CN117650861B - Wireless environment prediction method and device, electronic equipment and storage medium - Google Patents
Wireless environment prediction method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN117650861B CN117650861B CN202410106374.5A CN202410106374A CN117650861B CN 117650861 B CN117650861 B CN 117650861B CN 202410106374 A CN202410106374 A CN 202410106374A CN 117650861 B CN117650861 B CN 117650861B
- Authority
- CN
- China
- Prior art keywords
- data
- relation
- reference signal
- power spectrum
- antenna array
- 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
- 238000000034 method Methods 0.000 title claims abstract description 66
- 238000003860 storage Methods 0.000 title claims abstract description 22
- 230000005540 biological transmission Effects 0.000 claims abstract description 78
- 230000004807 localization Effects 0.000 claims abstract description 49
- 238000001228 spectrum Methods 0.000 claims description 134
- 238000010586 diagram Methods 0.000 claims description 86
- 238000013528 artificial neural network Methods 0.000 claims description 52
- 239000013598 vector Substances 0.000 claims description 49
- 238000000547 structure data Methods 0.000 claims description 34
- 238000000605 extraction Methods 0.000 claims description 18
- 238000004590 computer program Methods 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000004891 communication Methods 0.000 abstract description 20
- 238000012549 training Methods 0.000 description 18
- 239000011159 matrix material Substances 0.000 description 17
- 238000005259 measurement Methods 0.000 description 14
- 230000006870 function Effects 0.000 description 11
- 230000008569 process Effects 0.000 description 8
- 230000000694 effects Effects 0.000 description 6
- 238000005096 rolling process Methods 0.000 description 6
- 238000004422 calculation algorithm Methods 0.000 description 5
- 238000009826 distribution Methods 0.000 description 5
- 238000011084 recovery Methods 0.000 description 5
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 101100379079 Emericella variicolor andA gene Proteins 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000013480 data collection Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000005562 fading Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000003213 activating effect Effects 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 2
- 238000004220 aggregation Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000013441 quality evaluation Methods 0.000 description 2
- 230000003595 spectral effect Effects 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 238000007796 conventional method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 239000012633 leachable Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000005055 memory storage Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 238000001303 quality assessment method Methods 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000007493 shaping process Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002195 synergetic effect Effects 0.000 description 1
- 238000009827 uniform distribution Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/30—Monitoring; Testing of propagation channels
- H04B17/373—Predicting channel quality or other radio frequency [RF] parameters
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/042—Knowledge-based neural networks; Logical representations of neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/30—Monitoring; Testing of propagation channels
- H04B17/309—Measuring or estimating channel quality parameters
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/30—Monitoring; Testing of propagation channels
- H04B17/391—Modelling the propagation channel
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Electromagnetism (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Quality & Reliability (AREA)
- Probability & Statistics with Applications (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Algebra (AREA)
- Mathematical Optimization (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The application provides a wireless environment prediction method and device, electronic equipment and a storage medium, which belong to the technical field of communication, and achieve the aim of wireless environment prediction by inputting an antenna array transmission coefficient and average reference signal receiving power into a multi-grid localization statistical channel model.
Description
Technical Field
The present application relates to the field of communications technologies, and in particular, to a method and apparatus for predicting a wireless environment, an electronic device, and a storage medium.
Background
In the related art, an angular Power spectrum (Angular Power Spectrum, APS) is estimated from a reference signal received Power (REFERENCE SIGNAL RECEIVING Power, RSRP) by a Localized STATISTICAL CHANNEL Model (LSCM), and the quality of a wireless channel is estimated using the angular Power spectrum. Due to the influence of high dryness of the measurement matrix, the angle power spectrum predicted by the localization statistical channel model is inaccurate, so that the wireless environment prediction result is inaccurate. How to improve the accuracy of wireless channel quality assessment becomes a problem to be solved.
Disclosure of Invention
The embodiment of the application mainly aims to provide a wireless environment prediction method and device, electronic equipment and a storage medium, and aims to improve the accuracy of wireless environment prediction.
To achieve the above object, a first aspect of an embodiment of the present application provides a wireless environment prediction method, including:
acquiring an antenna array transmission coefficient of a base station;
acquiring average reference signal receiving power of a grid area in a preset period;
Inputting the antenna array transmission coefficients and the average reference signal received power into a multi-grid localization statistical channel model, wherein the multi-grid localization statistical channel model comprises a plurality of super-network layers, and the super-network layers comprise a Markov diagram neural network and a variational Bayesian inference module;
Carrying out feature extraction on the preset relationship diagram structural data through the Markov diagram neural network to obtain association relationship features; the preset relation diagram structure data is a heterogeneous Markov diagram representation, and is used for representing the association relation between the reference signal received power and the angle power spectrum of the grid region;
Predicting an angular power spectrum of the grid region by the variational Bayesian inference module based on the association relation features, the average reference signal received power and the antenna array transmission coefficient;
And predicting the wireless environment between the base station and the grid area according to the angle power spectrum.
In some embodiments, the markov neural network includes a first graph convolution network and a second graph convolution network, and the feature extraction is performed on the preset relationship graph structure data through the markov neural network to obtain the association relationship feature, which includes:
extracting first relation features from the preset relation graph structure data through the first graph convolution network to obtain first relation data; the preset relation diagram structure data comprise angle power spectrum priori data and angle power spectrum likelihood data, and the first relation data are used for representing the relation between the angle power spectrum priori data and the angle power spectrum likelihood data;
Extracting second relation features from the preset relation graph structural data through the second graph convolution network to obtain second relation data; the second relation data is used for representing the relation between the angle power spectrum likelihood data and the angle power spectrum priori data;
and obtaining the association relation characteristic according to the first relation data and the second relation data.
In some embodiments, the correlation feature includes first and second correlation data, and the predicting, by the variational bayesian inference module, the angular power spectrum of the grid region based on the correlation feature, the average reference signal received power, and the antenna array transmission coefficient includes:
updating the first relation data through the variation Bayesian inference module of the current super network layer based on the association relation characteristics, the average reference signal receiving power and the antenna array transmission coefficient to obtain first candidate relation data;
And predicting the angular power spectrum of the grid region by the variational Bayesian inference module of the next super network layer based on the first candidate relation data, the second relation data, the average reference signal received power and the antenna array transmission coefficient.
In some embodiments, the predicting, by the variational bayesian inference module of the next super network layer, the angular power spectrum of the grid region based on the first candidate relationship data, the second relationship data, the average reference signal received power and the antenna array transmission coefficients comprises:
updating the first candidate relationship data through the Markov diagram neural network of the next super network layer based on the second relationship data to obtain second candidate relationship data;
Updating the second relation data through the Markov diagram neural network of the next super network layer based on the first candidate relation data to obtain third candidate relation data;
And updating the second candidate relation data through the variation Bayesian inference module of the next super network layer based on the second candidate relation data, the third candidate relation data, the average reference signal receiving power and the antenna array transmission coefficient until the final candidate relation data output by the variation Bayesian inference module of the last super network layer is obtained, and taking the final candidate relation data as the angle power spectrum.
In some embodiments, the preset period includes a plurality of time steps, and the acquiring the average reference signal received power of the grid region in the preset period includes:
Acquiring the power of the base station for transmitting a reference signal to the grid area according to a preset beam forming vector, and acquiring the transmitting power;
acquiring a channel vector from the base station to the grid area in the time step;
determining the receiving power of the reference signal received by the grid area in the time step according to the transmitting power, the channel vector and the preset beam forming vector;
and determining the average reference signal receiving power according to the receiving power of all the time steps.
In some embodiments, the acquiring the antenna array transmission coefficients of the base station includes:
acquiring an antenna array guide vector of the base station;
and determining the antenna array transmission coefficient according to the antenna array steering vector and a preset beam forming vector.
In some embodiments, the multi-grid localization statistical channel model is trained according to the following steps:
Acquiring a sample antenna array transmission coefficient of the base station;
Acquiring sample average reference signal receiving power of the grid area in a preset period;
Predicting a target angle power spectrum of the grid region through a preset model based on the sample antenna array transmission coefficient and the sample average reference signal receiving power;
Performing loss calculation according to the target angle power spectrum, the sample antenna array transmission coefficient and the sample average reference signal receiving power to obtain unsupervised loss data;
and adjusting network parameters of the preset model according to the unsupervised loss data to obtain the multi-grid localization statistical channel model.
To achieve the above object, a second aspect of the embodiments of the present application proposes a wireless environment prediction apparatus including:
the first acquisition module is used for acquiring the antenna array transmission coefficient of the base station;
the second acquisition module is used for acquiring the average reference signal receiving power of the grid area in a preset period;
The input module is used for inputting the antenna array transmission coefficient and the average reference signal receiving power into a multi-grid localization statistical channel model, wherein the multi-grid localization statistical channel model comprises a plurality of super network layers, and the super network layers comprise a Markov diagram neural network and a variational Bayesian inference module;
the feature extraction module is used for carrying out feature extraction on the preset relationship diagram structure data through the Markov diagram neural network to obtain association relationship features; the preset relation diagram structure data is a heterogeneous Markov diagram representation, and is used for representing the association relation between the reference signal received power and the angle power spectrum of the grid region;
The prediction module is used for predicting the angular power spectrum of the grid region through the variation Bayesian inference module based on the association relation characteristic, the average reference signal received power and the antenna array transmission coefficient;
and the quality evaluation module is used for predicting the wireless environment between the base station and the grid area according to the angle power spectrum.
To achieve the above object, a third aspect of the embodiments of the present application provides an electronic device, which includes a memory and a processor, where the memory stores a computer program, and the processor implements the wireless environment prediction method described in the first aspect when executing the computer program.
To achieve the above object, a fourth aspect of the embodiments of the present application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the wireless environment prediction method described in the first aspect.
The wireless environment prediction, the wireless environment prediction device, the electronic equipment and the computer readable storage medium provided by the application are used for obtaining the average reference signal received power of the grid area in a preset period by obtaining the antenna array transmission coefficient of the base station, and inputting the antenna array transmission coefficient and the average reference signal received power into a multi-grid localization statistical channel model so that the multi-grid localization statistical channel model predicts an angle power spectrum according to the antenna array transmission coefficient and the average reference signal received power. The multi-grid localization statistical channel model includes a plurality of super-network layers including a Markov-map neural network and a variational Bayesian inference module. The preset relation diagram structure data is represented by a heterogeneous Markov diagram, a complex relation between the reference signal received power and the angle power spectrum of the grid region is contained, the characteristic extraction is carried out on the preset relation diagram structure data through a Markov diagram neural network, prior information is obtained from the diagram structure, so that the influence of high dryness of the measurement matrix is relieved, and the association relation characteristic is obtained. Based on the association relation characteristics, the average reference signal receiving power and the antenna array transmission coefficient, the angular power spectrum of the grid area is predicted by the variational Bayesian inference module, the problem that the variational Bayesian inference module is sensitive to the super-parameters of the probability map model is solved by introducing prior information of the inherent map structure, and the accuracy of the angular power spectrum prediction is improved. And the wireless environment between the base station and the grid area is predicted according to the angle power spectrum, so that the accuracy of the wireless environment prediction is improved.
Drawings
Fig. 1 is a flowchart of a wireless environment prediction method provided by an embodiment of the present application;
fig. 2 is a flowchart of step S110 in fig. 1;
fig. 3 is a flowchart of step S120 in fig. 1;
FIG. 4 is a schematic diagram of a multi-grid localization statistical channel model provided by an embodiment of the present application;
FIG. 5 is a flow chart of a training multi-grid localization statistical channel model provided by an embodiment of the present application;
FIG. 6 is a schematic representation of a heterogeneous Markov diagram representation provided by an embodiment of the present application;
FIG. 7 is another schematic representation of a heterogeneous Markov diagram representation provided by an embodiment of the present application;
fig. 8 is a flowchart of step S140 in fig. 1;
Fig. 9 is a flowchart of step S150 in fig. 1;
fig. 10 is a flowchart of step S920 in fig. 9;
fig. 11a is an effect diagram of a wireless environment prediction method according to an embodiment of the present application;
FIG. 11b is another effect diagram of a wireless environment prediction method according to an embodiment of the present application;
FIG. 11c is another effect diagram of a wireless environment prediction method according to an embodiment of the present application;
Fig. 12 is a schematic structural diagram of a wireless environment prediction apparatus according to an embodiment of the present application;
fig. 13 is a schematic diagram of a hardware structure of an electronic device according to an embodiment of the present application.
Detailed Description
The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.
It should be noted that although functional block division is performed in a device diagram and a logic sequence is shown in a flowchart, in some cases, the steps shown or described may be performed in a different order than the block division in the device, or in the flowchart. The terms first, second and the like in the description and in the claims and in the above-described figures, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of the application only and is not intended to be limiting of the application.
The localization statistical channel model plays a key role in wireless network optimization and can be used for providing statistical information about wireless environment, such as first-order angle power spectrum statistical characteristics and second-order angle power spectrum statistical characteristics. The localization statistical channel model can receive power calculation angle power spectrum from beam type reference signals so as to predict data transmission performance of the wireless network under different network parameters such as array azimuth, antenna inclination angle and the like. In the related art, an angular power spectrum is estimated from a reference signal received power by a localization statistical channel model, and the quality of a wireless channel is estimated using the angular power spectrum. Due to the influence of high dryness of the measurement matrix, the angle power spectrum predicted by the localization statistical channel model is inaccurate, so that the wireless environment prediction result is inaccurate. How to improve the accuracy of wireless environment prediction becomes a problem to be solved.
Based on the above, the embodiment of the application provides a wireless environment prediction method, a wireless channel quality evaluation device, electronic equipment and a computer readable storage medium, which improve the accuracy of wireless environment prediction.
The wireless environment prediction method, the wireless environment prediction device, the electronic device and the computer readable storage medium provided by the embodiments of the present application are specifically described by the following embodiments, and the wireless environment prediction method in the embodiments of the present application is first described.
The embodiment of the application provides a wireless environment prediction method, which relates to the technical field of communication. The wireless environment prediction method provided by the embodiment of the application can be applied to a terminal, a server and software running in the terminal or the server. In some embodiments, the terminal may be a smart phone, tablet, notebook, desktop, etc.; the server side can be configured as an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, and a cloud server for providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDNs, basic cloud computing services such as big data and artificial intelligent platforms and the like; the software may be an application or the like that implements the wireless environment prediction method, but is not limited to the above form.
The application is operational with numerous general purpose or special purpose computer system environments or configurations. For example: personal computers, server computers, hand-held or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, and the like. The application may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The application may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.
Consider a downlink multiple-input multiple-output (MultipleInputMultipleOutput, MIMO) radio system that includes a Base Station (BS) equipped with a set of radio channels includingAn antenna array of individual antennas.,ForThe number of antennas in the axial direction, representing the length of the antenna array,ForThe number of antennas in the axial direction indicates the width of the antenna array. The coverage area of a base station is divided into L grids. Base station uses codebookThe codebook includes M beams for transmitting reference signals to assist the receiving end in decoding and recovering the reference signals. /(I)ForBeamforming vectors for the individual beams. The base station scans the whole angle space and adjusts the angle and direction of the antennas in the antenna array so as to meet the communication requirements of different grid users. FirstSingle antenna User Equipment (UE) of the grid at theAnd measuring the channel impulse response of the reference signal at each moment to obtain the beam-type reference signal receiving power. The beam-forming reference signal received power includes channel information such as an angle of incidence (Angle of Departure, aoD) of the base station to the user equipment, a channel gain (CHANNEL GAIN), and the like. Channel gain, i.e., channel coefficient, is used to describe the attenuation and fading characteristics of a wireless channel. And the user equipment reports the beam type reference signal receiving power to the base station for the localization statistical channel modeling task to obtain a multi-grid localization statistical channel model. The multi-grid localization statistical channel model predicts an angle power spectrum from the beam type reference signal received power, predicts the wireless environment according to the angle power spectrum, can be applied to localization statistical channel modeling communication scenes, and falls to a network layer data processing module of a base station.
Fig. 1 is an optional flowchart of a method for predicting a wireless environment according to an embodiment of the present application, where the method in fig. 1 may include, but is not limited to, steps S110 to S160.
Step S110, obtaining an antenna array transmission coefficient of a base station;
step S120, obtaining average reference signal receiving power of a grid area in a preset period;
Step S130, inputting the antenna array transmission coefficient and the average reference signal receiving power into a multi-grid localization statistical channel model, wherein the multi-grid localization statistical channel model comprises a plurality of super-network layers, and the super-network layers comprise a Markov diagram neural network and a variation Bayesian inference module;
Step S140, extracting features of the preset relationship diagram structural data through a Markov diagram neural network to obtain association relationship features; the preset relation diagram structure data is a heterogeneous Markov diagram representation, and is used for representing the association relation between the reference signal received power and the angle power spectrum of the grid region;
Step S150, based on the association relation characteristics, the average reference signal received power and the antenna array transmission coefficient, predicting an angle power spectrum of the grid area through a variable dB leaf inference module;
Step S160, the wireless environment between the base station and the grid area is predicted according to the angle power spectrum.
Referring to fig. 2, in some embodiments, step S110 may include, but is not limited to, steps S210 to S220:
Step S210, an antenna array guide vector of a base station is obtained;
step S220, determining the antenna array transmission coefficient according to the antenna array steering vector and the preset beam forming vector.
In step S210 of some embodiments, the angle space is used to define a direction range of the base station antenna, divide a three-dimensional angle space of the base station, and uniformly discretize the free space into NV pitch angles and NH azimuth angles, so as to obtain N space angles.N is the number of space angles, NV is the number of pitch angles, and NH is the number of azimuth angles. FirstPitch angle of incidence (tilt AoD) of individual pitch angles is,. FirstAzimuth incidence angle (azimuth AoD) of azimuth,. It should be noted that the number of real propagation paths from the base station to the user equipment is less than N. The antenna array steering vector is used to describe the radiation or reception capabilities of the antenna array in different directions, including a plurality of differently oriented array steering vectors. The antenna array steering vector is a steering matrix, denoted S, thThe array steering vector for each direction (spatial angle) is expressed asDefinition. The array steering vector is calculated as shown in formula (1).
Formula (1)
Wherein,Representing the/>, in the x-axis directionThe array steering vectors for the individual spatial angles,,For the spacing distance of adjacent antennas in the x-axis direction,Is the wavelength of the reference signal; /(I)Representing the/>, in the y-axis directionArray steering vector of individual spatial angles,,Is the spacing distance of adjacent antennas in the y-axis direction; /(I)Represents the Kronecker (Kronecker) product of the matrix.
In step S220 of some embodiments, the preset beamforming vector includes beamforming vectors of M beams, the preset beamforming vector being expressed as,Represents theBeamforming vectors for the individual beams. And calculating a measurement matrix according to the antenna array steering vector S and the preset beam forming vector W, and taking the calculated measurement matrix as an antenna array transmission coefficient. The calculation method of the antenna array transmission coefficient is shown in formula (2).
Formula (2)
Wherein,Transmitting coefficients for the antenna array; h represents a conjugate transpose operation; /(I)The absolute value of the element is taken and squared.
It should be noted that, the measurement matrix has high coherence, which means that the measurement matrix has non-uniform columns and dense parallel columns. Non-uniform columns make small amplitude columns easier to ignore. Furthermore, dense parallel columns are also difficult to distinguish in this measurement matrix. Since the angular power spectrum is naturally sparse, the localization statistical channel modeling task can be constructed as a sparse recovery problem. However, the presence of high coherence of the measurement matrix can present challenges to the sparse recovery problem. Such a measurement matrix may lead to a higher probability of error in modeling the real channel path, failing to accurately model the channels of the wireless system, resulting in an off-line network optimization in downstream tasks failing to effectively improve the performance of the actual network, such as coverage and spectral efficiency. Sparsity refers to the fact that in the angular power spectrum to be recovered, most elements are zero or close to zero, while only a few have non-zero values.
Through the above steps S210 to S220, an antenna array transmission coefficient can be obtained to determine an angular power spectrum based on the antenna array transmission coefficient.
Referring to fig. 3, in some embodiments, the preset time period includes a plurality of time steps, and step S120 may include, but is not limited to, steps S310 to S340:
step S310, obtaining the power of a base station transmitting a reference signal to a grid area according to a preset beam forming vector, and obtaining the transmitting power;
Step S320, obtaining channel vectors from the base station to the grid area in a time step;
step S330, determining the receiving power of the reference signal received in the time step grid area according to the transmitting power, the channel vector and the preset beam forming vector;
Step S340, determining the average reference signal received power according to the received power of all time steps.
In step S310 of some embodiments, the transmit power of the base station is determined such that the base station is at the first transmit powerThe mth time step passes through the mth beamforming vectorToThe grid regions transmit reference signals.
In step S320 of some embodiments, the acquisition is at the firstTime step base station toChannel vector of individual grid regions. The channel vector includes information such as channel gain, multipath fading, noise, interference, etc. Multipath fading includes information such as delay differences, phase differences, etc. The noise may be thermal noise, interference noise. The interference is an interfering signal of other wireless devices. Channel vectorThe calculation method of (2) is shown in the formula (3).
Formula (3)
Wherein N is the number of space angles; represents the/> A spatial angle; /(I)ForArray steering vectors for the individual spatial angles; /(I)Represents theUnder the personal space angleThe grid area is at theComplex channel gains for individual time steps; h denotes a conjugate transpose operation.
In step S330 of some embodiments, the firstThe grid areas receive the/>, from the base stationReference signal of each time step, and measuring the received power of the reference signal. Specifically, the grid region reception/>, is determined from the transmit power, the channel vector, and the mth beamforming vectorThe mth beamforming vector is used to transmit the received power of the reference signal for a time step. The calculation method of the received power is shown in formula (4).
Formula (4)
Wherein,Represents theEach beam; /(I)Represents theA plurality of grid regions; /(I)Represents theA plurality of time steps; /(I)Representing the received power; p represents the transmit power; /(I)Representing a channel vector; /(I)ForAnd a beamforming vector.
The base station uses all beamforming vectors in the codebook at time step t toA grid region transmitting a reference signal, the grid region receiving theThe received power of the individual time step reference signal is,Where M is the number of beam shaping vectors in the codebook.
In step S340 of some embodiments, an average value of the received power of the grid area in all time steps is calculated, so as to obtain an average reference signal received power of the grid area in a preset period. The calculation method of the average reference signal received power is shown in formula (5).
Formula (5)
Wherein,Represents theAverage reference signal received power of the grid areas in a preset period; t represents the number of time steps in a preset period; /(I)Represents theThe grid area is at theAnd receiving power for each time step.
Through the above steps S310 to S340, an average reference signal received power can be obtained to estimate an angular power spectrum based on the antenna array transmission coefficient and the average reference signal received power.
In order not to lose generality, assume complex channel gainsAt a differentThe potential paths are mutually independent, and the phase is obeyed atUniform distribution within a range, thus defining. Due to the static environment, whenAngular power spectrumRemains unchanged and due to the limited number of scatterers and environmental stability, angular power spectrumIs typically sparse. The average reference signal received power, antenna array transmission coefficient and angular power spectrum can be obtained to satisfy the following relationship:
Formula (6)
Wherein,Representing an average value for T consecutive time steps, as desired; /(I)Represents theThe expected average reference signal received power of each grid area in a preset period T; /(I)Representing the antenna array transmission coefficients; /(I)Represents theThe angular power spectrum of the individual grid regions is a desire for complex channel gains in a number of different directions.
It should be noted that, the scatterer refers to an object or medium encountered by the wireless signal during the propagation process, and these objects or mediums can scatter the energy of the signal. These objects are typically randomly distributed and may include buildings, trees, floors, obstructions, and the like. An environment generally refers to a location or geographic area where a wireless signal propagates. The environment may be various geographic and topographical conditions in cities, rural areas, indoor, outdoor, mountainous areas, oceans, and the like.
Due to the noise present in the radio channel, it can be derived from equation (5) and equation (6):
Formula (7)
Wherein,ForNoise present in the individual grid areas.
Determining that the localized statistical channel modeling task is a slave according to equation (7)AndIn co-estimation of angular power spectra/>, of a plurality of gridsI.e. the angular power spectrum of the different grid areas is estimated from the antenna array transmission coefficients and the average reference signal received power.
In step S130 of some embodiments, the antenna array transmission coefficients and the average reference signal received power are input to a multi-grid localization statistical channel model, and an angular power spectrum of the plurality of grid regions is predicted based on the multi-grid localization statistical channel model. The network structure of the multi-grid localization statistical channel model is shown in fig. 4. The multi-grid localization statistical channel model includes a plurality of super-network layers, each including a markov-map neural network (Markov Graph Neural Network, MGNN) and a variational bayesian inference module (Variational Bayesian Inference, VBI).
In fig. 4, the multi-grid localization statistical channel model includes j+1 super network layers, super layer 1, super layer 2, super layer J, and output layer, respectively. Transmitting coefficients of antenna arrayAnd average reference signal received powerInputting the multi-grid localization statistical channel model, and enabling the variational Bayesian inference module of the output layer to output an angle power spectrum/>, through the synergistic effect between the Markov diagram neural network and the variational Bayesian inference module in each super network layer。
In a multi-grid localized statistical channel modeling scenario, reference signal received power is collected from different grids within a particular site. In this case, the spatial consistency nature of the wireless channels means that there is structural sparsity between the angular power spectra of adjacent grids, which helps alleviate the high coherence problem of the measurement matrix. However, conventional methods based on LASSO regression models have difficulties in effectively modeling structural sparsity. To address this challenge, methods based on variational Bayesian inference have been proposed to address such sparse recovery problems by introducing a hierarchical sparse prior. However, existing methods based on variational Bayesian inference are still sensitive to the hyper-parameters defined in the probability map model. Furthermore, these methods, such as the Turbo-VBI algorithm, have long execution times due to the use of a double iterative loop, which is inefficient for the task of multi-grid localization statistical channel modeling.
In the related art, in order to solve the problem of recovering sparse signals in wireless communication, a fixed point network for sparse channel estimation is introduced. For example, a noise reduction approximation information transfer (AMP) network based on learning, a multi-measurement vector learning approximation information transfer (AMP) network. However, these methods typically rely on offline training and extensive tag data collection, which in a multi-grid localization statistical channel modeling scenario can suffer from the fact that ground truth angle power spectrum data is not available, thus presenting significant limitations. To address this problem, online training strategies for sparse channel tracking have been introduced. However, these methods typically require a large amount of training data to achieve good predictive performance. The reason for this is that these methods only use existing neural network models and do not take full advantage of prior information obtained from the inherent graph structure in known physical mechanisms and sparse signal recovery problems.
In order to solve the defects that the traditional method based on the variational Bayesian inference is sensitive to super-parameters, the existing neural network for solving the sparse signal recovery problem in wireless communication depends on offline training and requires detailed label data collection or a large number of training data sets, the embodiment of the application combines the graph neural network based on the heterogeneous Markov graph with a variational Bayesian inference model to carry out grid localization statistical channel modeling. The embodiment of the application solves the problem that the variational Bayesian inference is sensitive to the super-parameters through the graph neural network based on the heterogeneous Markov graph, and simultaneously fully utilizes the inherent graph structure in the variational Bayesian inference algorithm to acquire prior information from the graph structure. Compared with the prior art, the method has the advantages that the requirement for a large amount of training data is remarkably reduced, and the efficiency, the effectiveness and the adaptability of the model are improved.
According to the method, the non-supervision training method is adopted to train the multi-grid localization statistical channel model, and label data collection is not needed, so that an efficient online training process is promoted.
Referring to fig. 5, in some embodiments, the training process of the multi-grid localization statistical channel model may include, but is not limited to, steps S510 to S550:
Step S510, obtaining a sample antenna array transmission coefficient of a base station;
step S520, obtaining the sample average reference signal receiving power of the grid area in a preset period;
step S530, predicting a target angle power spectrum of the grid area through a preset model based on the sample antenna array transmission coefficient and the sample average reference signal receiving power;
step S540, loss calculation is carried out according to the target angle power spectrum, the sample antenna array transmission coefficient and the sample average reference signal receiving power, and unsupervised loss data is obtained;
Step S550, network parameters of the preset model are adjusted according to the unsupervised loss data, and the multi-grid localization statistical channel model is obtained.
In step S510 of some embodiments, sample antenna array transmission coefficients of the base station are obtained from the training data setThe method for calculating the transmission coefficient of the sample antenna array is the same as the transmission coefficient of the antenna array, and refer to step S210 to step S220, which are not described herein again.
In step S520 of some embodiments, a sample average reference signal received power for each grid region over a preset period is obtained from the training data set. The method for calculating the average reference signal received power of the samples is the same as the average reference signal received power, and refer to step S310 to step S340, which are not repeated here.
In step S530 of some embodiments, the sample antenna array transmission coefficient and the sample average reference signal received power are input to a preset model, and a target angular power spectrum of each of the plurality of grid regions is predicted and output by the preset model. If the preset model is MVBGNN, then) Wherein the network structure of the preset model is shown in figure 4Representing the sample antenna array transmission coefficient,Representing the sample average reference signal received power. It can be appreciated that/>, the target angle power spectrumThe positions of non-zero elements in the elements indicate that paths from the base station antennas corresponding to the dividing grids and the positions represented by the corresponding dividing free space to the user equipment exist, so that statistical channel modeling is realized.
It should be noted that MVB-GNN is a markov-varying db-law inference architecture enhanced by a graph neural network, whose forward data stream originates from the messaging process of the db-law inference method, which architecture can outperform the traditional db-law inference method by automatically learning structural sparseness and ensuring rapid implementation.
It should be further noted that, using the reference signal received power collected by the random time T-stream, the reference signal received power can be obtained from the reconstructionTo obtain the desired reference signal received power.
In step S540 of some embodiments, considering that there is a limit to the acquisition of tagged data, embodiments of the present application design an unsupervised loss function to avoid the need for tagged data. The unsupervised loss function is defined as shown in equation (8).
Wherein LOSS represents an unsupervised LOSS function; Is a set of learnable weights (weights) and biases (bias); /(I) 、Respectively expressNorms,A norm; /(I)Representing a sample antenna array transmission coefficient; /(I)Representing the sample average reference signal received power; /(I)Is a super parameter for passingNorms to forceThinning.
Substituting the target angle power spectrum, the sample antenna array transmission coefficient and the sample average reference signal receiving power into an unsupervised loss function shown in a formula (8) to perform loss calculation, and obtaining unsupervised loss data. The training loss shown in equation (8) contains only one data pointThe angular power spectrum requirement on a plurality of real grids is eliminated, and the neural enhancement variable decibel leaf-based inferred graph model (MVB-GNN) can be subjected to online self-adaption.
In step S550 of some embodiments, network parameters of the preset model are updated online according to the unsupervised loss data by minimizing the unsupervised loss dataAnd obtaining the multi-grid localization statistical channel model. The network parameters are learnable weights and biases. The multi-grid localization statistics channel model has the capability of automatically learning structural sparsity, utilizing graph structures and promoting online training, and shows the versatility and wide applicability in solving practical challenges and scenes.
The above steps S510 to S550 avoid the need for the labeled data by an unsupervised online training manner, and improve the feasibility of the model in practical application.
The traditional variational Bayesian inference method obtains angular power spectra of multiple grids through likelihood probability (likelihood) and prior probability (priority)Maximum a posteriori probability (Maximum A Posteriori, MAP) estimates of (a) as shown in equation (9). /(I)
Formula (9)
Wherein,Is the maximum posterior probability; /(I)Is a likelihood probability; /(I)Is a priori probability.
To achieve efficient inference, hidden variables (latent variables) are introduced into the likelihood probability portion and the prior probability portion. The hidden variable of the likelihood probability part is the precision of the noise, expressed as. To describeThe structural sparsity in (a) introduces precision and support (support) as hidden variables in the early probability part, and the precision is expressed asSupport is expressed as. Based on the probability model, classical variational Bayesian inference updates approximate conditional marginal posterior distribution/>, in an alternating fashion、、AndTo obtain angular power spectra/>, of a plurality of gridsIs used for the maximum a posteriori probability estimation of (c).
Based on the spatial consistency attribute of the wireless channels, the similar structure between adjacent grid channels is utilized, the similar structure can lead to structural sparsity, better channel modeling performance can be realized, and the problem brought by high coherence is relieved. In the variant bayesian inference however,Is unable to capture the characteristics of the angular power spectral structure sparse mode from adjacent grids. To address this problem, embodiments of the present application employ a priori design of a heterogeneous Markov diagram representation in the form of a Markov chain and construct a Markov diagram neural network based on the heterogeneous Markov diagram representation.
And extracting features of the preset relationship diagram structural data through a Markov diagram neural network to obtain association relationship features. The preset relation diagram structure data is represented by a heterogeneous Markov diagram, the diagram contains a complex relation between the reference signal received power and the angular power spectrum with sparse multi-grid structure, and the preset relation diagram structure data is used for representing the association relation between the reference signal received power and the angular power spectrum of the grid region. The markov diagram representation is shown in fig. 6. Consider a Markov chain with two states, introducing two heterogeneous vertex setsAnd。Prior probability including L grid regions,Likelihood probability comprising L grid regions. To ensure structural sparsity of the angular power spectrum between adjacent grid regions,Edges between the middle vertices are symmetrical in both directions so that sub-graph a has bi-directional edges. /(I)AndEdges between vertices should not ignore the potential dependencies of accuracy and support between different grid areas, so a fully connected heterogeneous bipartite graph is used to represent the relationship between likelihood and prior, as shown in sub-graph B. In addition, in order to make GNN have better generalization performance, the two types of vertices of each grid region are decomposed into N types of vertices, respectively, resulting in a heterogeneous markov diagram representation as shown in fig. 7. FirstVertices of the grid areas are decomposed intoAnd. Here, two sets of vertices followWhereinRepresenting the number of vertices of subgraph A,The number of vertexes of the sub-graph B is represented, L is the number of grid areas, and N is the number of vertexes which need to be decomposed for each vertex. /(I)、、Representing sets of edges connecting sub-graph a to sub-graph a, sub-graph a to sub-graph B, and sub-graph B to sub-graph a, respectively. The set of edges satisfies,。Representing the number of edges connecting adjacent vertices in sub-graph A,Representing the number of edges connecting vertices in sub-graph A with vertices in sub-graph B,Representing the number of edges connecting the vertices in sub-graph B with the vertices in sub-graph a. For the/>, in sub-graph AVerticesExpressed byA collection of neighbor vertices to which an edge in (a) is connected,Expressed byA collection of neighbor vertices to which an edge of (a) is connected. For the/>, in sub-graph BVerticesExpressed byA set of neighbors to which it is connected. /(I). It will be appreciated that ifOrI.e. the first or last grid for which the grid area is the firstVerticesOtherwise。
Based on the heterogeneous Markov diagram representation, the vertex and the introduction are carried outAndRelated eigenvectorsAnd,. WillAssigned asWillAssigned as。AndMay be obtained by staggering the use of a markov neural network and a variational bayesian inference module. To handle aggregation (aggregation) of three heterogeneous edge sets and feature updates of two heterogeneous vertex sets in fig. 7, two different graph volume integrating children, a first graph volume integrating network and a second graph volume integrating network, respectively, are added in each layer of the markov graph neural network to effectively capture relevant information. The following describes a process of extracting association characteristics of the preset relationship diagram structural data by using the markov diagram neural network.
Referring to FIG. 8, in some embodiments, a Markov neural network includes K network layers at which to introduce、AndThe three different learnable weights and biases, each network layer including a first graph roll-up network and a second graph roll-up network, step S140 may include, but is not limited to including steps S810 through S830:
Step S810, extracting first relation features from preset relation diagram structure data through a first diagram convolution network to obtain first relation data; the preset relation diagram structure data comprise angle power spectrum priori data and angle power spectrum likelihood data, and the first relation data are used for representing the relation between the angle power spectrum priori data and the angle power spectrum likelihood data;
step S820, extracting second relation features from the preset relation diagram structure data through a second diagram convolution network to obtain second relation data; the second relation data is used for representing the relation between the angle power spectrum likelihood data and the angle power spectrum priori data;
Step S830, obtaining the association relation feature according to the first relation data and the second relation data.
In step S810 of some embodiments, the preset relationship graph structure data includes angular power spectrum prior data and angular power spectrum likelihood data, the angular power spectrum prior data representing the vertex setAngular power spectrum likelihood data represents vertex set. The first relation data is used for representing angle power spectrum priori dataRelationship between the power spectrum of angle priori dataAnd angular power spectrum likelihood dataIs a relationship of (3). The process of extracting the first relation characteristic from the preset relation diagram structure data through the first diagram convolution network is as follows: initializing the/>, in the preset relationship graph structure data, for each vertexGrid area numberThe accuracy of each vertex, supporting a probability equal to 1, yieldsAndWherein. Two groups of vertexes/>, are introduced through preset relational graph structural dataAndRelated eigenvectorsAndWillAssigned toWill beAssigned to. The identification of the K network layers is 0, 1, … and K-1 in sequence. WillLayer 0 assigned,Layer 0 assigned. For the k-th layer, the first graph rolling network performs the first relation feature extraction in the manner shown in the formula (10).
Formula (10)
Wherein,ForLayer output of the first layerFirst relationship data for the vertices; /(I)Activating a function for a ReLU;、 For/> Bias of first graph rolling network of layer,Representing the offset connecting subgraph A to the edge of subgraph A,Representing a bias of edges connecting sub-graph a to sub-graph B; /(I)、ForWeights of the first graph of the layers, volumes network,Weights representing edges connecting subgraph A to subgraph A,Representing weights of edges connecting sub-graph a to sub-graph B; /(I)Expressed byEdges andA set of neighbor vertices connected by vertices; /(I)Expressed byEdges andA set of neighbor vertices connected by vertices; /(I)Represents theVertices andThe product of the square roots of the degrees of the vertices; /(I)For the last network layer, i.e.Layer output of the first layerFirst relationship data for the vertices; /(I)For the last network layer, i.e. the firstLayer output of the first layerSecond relationship data for the vertices.
In step S820 of some embodiments, the second relationship data is used to characterize the angular power spectrum likelihood dataWith angle power spectrum prior dataIs a relationship of (3). For the k layer, the second graph rolling network performs the second relation feature extraction on the preset relation graph structure data in the manner shown in the formula (11).
Wherein,ForLayer output of the first layerSecond relationship data for the vertices; /(I)Activating a function for a ReLU; For/> Bias of second graph rolling network of layer,Representing a bias connecting sub-graph B to the edge of sub-graph a; /(I)ForThe second graph of the layers rolls the weights of the network; /(I)Expressed byEdges andA set of neighbor vertices connected by vertices; /(I)Represents theVertices andThe product of the square roots of the degrees of the vertices; /(I)For the last network layer, i.e. the firstLayer output of the first layerFirst relationship data for the vertices.
In step S830 of some embodiments, the first relationship data and the second relationship data output by the kth layer are input to the first graph neural network of the kth+1 layer, to obtain the first relationship data output by the kth+1 layer. And inputting the first relation data output by the k layer into a second graph neural network of the k+1 layer to obtain second relation data output by the k+1 layer. And by analogy, the first relation data and the second relation data output by the last layer are used as association relation features. Specifically, the first and second relationship data output by the last layer are respectively expressed as、WillAssigning intermediate parametersWillAssigned to intermediate parametersBy means of intermediate parametersUpdate initialized probabilityBy means of intermediate parametersUpdate initialized probabilityAccording to updatedAndDeterminationAndAnd obtaining the association relation characteristic.
If the Markov neural network comprises 2 network layers, three different learnable weights and biases set in the network are、And,. Layer 0 and layer 1 leachable weights、Layer 0 and layer 1 bias、. ForThe nth vertex of the grid area is marked as/>, at the node of the whole preset relation diagram structural data. Initializing preset relation diagram structure data to obtainAnd. WillAssigned toWillAssigned to. Based on the formula (10), carrying out feature extraction output/> on the preset relation diagram structure data through a first diagram convolution network of the 0 th layer. Based on the formula (11), carrying out feature extraction output/> on the preset relation diagram structure data through a second diagram convolution network of the 0 th layer. Based onAndFeature extraction, output/>, through a first graph rolling network of layer 1. Based onFeature extraction, output/>, through a first graph rolling network of layer 1. WillAndAs the association relationship feature.
Through the steps S810 to S830, the characteristics of the sparse mode of the angular power spectrum structure from the adjacent grids can be captured based on the graph structure, so as to alleviate the problem of high dryness of the measurement matrix, thereby improving the accuracy of the angular power spectrum prediction.
Based on the association relation characteristics, the average reference signal received power and the antenna array transmission coefficient, the angular power spectrum of the grid area is predicted by the variable decibel leaf inference module.
Referring to FIG. 9, in some embodiments, the associative relationship feature includes first relationship data and second relationship data, the first relationship data being represented asThe second relationship data is expressed asStep S150 may include, but is not limited to, steps S910 to S920:
Step S910, updating the first relation data by a variation Bayesian inference module of the current super network layer based on the association relation characteristics, the average reference signal receiving power and the antenna array transmission coefficient to obtain first candidate relation data;
Step S920, based on the first candidate relation data, the second relation data, the average reference signal received power and the antenna array transmission coefficient, predicting the angular power spectrum of the grid region by the dB-in-signal-of-noise inference module of the next super network layer.
In step S910 of some embodiments, the variational bayesian inference module has an implicit variable signal precisionAnd noise precision. Initializing superparameter Initializing hidden variables、. Based on association relation features (/ >)And) Average reference signal received powerAntenna array transmission coefficientPrecisionAnd noiseBayesian inference is performed by a variational Bayesian inference module to infer first relationship dataAnd updating to obtain first candidate relation data. The specific procedure of Bayesian inference is as follows:
Updating angle power spectrum I.e. update theAngular power spectrum of individual grid regionsMean,,ForAngular power spectrum of individual grid regionsIs a function of the variance of (a),Diag represents selectionIs a diagonal element of (c). Update signal accuracyBy usingUpdate,,. Update support v, i.e. update first relationship data,,,As a gamma function,As a second order gamma function. For positive integers n,The | represents factoring.,ForA first order derivative of z.. Updating noise precisionByUpdateWherein,M is the number of beams and tr represents the trace of the matrix.
In step S920 of some embodiments, based on the first candidate relationship data (updated) Second relationship dataAverage reference signal received powerAntenna array transmission coefficientUpdatedAndAnd predicting the angular power spectrum of the grid region by a hyper-parameter through a variable decibel leaf inference module of the next hyper-network layer. /(I)
Through the steps S910 to S920, the prior graph structure data output by the previous-variation db-leaf inference module can be fully utilized, so that the accuracy of angle power spectrum estimation is improved. At the same time, the need for sample size is reduced during the training phase.
Referring to fig. 10, in some embodiments, step S920 may include, but is not limited to, steps S1010 through S1030:
step S1010, updating the first candidate relation data through a Markov diagram neural network of the next super network layer based on the second relation data to obtain second candidate relation data;
Step S1020, updating the second relation data through a Markov diagram neural network of the next super network layer based on the first candidate relation data to obtain third candidate relation data;
Step S1030, based on the second candidate relationship data, the third candidate relationship data, the average reference signal receiving power and the antenna array transmission coefficient, updating the second candidate relationship data by the variable dB leaf inference module of the next super network layer until obtaining the final candidate relationship data output by the variable dB leaf inference module of the last super network layer, and taking the final candidate relationship data as an angle power spectrum.
In step S1010 of some embodiments, the second relationship data is storedFirst candidate relationship dataAnd (3) inputting the first candidate relationship data into a Markov diagram neural network of the next super network layer, and updating the first candidate relationship data through the Markov diagram neural network based on the formula (10) and the formula (11) to obtain second candidate relationship data.
In step S1020 of some embodiments, based on the first candidate relationship dataBased on equation (10) and equation (11), the second relationship data/>, is transformed by the Markov-map neural network of the next super-network layerAnd updating to obtain third candidate relation data.
In step S1030 of some embodiments, the second candidate relationship data is storedThird candidate relationship dataAverage reference signal received powerAntenna array transmission coefficientUpdatedAndThe variable dB leaf inference module of the super parameter input to the next super network layer carries out the/>, on the second candidate relation dataUpdating until final candidate relation data/>, which is output by a variable dB leaf inference module of the last super network layer, is obtainedFinal candidate relationship dataAs angular power spectrum。
Through the steps S1010 to S1030, the structural sparsity between the angle power spectrums of the adjacent grid areas can be captured, and the problem of high dryness of the measurement matrix is relieved. Meanwhile, the prior diagram structure data of the variational Bayesian inference module is fully utilized, and the accuracy of angle power spectrum prediction is improved.
In step S160 of some embodiments, the wireless environment refers to an environment in which data transmission and communication are performed through a wireless communication technology without using a wired connection. The angle power spectrum represents the power distribution of the signal in different space angles, and the wireless environment is predicted according to the statistical characteristics of the angle power spectrum by analyzing the statistical characteristics of the shape, the average value, the peak position and the peak power of the angle power spectrum, so as to obtain the signal strength, the multipath effect, the channel capacity, the receiving end performance and the like, and generate a wireless environment prediction result. The receiver performance can be measured by signal-to-noise ratio. After obtaining the wireless environment prediction result, the base station can select a proper scheduling algorithm and downlink data block size to ensure that the user terminal obtains the optimal downlink performance under different wireless environments.
In some embodiments, the maximum peak power of the angular power spectrum may be taken as the signal strength. The degree of multipath can be estimated from the number of peak powers in the angular power spectrum, the intensity of the peak powers, and the distribution of the peak powers. The greater the amount of peak power, the more significant the multipath effect. Multipath effects can result in larger amplitude variations in peak power, which can be more severe if the intensity of the peak power is greater. The wider the peak power is distributed over different angular ranges, the more the multipath is. Channel capacity, signal to noise ratio, etc. are calculated from the angular power spectrum.
Let the Markov diagram neural network be represented as Markov-GNN, the variational Bayesian inference module be represented as VBI, the multi-grid localization statistical channel model be represented as MVB-GNN, and the overall forward data path of the multi-grid localization statistical channel model in the embodiment of the application is as follows.
Acquiring average reference signal received power for multiple grid regionsAntenna array transmission coefficient. Initializing superparameterInitialize gradient variable. For each grid regionMarkov-map neural network pair/>, through the j-th super-network layerUpdating, i.e.. Output based on Markov diagram neural network、、、Super-parametersAngular power spectrum estimation, i.e./>, is performed by a variable dB leaf inference module of the jth super network layer. The last super-network layer is the Output layer, and the Markov-FIG. neural network of the Output layer is represented as Output Markov-GNN,. Output of Markov map neural network for last super network layerThe variable dB leaf inference module input to the last super network layer predicts the angle power spectrum, namelyWill beAssigned toObtain the angle power spectrum. It should be noted that, the previous super network layer of the output layer is the J-th super network layer.
The number of antennas is set to 64, the number of spatial angles is set to 208, the number of beams is set to 32, and the number of grids is set to 20. Consider the use of a Uniform Planar Array (UPA) and a Discrete Fourier Transform (DFT) codebook. The real APS (angular power spectrum) of 20 grids was simulated with a path from none to have a probability of 0.001 and a path from have to have a probability of 0.099, and an angular power spectrum APS having 3 non-zero element sparse vectors as the first grid was randomly and uniformly selected. Then, each training data point is generated and gaussian noise is added with a signal-to-noise ratio (SNR) of 10 dB. The performance of the MVB-GNN and other baseline algorithms proposed by the embodiments of the present application were compared. GroupLASSO, conventional Markov-VBI, and MLP-VBI were chosen as baseline algorithms. For GroupLASSO, the row sparseness case is considered. The initial path required in the conventional Markov-VBI is set to 0.02 and 0.08 with probability from none to none and path from none to none. For MLP-VBI, the Markov-GNN module in MVB-GNN is replaced with a simple 3-layer multi-layer perceptron (MLP) module. In addition, three test data sets, each containing 200 samples, common in wireless communication studies, were used for comparison of performance. In the simulated data set (Simulated Dataset), test data is generated in the same manner as training data. In the i.i.d gaussian dataset (i.i.d Gaussian Dataset), test data are generated by using a gaussian distribution matrix of independent co-distributions. RSRP is generated over different grid areas from DeepMIMO datasets. The Accuracy (ACC) of the methods in the simulated data set is shown in fig. 11 a. The accuracy of each method in the i.i.d. gaussian dataset is shown in figure 11 b. The accuracy of the dataset at DeepMIMO for each method is shown in figure 11 c. 11a, 11b and 11c, it can be seen that the accuracy of the angle power spectrum prediction performed by the multi-grid localization statistical channel model according to the embodiment of the application is highest.
Referring to fig. 12, an embodiment of the present application further provides a wireless environment prediction apparatus, which may implement the above wireless environment prediction method, where the apparatus includes:
a first obtaining module 1210, configured to obtain an antenna array transmission coefficient of a base station;
a second obtaining module 1220, configured to obtain an average reference signal received power of the grid area in a preset period;
An input module 1230, configured to input the antenna array transmission coefficient and the average reference signal received power into a multi-grid localization statistical channel model, where the multi-grid localization statistical channel model includes a plurality of super-network layers, and the super-network layers include a markov diagram neural network and a variational bayesian inference module;
A feature extraction module 1240, configured to perform feature extraction on the preset relationship graph structure data through a markov neural network, so as to obtain an association relationship feature; the preset relation diagram structure data is a heterogeneous Markov diagram representation, and is used for representing the association relation between the reference signal received power and the angle power spectrum of the grid region;
A prediction module 1250, configured to predict an angular power spectrum of the grid region by using the variational dawster inference module based on the association characteristic, the average reference signal received power and the antenna array transmission coefficient;
a radio environment prediction module 1260, configured to predict a radio environment from the base station to the grid region according to the angular power spectrum.
The specific implementation of the wireless environment prediction apparatus is basically the same as the specific embodiment of the wireless environment prediction method described above, and will not be described herein.
The embodiment of the application also provides electronic equipment, which comprises a memory and a processor, wherein the memory stores a computer program, and the processor realizes the wireless environment prediction method when executing the computer program. The electronic equipment can be any intelligent terminal including a tablet personal computer, a vehicle-mounted computer and the like.
Referring to fig. 13, fig. 13 illustrates a hardware structure of an electronic device according to another embodiment, the electronic device includes:
The processor 1310 may be implemented by a general-purpose CPU (central processing unit), a microprocessor, an application-specific integrated circuit (ApplicationSpecificIntegratedCircuit, ASIC), or one or more integrated circuits, etc. for executing related programs to implement the technical solution provided by the embodiments of the present application;
Memory 1320 may be implemented in the form of read-only memory (ReadOnlyMemory, ROM), static storage, dynamic storage, or random access memory (RandomAccessMemory, RAM), among others. Memory 1320 may store an operating system and other application programs, and when the technical solutions provided by the embodiments of the present disclosure are implemented in software or firmware, relevant program codes are stored in memory 1320 and invoked by processor 1310 to perform the wireless environment prediction method of the embodiments of the present disclosure;
an input/output interface 1330 for implementing information input and output;
Communication interface 1340, configured to implement communication interaction between the present device and other devices, where communication may be implemented through a wired manner (e.g., USB, network cable, etc.), or may be implemented through a wireless manner (e.g., mobile network, WIFI, bluetooth, etc.);
Bus 1350 for transferring information between components of the device (e.g., processor 1310, memory 1320, input/output interface 1330, and communication interface 1340);
wherein processor 1310, memory 1320, input/output interface 1330, and communication interface 1340 implement a communication connection between each other within the device via bus 1350.
The embodiment of the application also provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program realizes the wireless environment prediction method when being executed by a processor.
The memory, as a non-transitory computer readable storage medium, may be used to store non-transitory software programs as well as non-transitory computer executable programs. In addition, the memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory optionally includes memory remotely located relative to the processor, the remote memory being connectable to the processor through a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The embodiments described in the embodiments of the present application are for more clearly describing the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application, and those skilled in the art can know that, with the evolution of technology and the appearance of new application scenarios, the technical solutions provided by the embodiments of the present application are equally applicable to similar technical problems.
It will be appreciated by persons skilled in the art that the embodiments of the application are not limited by the illustrations, and that more or fewer steps than those shown may be included, or certain steps may be combined, or different steps may be included.
The above described apparatus embodiments are merely illustrative, wherein the units illustrated as separate components may or may not be physically separate, i.e. may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
Those of ordinary skill in the art will appreciate that all or some of the steps of the methods, systems, functional modules/units in the devices disclosed above may be implemented as software, firmware, hardware, and suitable combinations thereof.
The terms "first," "second," "third," "fourth," and the like in the description of the application and in the above figures, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the application described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
It should be understood that in the present application, "at least one (item)" means one or more, and "a plurality" means two or more. "and/or" for describing the association relationship of the association object, the representation may have three relationships, for example, "a and/or B" may represent: only a, only B and both a and B are present, wherein a, B may be singular or plural. The character "/" generally indicates that the context-dependent object is an "or" relationship. "at least one of" or the like means any combination of these items, including any combination of single item(s) or plural items(s). For example, at least one (one) of a, b or c may represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", wherein a, b, c may be single or plural.
In the several embodiments provided by the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, the above-described division of units is merely a logical function division, and there may be another division manner in actual implementation, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.
The units described above as separate components may or may not be physically separate, and components shown as units may or may not be physical units, may be located in one place, or may be distributed over a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.
The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be embodied in essence or a part contributing to the prior art or all or part of the technical solution in the form of a software product stored in a storage medium, including multiple instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method of the various embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory RAM), a magnetic disk, or an optical disk, or other various media capable of storing a program.
The preferred embodiments of the present application have been described above with reference to the accompanying drawings, and are not thereby limiting the scope of the claims of the embodiments of the present application. Any modifications, equivalent substitutions and improvements made by those skilled in the art without departing from the scope and spirit of the embodiments of the present application shall fall within the scope of the claims of the embodiments of the present application.
Claims (7)
1. A wireless environment prediction method, the wireless environment prediction method comprising:
acquiring an antenna array transmission coefficient of a base station;
acquiring average reference signal receiving power of a grid area in a preset period;
Inputting the antenna array transmission coefficients and the average reference signal received power into a multi-grid localization statistical channel model, wherein the multi-grid localization statistical channel model comprises a plurality of super-network layers, and the super-network layers comprise a Markov diagram neural network and a variational Bayesian inference module;
Carrying out feature extraction on the preset relationship diagram structural data through the Markov diagram neural network to obtain association relationship features; the preset relation diagram structure data is a heterogeneous Markov diagram representation, and is used for representing the association relation between the reference signal received power and the angle power spectrum of the grid region;
Predicting an angular power spectrum of the grid region by the variational Bayesian inference module based on the association relation features, the average reference signal received power and the antenna array transmission coefficient;
Predicting a wireless environment between the base station and the grid area according to the angle power spectrum; the markov neural network comprises a first graph convolution network and a second graph convolution network, and the feature extraction is carried out on the preset relation graph structure data through the markov neural network to obtain the association relation feature, and the method comprises the following steps:
extracting first relation features from the preset relation graph structure data through the first graph convolution network to obtain first relation data; the preset relation diagram structure data comprise angle power spectrum priori data and angle power spectrum likelihood data, and the first relation data are used for representing the relation between the angle power spectrum priori data and the angle power spectrum likelihood data;
Extracting second relation features from the preset relation graph structural data through the second graph convolution network to obtain second relation data; the second relation data is used for representing the relation between the angle power spectrum likelihood data and the angle power spectrum priori data;
Obtaining the association relation characteristic according to the first relation data and the second relation data;
the association relation feature comprises first relation data and second relation data, and the predicting, by the variational bayesian inference module, the angular power spectrum of the grid region based on the association relation feature, the average reference signal received power and the antenna array transmission coefficient comprises:
updating the first relation data through the variation Bayesian inference module of the current super network layer based on the association relation characteristics, the average reference signal receiving power and the antenna array transmission coefficient to obtain first candidate relation data;
predicting an angular power spectrum of the grid region by the variational bayesian inference module of the next super network layer based on the first candidate relationship data, the second relationship data, the average reference signal received power and the antenna array transmission coefficient;
The predicting, by the variational bayesian inference module of the next super-network layer, the angular power spectrum of the grid region based on the first candidate relationship data, the second relationship data, the average reference signal received power and the antenna array transmission coefficient includes:
updating the first candidate relationship data through the Markov diagram neural network of the next super network layer based on the second relationship data to obtain second candidate relationship data;
Updating the second relation data through the Markov diagram neural network of the next super network layer based on the first candidate relation data to obtain third candidate relation data;
And updating the second candidate relation data through the variation Bayesian inference module of the next super network layer based on the second candidate relation data, the third candidate relation data, the average reference signal receiving power and the antenna array transmission coefficient until the final candidate relation data output by the variation Bayesian inference module of the last super network layer is obtained, and taking the final candidate relation data as the angle power spectrum.
2. The wireless environment prediction method according to claim 1, wherein the preset time period includes a plurality of time steps, and the acquiring the average reference signal received power of the grid region in the preset time period includes:
Acquiring the power of the base station for transmitting a reference signal to the grid area according to a preset beam forming vector, and acquiring the transmitting power;
acquiring a channel vector from the base station to the grid area in the time step;
determining the receiving power of the reference signal received by the grid area in the time step according to the transmitting power, the channel vector and the preset beam forming vector;
and determining the average reference signal receiving power according to the receiving power of all the time steps.
3. The method according to any one of claims 1 to 2, wherein the acquiring the antenna array transmission coefficient of the base station includes:
acquiring an antenna array guide vector of the base station;
and determining the antenna array transmission coefficient according to the antenna array steering vector and a preset beam forming vector.
4. The method according to any one of claims 1 to 2, wherein the multi-grid localization statistical channel model is trained according to the following steps:
Acquiring a sample antenna array transmission coefficient of the base station;
Acquiring sample average reference signal receiving power of the grid area in a preset period;
Predicting a target angle power spectrum of the grid region through a preset model based on the sample antenna array transmission coefficient and the sample average reference signal receiving power;
Performing loss calculation according to the target angle power spectrum, the sample antenna array transmission coefficient and the sample average reference signal receiving power to obtain unsupervised loss data;
and adjusting network parameters of the preset model according to the unsupervised loss data to obtain the multi-grid localization statistical channel model.
5. A wireless environment prediction apparatus, the apparatus comprising:
the first acquisition module is used for acquiring the antenna array transmission coefficient of the base station;
the second acquisition module is used for acquiring the average reference signal receiving power of the grid area in a preset period;
The input module is used for inputting the antenna array transmission coefficient and the average reference signal receiving power into a multi-grid localization statistical channel model, wherein the multi-grid localization statistical channel model comprises a plurality of super network layers, and the super network layers comprise a Markov diagram neural network and a variational Bayesian inference module;
the feature extraction module is used for carrying out feature extraction on the preset relationship diagram structure data through the Markov diagram neural network to obtain association relationship features; the preset relation diagram structure data is a heterogeneous Markov diagram representation, and is used for representing the association relation between the reference signal received power and the angle power spectrum of the grid region;
The prediction module is used for predicting the angular power spectrum of the grid region through the variation Bayesian inference module based on the association relation characteristic, the average reference signal received power and the antenna array transmission coefficient;
The wireless environment prediction module is used for predicting the wireless environment between the base station and the grid area according to the angle power spectrum;
The markov neural network comprises a first graph convolution network and a second graph convolution network, and the feature extraction is carried out on the preset relation graph structure data through the markov neural network to obtain the association relation feature, and the method comprises the following steps:
extracting first relation features from the preset relation graph structure data through the first graph convolution network to obtain first relation data; the preset relation diagram structure data comprise angle power spectrum priori data and angle power spectrum likelihood data, and the first relation data are used for representing the relation between the angle power spectrum priori data and the angle power spectrum likelihood data;
Extracting second relation features from the preset relation graph structural data through the second graph convolution network to obtain second relation data; the second relation data is used for representing the relation between the angle power spectrum likelihood data and the angle power spectrum priori data;
obtaining the association relation characteristic according to the first relation data and the second relation data;
the association relation feature comprises first relation data and second relation data, and the predicting, by the variational bayesian inference module, the angular power spectrum of the grid region based on the association relation feature, the average reference signal received power and the antenna array transmission coefficient comprises:
updating the first relation data through the variation Bayesian inference module of the current super network layer based on the association relation characteristics, the average reference signal receiving power and the antenna array transmission coefficient to obtain first candidate relation data;
Predicting an angular power spectrum of the grid region by the variational bayesian inference module of the next super network layer based on the first candidate relationship data, the second relationship data, the average reference signal received power and the antenna array transmission coefficient;
The predicting, by the variational bayesian inference module of the next super-network layer, the angular power spectrum of the grid region based on the first candidate relationship data, the second relationship data, the average reference signal received power and the antenna array transmission coefficient includes:
updating the first candidate relationship data through the Markov diagram neural network of the next super network layer based on the second relationship data to obtain second candidate relationship data;
Updating the second relation data through the Markov diagram neural network of the next super network layer based on the first candidate relation data to obtain third candidate relation data;
And updating the second candidate relation data through the variation Bayesian inference module of the next super network layer based on the second candidate relation data, the third candidate relation data, the average reference signal receiving power and the antenna array transmission coefficient until the final candidate relation data output by the variation Bayesian inference module of the last super network layer is obtained, and taking the final candidate relation data as the angle power spectrum.
6. An electronic device comprising a memory and a processor, the memory storing a computer program, the processor implementing the wireless environment prediction method of any one of claims 1 to 4 when the computer program is executed.
7. A computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the wireless environment prediction method of any one of claims 1 to 4.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410106374.5A CN117650861B (en) | 2024-01-25 | 2024-01-25 | Wireless environment prediction method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410106374.5A CN117650861B (en) | 2024-01-25 | 2024-01-25 | Wireless environment prediction method and device, electronic equipment and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN117650861A CN117650861A (en) | 2024-03-05 |
CN117650861B true CN117650861B (en) | 2024-05-17 |
Family
ID=90047999
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202410106374.5A Active CN117650861B (en) | 2024-01-25 | 2024-01-25 | Wireless environment prediction method and device, electronic equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117650861B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN118289652B (en) * | 2024-06-06 | 2024-08-09 | 山东济矿鲁能煤电股份有限公司阳城煤矿 | Unmanned system of monorail crane |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115378521A (en) * | 2021-05-18 | 2022-11-22 | 华为技术有限公司 | Method, device and related equipment for determining wireless channel multipath information |
CN116155412A (en) * | 2023-02-02 | 2023-05-23 | 深圳市大数据研究院 | Wireless channel evaluation method and system |
CN116980933A (en) * | 2023-07-06 | 2023-10-31 | 深圳市大数据研究院 | Antenna parameter adjustment method and device, electronic equipment and storage medium |
-
2024
- 2024-01-25 CN CN202410106374.5A patent/CN117650861B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115378521A (en) * | 2021-05-18 | 2022-11-22 | 华为技术有限公司 | Method, device and related equipment for determining wireless channel multipath information |
CN116155412A (en) * | 2023-02-02 | 2023-05-23 | 深圳市大数据研究院 | Wireless channel evaluation method and system |
CN116980933A (en) * | 2023-07-06 | 2023-10-31 | 深圳市大数据研究院 | Antenna parameter adjustment method and device, electronic equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN117650861A (en) | 2024-03-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103874118B (en) | Radio Map bearing calibrations in WiFi indoor positionings based on Bayesian regression | |
CN117650861B (en) | Wireless environment prediction method and device, electronic equipment and storage medium | |
Zhang et al. | An efficient machine learning approach for indoor localization | |
Wang et al. | Sparse bayesian learning-based 3D radio environment map construction—Sampling optimization, scenario-dependent dictionary construction and sparse recovery | |
CN110412499B (en) | Broadband DOA estimation algorithm based on RSS algorithm under compressed sensing theory | |
CN111707985A (en) | Off-grid DOA estimation method based on covariance matrix reconstruction | |
CN107576931B (en) | Covariance low-dimensional iteration sparse reconstruction-based correlation/coherent signal direction-of-arrival estimation method | |
Gupta et al. | Machine learning-based urban canyon path loss prediction using 28 ghz manhattan measurements | |
CN106291452B (en) | A kind of direction of arrival estimation method based on enhanced greedy algorithm | |
He et al. | Random forests based path loss prediction in mobile communication systems | |
CN109375153A (en) | A kind of multi-path dense signal angle estimation method based on impulse response compressed sensing | |
Liu et al. | UAV-aided radio map construction exploiting environment semantics | |
CN114386321A (en) | AOA and TOF joint estimation method and device for indoor positioning and storage medium | |
Zhong et al. | A distributed particle filtering approach for multiple acoustic source tracking using an acoustic vector sensor network | |
CN115236584A (en) | Meter-wave radar low elevation angle estimation method based on deep learning | |
Wu et al. | A holistic approach to reconstruct data in ocean sensor network using compression sensing | |
CN109407051A (en) | The generalized rank target Sensor Network space-time data localization method of convex optimization fusion graph theory | |
Wang et al. | Sparse bayesian learning-based hierarchical construction for 3D radio environment maps incorporating channel shadowing | |
CN106507473A (en) | A kind of indoor orientation method and device based on compressed sensing algorithm | |
Yang et al. | A correlation-aware sparse Bayesian perspective for DOA estimation with off-grid sources | |
Li et al. | DeFe: Indoor localization based on channel state information feature using deep learning | |
Sun et al. | Wireless communication indoor positioning method in 5G sub-station using deep neural network and location fingerprint algorithm | |
CN116634358A (en) | Terminal positioning method and device and nonvolatile storage medium | |
CN103607768B (en) | Target device localization method under a kind of de-centralized scene and relevant device | |
Gill et al. | Three-way deep neural network for radio frequency map generation and source localization |
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 |