference Multipath Errors value of i-th satellite, Md1ijmin、Md2ijminIt is that Double deference multipath misses
The minimum value of data, M in differenced1ijmax、Md2ijmaxIt is the maximum value of data in i-th satellite Double deference Multipath Errors, M 'd1ij、
M′d2ijIt is the Double deference Multipath Errors data after normalization.
The input data obtained after sample preprocessing is the matrix data of n row 2m column, such as formula 6) shown in；Wherein n is
The epoch number of difference Multipath Errors data, station, the common number of satellites in monitoring station on the basis of m；It is each to be classified as an input data,
So the input layer of neural network is 2m；
T in formulanM′d1mRepresent tnThe Double deference Multipath Errors of moment (epoch) the m satellite p1 frequency range；tnM′d2mGeneration
Table tnThe Double deference Multipath Errors of the m satellite p2 frequency range of moment.
(3), sample training and post-processing: being arranged corresponding training parameter, pre- to sample using BP neural network algorithm
Treated, and data are trained, wherein input is formula 6) shown in data in n row 2m column matrix, export and resolve software for GNSS
Read the coordinate time sequence in tri- directions monitoring station X, Y, Z that training obtains；Software, which is resolved, by GNSS reads trained obtain
Tri- directions monitoring station X, Y, Z coordinate time sequence, multipath corrected value is predicted, and correct tri- sides X, Y, Z
To coordinate value.
Preferably, network topology structure determines specifically: according to sample data training as a result, calculating root-mean-square error
Rmse adjusts different hidden layer neuron numbers, network weight, then recalculates rmse, after repeating repeatedly, selection
The smallest neuron number of rmse and network weight are as optimum network topological structure；
Formula 7) in n be prediction points, o be network desired output, p be network reality output.
(4), model modification: when data of monitoring point length reaches a period in step (1), by newly-increased observation number
It has been obtained before being merged into according to the Double deference Multipath Errors data after the normalization obtained by step (1) and step (2) processing
To normalization after Double deference Multipath Errors data in form the Double deference Multipath Errors data after new normalization, and
Again new samples data are trained, according to the training result of new samples data, update model parameter；Model parameter is BP mind
Through the network weight and biasing coefficient in network.
The invention has the following advantages:
A kind of GNSS multipath effect correcting method based on BP neural network technology of the present invention, by utilizing BP nerve net
The signal forward-propagating of network algorithm and the backpropagation of error model Multipath Errors, can effectively weaken the movement of satellite
Multipath effect error caused by situation, elevation of satellite, signal-to-noise ratio and monitoring point ambient environmental conditions, trains simultaneously
Tri- direction result monitoring station X, Y, Z cyclic fluctuation be improved significantly, and can be according to newly-increased Multipath Errors
Data carry out dynamic update to model, can utmostly weaken multipath effect periodic error caused by monitoring point.
It is more three times using unitary in a kind of GNSS multipath effect correcting method based on BP neural network technology of the present invention
Item formula carries out Multipath Errors fitting, is conducive to Multipath Errors and the satellite motion period of each satellite Pseudo-range Observations of monitoring
Periodically correlation is presented, so that error curve is relatively smooth, regression criterion is relatively small.
A kind of GNSS multipath effect correcting method based on BP neural network technology of the present invention, by using at Double deference
Can effectively eliminate the common error to satellite monitoring of monitoring station and base station, at the same double difference treated Multipath Errors into
Row difference more meets the carrier phase difference algorithm of GNSS software resolving, so as to reach better Correction of Errors effect.
A kind of GNSS multipath effect correcting method based on BP neural network technology of the present invention, passes through normalized energy
It is enough that the value range of Double deference treated Multipath Errors data is effectively contracted to (0,1), greatly promote subsequent BP nerve
The training operation efficiency of network model and the resolving efficiency of GNSS software.
Other than objects, features and advantages described above, there are also other objects, features and advantages by the present invention.
Below with reference to figure, the present invention is described in further detail.
Detailed description of the invention
The attached drawing constituted part of this application is used to provide further understanding of the present invention, schematic reality of the invention
It applies example and its explanation is used to explain the present invention, do not constitute improper limitations of the present invention.In the accompanying drawings:
Fig. 1 is BP neural network structural schematic diagram of the present invention；
Fig. 2 is difference Multipath Errors figure (a, P1 frequency range difference in the embodiment of the present invention after the normalization of No. 1 satellite
Multipath Errors, b, P2 frequency range difference Multipath Errors)；
Fig. 3 is model predication value and initial data comparison diagram (a, X-direction mould of the present invention after BP neural network algorithm
Type predicted value and initial data compare, and b, Y-direction model predication value and initial data compare, c, Z-direction model predication value and original
Beginning data to).
Specific embodiment
The embodiment of the present invention is described in detail below in conjunction with attached drawing, but the present invention can be limited according to claim
Fixed and covering multitude of different ways is implemented.
A kind of GNSS multipath effect correcting method based on BP neural network technology provided by the invention is in BP nerve
Implement on the basis of application of net, BP neural network algorithm includes an input layer, a hidden layer and an output layer
(referring to Fig. 1), 2 times wherein to stand on the basis of input layer number with the common number of satellites in monitoring station, there are three output layers
Neuron, respectively the coordinate time sequence in tri- direction monitoring point X, Y, Z；BP neural network activation primitive uses Sigmoid letter
Number:
A kind of GNSS multipath effect correcting method based on BP neural network technology of the present invention, specifically includes following step
It is rapid:
(1), Multipath Errors extract；One period base station of continuous acquisition and the almanac data of monitoring station, carrier wave and
Pseudo-range Observations, to the data (almanac data, carrier wave and Pseudo-range Observations) of acquisition as the following formula 1) it handles, extract base station
The Multipath Errors MP sequence of satellites different with monitoring station；MP sequence by several epoch Mp1、Mp2Multipath Errors value group
At；Simultaneously to formula 1) M that is calculatedp1、Mp2Multipath Errors value carries out sample pre-selection, obtains not comprising ambiguity information
Multipath Errors MP sequence；The period acquired in the present embodiment is 30 days；
Formula 1) in f1、f2For the frequency of carrier phase observation data；φ1、φ2For carrier phase observation data；P1, p2 are pseudorange
Observation；Mp1、Mp2For Multipath Errors value；
Sample pre-selection in step (1), specifically: 1. according to formula 1) acquire the M of each satellite current epochp1、Mp2Multichannel
The mean value of diameter error amount；2. will be according to formula 1) M of subsequent each epoch that is calculatedp1、Mp2Multipath Errors value is individually subtracted
1. mean value that step calculates obtains the M not comprising ambiguity informationp1、Mp2Multipath Errors value；To obtain being gone through by several
Member and the M for not including ambiguity informationp1、Mp2The MP sequence data of Multipath Errors value composition.
(2), sample preprocessing: the multichannel that will be calculated in step (1) by the method for the multiple fitting of a polynomial of unitary
Elimination of rough difference in diameter error MP sequence obtains the clean MP sequence data not comprising rough error；
The multiple fitting of a polynomial of unitary is the fitting of unitary cubic polynomial, specifically carries out elimination of rough difference as follows:
(2.1), assume polynomial equation are as follows: y=a0+a1x+a2x2+a3x3；
(2.2), the sum of square of deviations of Multipath Errors are as follows:
(2.3), according to R2Minimum principle finds out polynomial coefficient: a0、a1、a2、a3；
(2.4), the fitting Multipath Errors value of various time points is found out using multinomial, and the multipath after fitting is missed
Difference presses formula 2) processing, it will be greater than the data rejecting of the middle error of 3 times of Multipath Errors values；
Formula 2) in yjFor coefficient a0、a1、a2、a3The match value at j moment, M after fittingp1j、Mp2jFor the Multipath Errors at j moment
Value, 3V are the middle error of 3 times of Multipath Errors values.
It is more three times using unitary in a kind of GNSS multipath effect correcting method based on BP neural network technology of the present invention
Item formula carries out Multipath Errors fitting, is conducive to Multipath Errors and the satellite motion period of each satellite Pseudo-range Observations of monitoring
Periodically correlation is presented, so that error curve is relatively smooth, regression criterion is relatively small.
Due to generally using carrier phase difference algorithm in deformation monitoring, it need to be to each total in monitoring station and base station
Double deference processing is carried out with the MP sequence data after the excluding gross error of satellite, obtains the Double deference time series of Multipath Errors,
Then the data in Double deference time series are normalized；Double deference processing includes single poor processing and double difference processing；
Single poor processing are as follows: by the M after the excluding gross error of common satellite each in monitoring station and base stationp1、Mp2Data press formula 3)
Processing；
Formula 3) in MSupervise p1i、MSupervise p2i、MBase p2i、MBase p1i, i-th satellite in monitoring station and base station for being calculated for step (1)
The Multipath Errors of p1, p2 frequency range；Md1i、Md2iFor single difference Multipath Errors value of i-th satellite p1, p2 frequency range；
Double difference processing are as follows: the M that will be obtained after single poor processingd1i、Md2iSingle difference Multipath Errors value and reference satellite it is more
Tracking error value makees difference processing, forms Double deference Multipath Errors value；The elevation angle that base station or monitoring station monitor is maximum
Satellite is as reference satellite；The maximum satellite of elevation angle for selecting base station or monitoring station to monitor as reference satellite, be due to
The elevation angle of satellite is bigger, and the signal strength of satellite is higher, the signal of satellite launch by ionosphere and it is tropospheric influence it is small, more
It is accurate true.
Formula 4) in Md1ij、Md2ijFor the Double deference multipath of i-th satellite and reference satellite j moment (epoch) p1, p2 frequency range
Error amount.
By using the common error to satellite monitoring that can effectively eliminate monitoring station and base station at Double deference, simultaneously
Double difference treated Multipath Errors carry out the carrier phase difference algorithm that difference more meets the resolving of GNSS software, so as to reach more
Good Correction of Errors effect.
Normalized are as follows: by the data M in Double deference time seriesd1ij、Md2ijBy formula 5) processing, while will normalization
M ' afterwardsd1ij、M′d2ijDouble deference Multipath Errors data save respectively；
Md1ij、Md2ijIt is the Double deference Multipath Errors value of i-th satellite, Md1ijmin、Md2ijminIt is that Double deference multipath misses
The minimum value of data, M in differenced1ijmax、Md2ijmaxIt is the maximum value of data in i-th satellite Double deference Multipath Errors, M 'd1ij、
M′d2ijIt is the Double deference Multipath Errors data after normalization.
Effectively the value range of Double deference treated Multipath Errors data can be contracted to by normalized
(0,1) greatly promotes the training operation efficiency of subsequent BP neural network model and the resolving efficiency of GNSS software, more meets reality
Applicable cases.
The input data obtained after sample preprocessing is the matrix data of n row 2m column, such as formula 6) shown in；Wherein n is
The epoch number of difference Multipath Errors data, station, the common number of satellites in monitoring station on the basis of m；It is each to be classified as an input data,
So the input layer of neural network is 2m；
T in formulanM′d1mRepresent tnThe Double deference Multipath Errors of moment (epoch) the m satellite p1 frequency range；tnM′d2mGeneration
Table tnThe Double deference Multipath Errors of the m satellite p2 frequency range of moment.
(3), sample training and post-processing: being arranged corresponding training parameter, pre- to sample using BP neural network algorithm
Treated, and data are trained, wherein input is formula 6) shown in data in n row 2m column matrix, export and resolve software for GNSS
Read the coordinate time sequence in tri- directions monitoring station X, Y, Z that training obtains；Software, which is resolved, by GNSS reads trained obtain
Tri- directions monitoring station X, Y, Z coordinate time sequence, multipath corrected value is predicted, and correct tri- sides X, Y, Z
To coordinate value.
A kind of GNSS multipath effect correcting method based on BP neural network technology includes that network topology structure determines, tool
Body are as follows: according to sample data training as a result, calculating root-mean-square error rmse, adjusts different hidden layer neuron numbers, net
Network weight, then recalculates rmse, repeat repeatedly after, select the smallest neuron number of rmse and network weight as
Optimum network topological structure；
Formula 7) in n be prediction points (epoch number), o is the desired output of network, i.e. the time sequence in tri- direction monitoring point xyz
Column, p are the reality output of network.
(4), model modification: when data of monitoring point length reaches 30 days, newly-increased observation data by step (1) and are walked
Suddenly the double difference after the normalization that the Double deference Multipath Errors data after the normalization that (2) processing obtains have obtained before being merged into
Form the Double deference Multipath Errors data after new normalization in point Multipath Errors data, and again to new samples data into
Row training, the training result according to new samples data calculate root-mean-square error rmse, readjust the model ginseng of BP neural network
Number, model parameter are network weight and biasing coefficient in BP neural network.
15 common satellites observation in 30 days in base station and monitoring station measurement gnss big-dipper satellite is selected in the present embodiment
The basic data that base station and monitoring station are observed successively is carried out step (1) and step by basic data of the data as the present embodiment
Suddenly the processing (formula 1) of (2), formula 2), formula 3), formula 4), 5) processing), the Double deference multipath after obtaining the normalization of each satellite misses
Difference data；Double deference Multipath Errors data after normalization (since satellite is more, only show No. 1 Beidou as shown in Figure 2 here
P1, P2 frequency range Double deference Multipath Errors time series of satellite)；The input layer nerve of BP neural network algorithm in the present embodiment
First number is 2 × 15=30；Output layer neuron number is three, is the coordinate time sequence in tri- directions monitoring station X, Y, Z
Column；Double deference Multipath Errors data after the normalization of each satellite are subjected to sample training, setting convergence error is 2mm, training
Number 5000；It is black shown in solid in result such as Fig. 3 after training.
As can be drawn from Figure 3, through a kind of GNSS multipath effect correction side based on BP neural network technology of the present invention
After method, the cyclic fluctuation that GNSS resolves the coordinate time sequence in tri- directions monitoring station X, Y, Z that software calculates is more original
Data have clear improvement, and illustrate that it is effective that multipath correction is carried out using BP neural network algorithm；Also illustrate this hair
It is bright by being modeled using the signal forward-propagating of BP neural network algorithm and the backpropagation of error to Multipath Errors, can
Effectively weaken multipath caused by motion conditions, elevation of satellite, signal-to-noise ratio and the monitoring point ambient environmental conditions of satellite to imitate
The cyclic fluctuation in tri- direction result monitoring station X, Y, Z answering error, while training is obviously improved, and being capable of root
Dynamic update is carried out to model according to newly-increased Multipath Errors data, can utmostly weaken multipath effect and monitoring point is made
At periodic error.
Claims (6)
1. a kind of GNSS multipath effect correcting method based on BP neural network technology, which comprises the following steps:
(1), Multipath Errors extract: the almanac data, carrier wave and pseudorange of continuous acquisition one period base station and monitoring station
Observation handles the data of acquisition, extracts the Multipath Errors MP sequence of base station and each satellite in monitoring station；MP sequence
Arrange the M by several epochp1、Mp2Multipath Errors value composition；Simultaneously to the M in MP sequencep1、Mp2Multipath Errors value carries out
Sample pre-selection, obtains the Multipath Errors MP sequence not comprising ambiguity information；
(2), sample preprocessing: the multipath being calculated in step (1) is missed by the method for the multiple fitting of a polynomial of unitary
Elimination of rough difference in poor MP sequence obtains the clean MP sequence data not comprising rough error；
Double deference processing is carried out to the MP sequence data after the excluding gross error of satellite common in monitoring station and base station, obtains multichannel
The Double deference time series data of diameter error, then Double deference time series data is normalized；
Wherein, single poor processing are as follows: by the M after the excluding gross error of common satellite each in monitoring station and base stationp1、Mp2Data press formula 3)
Processing；
Formula 3) in MSupervise pli、MSupervise p2i、MBase p2i、MBase p1iMonitoring station and base station i-th satellite p1, the p2 being calculated for step (1)
The Multipath Errors of frequency range；Md1i、Md2iFor single difference Multipath Errors value of i-th satellite p1, p2 frequency range；
Double difference processing are as follows: the M that will be obtained after single poor processingd1i、Md2iThe multipath of single difference Multipath Errors value and reference satellite
Error amount makees difference processing, forms Double deference Multipath Errors value；The height that reference satellite is selected as base station or monitoring station monitors
Spend the maximum satellite in angle；
Formula 4) in Md1ij、Md2ijFor the Double deference Multipath Errors value of i-th satellite and reference satellite j moment p1, p2 frequency range；
(3), sample training and post-processing: being arranged corresponding training parameter, using BP neural network algorithm to sample preprocessing
Data afterwards are trained, and obtain the coordinate time sequence in tri- directions monitoring station X, Y, Z；Software, which is resolved, by GNSS reads instruction
The coordinate time sequence in tri- directions monitoring station X, Y, the Z got, predicts multipath corrected value, and correct X, Y, Z
The coordinate value in three directions；
(4), model modification: when data of monitoring point length reaches a period in step (1), newly-increased observation data are led to
What the Double deference Multipath Errors data after crossing step (1) and the obtained normalization of step (2) processing had obtained before being merged into
Double deference Multipath Errors data after forming new normalization in Double deference Multipath Errors data after normalization, and again
New samples data are trained, model parameter is updated.
2. a kind of GNSS multipath effect correcting method based on BP neural network technology according to claim 1, special
Sign is, the M in step (1) the MP sequencep1、Mp2Multipath Errors value is by formula 1) it is calculated:
Formula 1) in f1、f2For the frequency of carrier phase observation data；φ1、φ2For carrier phase observation data；P1, p2 are pseudorange observation
Value；Mp1、Mp2For Multipath Errors value；
Step (1) in sample pre-selection, specifically: 1. according to formula 1) acquire the M of each satellite current epochp1、Mp2Multipath misses
The mean value of difference；2. will be according to formula 1) M of subsequent each epoch that is calculatedp1、Mp2Step is individually subtracted in Multipath Errors value
1. the mean value calculated obtains the M not comprising ambiguity informationp1、Mp2Multipath Errors value；To obtaining by several epoch and
M not comprising ambiguity informationp1、Mp2The MP sequence data of Multipath Errors value composition.
3. a kind of GNSS multipath effect correcting method based on BP neural network technology according to claim 1, special
Sign is, the multiple fitting of a polynomial of unitary in the step (2) is the fitting of unitary cubic polynomial, specifically as follows into
Row elimination of rough difference:
Y=a0+a1x+a2x2+a3x3:
(2.1), assume polynomial equation are as follows:
(2.2), the sum of square of deviations of Multipath Errors are as follows:
(2.3), according to R2Minimum principle finds out polynomial coefficient: a0、a1、a2、a3；
(2.4), the fitting Multipath Errors value of various time points is found out using multinomial, and by the Multipath Errors value after fitting
By formula 2) processing, it will be greater than the data rejecting of the middle error of 3 times of Multipath Errors values；
Formula 2) in yjFor coefficient a0、a1、a2、a3The match value at j moment, M after fittingp1j、Mp2jFor the Multipath Errors value at j moment,
3V is the middle error of 3 times of Multipath Errors values.
4. a kind of GNSS multipath effect correcting method based on BP neural network technology according to claim 1, special
Sign is, normalized in the step (2) are as follows: by the data M in Double deference time seriesd1ij、Md2ijBy formula 5) processing,
Simultaneously by the M ' after normalizationd1ij、M′d2ijDouble deference Multipath Errors data save respectively；
Md1ij、Md2ijIt is the Double deference Multipath Errors value of i-th satellite, Md1ijmin、Md2ijminIt is in Double deference Multipath Errors
The minimum value of data, Md1ijmax、Md2ijmaxIt is the maximum value of data in i-th satellite Double deference Multipath Errors, M 'd1ij、
M′d2ijIt is the Double deference Multipath Errors data after normalization.
5. a kind of GNSS multipath effect correcting method based on BP neural network technology according to claim 1, special
Sign is that BP neural network algorithm includes an input layer, a hidden layer and an output layer；Input layer number is
2 times of common number of satellites in base station and monitoring station；Output layer totally three neurons, respectively tri- sides monitoring station X, Y, Z
To coordinate time sequence；BP neural network activation primitive uses Sigmoid function.
6. a kind of GNSS multipath effect correcting method based on BP neural network technology according to claim 1, special
Sign is, further includes that network topology structure determines before the step (4), specifically: according to sample data training as a result,
Root-mean-square error rmse is calculated, different hidden layer neuron numbers, network weight is adjusted, then recalculates rmse, is repeated
After carrying out repeatedly, select the smallest neuron number of rmse and network weight as optimum network topological structure；
Formula 7) in n be prediction points, o be network desired output, p be network reality output.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910254859.8A CN109738926B (en) | 2019-04-01 | 2019-04-01 | A kind of GNSS multipath effect correcting method based on BP neural network technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910254859.8A CN109738926B (en) | 2019-04-01 | 2019-04-01 | A kind of GNSS multipath effect correcting method based on BP neural network technology |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109738926A CN109738926A (en) | 2019-05-10 |
| CN109738926B true CN109738926B (en) | 2019-06-28 |
Family
ID=66371397
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910254859.8A Active CN109738926B (en) | 2019-04-01 | 2019-04-01 | A kind of GNSS multipath effect correcting method based on BP neural network technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109738926B (en) |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111323795B (en) * | 2020-03-20 | 2022-03-22 | 湖南联智科技股份有限公司 | Multi-path error weakening method in Beidou deformation monitoring |
| CN111562599B (en) * | 2020-05-14 | 2021-07-16 | 中国北方工业有限公司 | Method for establishing RTD pseudo-range correction forecasting model of Beidou satellite network |
| CN112099067A (en) * | 2020-08-25 | 2020-12-18 | 中国铁路设计集团有限公司 | Deformation monitoring GNSS multi-path effect correction method based on local mean decomposition filtering |
| CN112068161B (en) * | 2020-09-17 | 2024-03-29 | 国网浙江省电力有限公司检修分公司 | Multipath error reduction method and device |
| CN112926190B (en) * | 2021-01-28 | 2024-04-16 | 东南大学 | Multi-path weakening method and device based on VMD algorithm |
| CN113189624B (en) * | 2021-04-30 | 2023-10-03 | 中山大学 | Self-adaptive classification multipath error extraction method and device |
| CN113609757B (en) * | 2021-07-13 | 2024-03-08 | 中国科学院国家空间科学中心 | GNSS occultation troposphere parameter correction method based on random forest regression |
| CN114488227B (en) * | 2022-01-26 | 2023-10-20 | 西南交通大学 | Multipath error correction method based on spatial correlation |
| CN115877425B (en) * | 2022-12-22 | 2024-09-10 | 国汽大有时空科技(安庆)有限公司 | SSR grid master station determining method and device based on AI model |
| CN115993668B (en) * | 2023-03-22 | 2023-05-30 | 成都云智北斗科技有限公司 | Polynomial correction and neural network-based PWV reconstruction method and system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5726659A (en) * | 1995-09-21 | 1998-03-10 | Stanford University | Multipath calibration in GPS pseudorange measurements |
| CN103430046A (en) * | 2011-03-22 | 2013-12-04 | 天宝导航有限公司 | Gnss signal processing with known position for reconvergence |
| CN106488548A (en) * | 2016-10-25 | 2017-03-08 | 北京邮电大学 | A kind of determination method and device of indoor multipath error |
| CN107703418A (en) * | 2017-08-30 | 2018-02-16 | 上海交通大学 | Shelf depreciation location error compensation method based on more radial base neural nets |
| CN207779468U (en) * | 2018-02-28 | 2018-08-28 | 湖南联智桥隧技术有限公司 | Integral type slope monitoring stake based on Beidou satellite navigation system |
| CN108957502A (en) * | 2018-06-04 | 2018-12-07 | 安徽理工大学 | GNSS multisystem Multipath Errors based on deep learning LSTM weaken method in real time |
-
2019
- 2019-04-01 CN CN201910254859.8A patent/CN109738926B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5726659A (en) * | 1995-09-21 | 1998-03-10 | Stanford University | Multipath calibration in GPS pseudorange measurements |
| CN103430046A (en) * | 2011-03-22 | 2013-12-04 | 天宝导航有限公司 | Gnss signal processing with known position for reconvergence |
| CN106488548A (en) * | 2016-10-25 | 2017-03-08 | 北京邮电大学 | A kind of determination method and device of indoor multipath error |
| CN107703418A (en) * | 2017-08-30 | 2018-02-16 | 上海交通大学 | Shelf depreciation location error compensation method based on more radial base neural nets |
| CN207779468U (en) * | 2018-02-28 | 2018-08-28 | 湖南联智桥隧技术有限公司 | Integral type slope monitoring stake based on Beidou satellite navigation system |
| CN108957502A (en) * | 2018-06-04 | 2018-12-07 | 安徽理工大学 | GNSS multisystem Multipath Errors based on deep learning LSTM weaken method in real time |
Non-Patent Citations (3)
| Title |
|---|
| An Adaptive MultipathMitigation Filter for GNSS Applications;Chung-Liang Chang et al;《EURASIP Journal on Advances in Signal Processing》;20081231;第1-10页 |
| BP 神经网络在差分GPS定位技术中的应用;刘英英;《东北大学学报（自然科学版）》;20081130;第29卷(第11期);第1536-1539页 |
| 基于BP算法的自适应多径消除方法研究;杨鹏;《中国优秀硕士学位论文全文数据库 信息科技辑》;20120515(第5期);第I136-899页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109738926A (en) | 2019-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109738926B (en) | A kind of GNSS multipath effect correcting method based on BP neural network technology | |
| EP3514823A2 (en) | Machine learning-based semiconductor manufacturing yield prediction system and method | |
| CN105654069B (en) | Based on LpThe increment subspace method for tracking target of norm regularization | |
| CN103197335B (en) | Method using improved regularization method to restrain difference global positioning system (DGPS) integer ambiguity ill-condition | |
| CN104392414A (en) | Establishing method of regional CORS coordinate time series noise model | |
| CN112634608B (en) | Regional traffic flow prediction method and system | |
| CN106971189B (en) | A kind of noisy method for recognising star map of low resolution | |
| CN112132365A (en) | Multi-scale feature-based forecast wind speed correction method and system | |
| CN115856963A (en) | High-precision positioning algorithm based on deep neural network learning | |
| Czaplewski et al. | Improvement in accuracy of determining a vessel’s position with the use of neural networks ana robust m-estimation | |
| CN116310852A (en) | Double-time-phase remote sensing image unsupervised classification and change detection method and system | |
| Febvre et al. | Joint calibration and mapping of satellite altimetry data using trainable variational models | |
| Zhou et al. | Calculation of fractal dimension based on artificial neural network and its application for machined surfaces | |
| Gerg et al. | Real-time, deep synthetic aperture sonar (SAS) autofocus | |
| CN104463245B (en) | A kind of target identification method | |
| CN106501828B (en) | A kind of high accuracy pseudo range one-point positioning method based on fuzzy logic weighting | |
| Silva et al. | Deformation fringes detection in SAR interferograms using deep learning | |
| CN109164471B (en) | Region division method based on principal component analysis | |
| CN116026325A (en) | Navigation method and related device based on neural process and Kalman filtering | |
| Tripathi et al. | A survey: optimization algorithms in deep learning | |
| CN111008549A (en) | UUV platform DVL signal distortion reconstruction method based on sample entropy and IFOA-GRNN | |
| Scholz et al. | Sim2real for environmental neural processes | |
| Kellermann et al. | Estimation of unknown system states based on an adaptive neural network and Kalman filter | |
| CN112801955A (en) | Plankton detection method under unbalanced population distribution condition | |
| CN112014324A (en) | Water body nutrition state evaluation method based on multi-source monitoring data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP03 | Change of name, title or address |
Address after: 410200 No. 168, Second Section of Yanhe Road, Wangcheng Economic and Technological Development Zone, Changsha City, Hunan Province Patentee after: Hunan Lianzhi Technology Co.,Ltd. Address before: 410019 Yanhe Road, Wangcheng District, Changsha, Hunan Province, No. 168 Patentee before: HUNAN LIANZHI BRIDGE AND TUNNEL TECHNOLOGY Co.,Ltd. |
|
| CP03 | Change of name, title or address | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: A GNSS multipath effect correction method based on BP neural network technology Effective date of registration: 20220920 Granted publication date: 20190628 Pledgee: China Everbright Bank Co.,Ltd. Changsha Huafeng Sub branch Pledgor: Hunan Lianzhi Technology Co.,Ltd. Registration number: Y2022430000088 |
|
| PE01 | Entry into force of the registration of the contract for pledge of patent right | ||
| PC01 | Cancellation of the registration of the contract for pledge of patent right |
Date of cancellation: 20231106 Granted publication date: 20190628 Pledgee: China Everbright Bank Co.,Ltd. Changsha Huafeng Sub branch Pledgor: Hunan Lianzhi Technology Co.,Ltd. Registration number: Y2022430000088 |
|
| PC01 | Cancellation of the registration of the contract for pledge of patent right |