CN115622596B - Rapid beam alignment method based on multi-task learning - Google Patents
Rapid beam alignment method based on multi-task learning Download PDFInfo
- Publication number
- CN115622596B CN115622596B CN202211591391.XA CN202211591391A CN115622596B CN 115622596 B CN115622596 B CN 115622596B CN 202211591391 A CN202211591391 A CN 202211591391A CN 115622596 B CN115622596 B CN 115622596B
- Authority
- CN
- China
- Prior art keywords
- task
- base station
- codebook
- pilot signals
- beam alignment
- 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 35
- 238000005259 measurement Methods 0.000 claims abstract description 23
- 230000005540 biological transmission Effects 0.000 claims abstract description 21
- 238000001514 detection method Methods 0.000 claims abstract description 15
- 238000013136 deep learning model Methods 0.000 claims abstract description 8
- 238000012549 training Methods 0.000 claims description 24
- 239000013598 vector Substances 0.000 claims description 23
- 230000006870 function Effects 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 239000000654 additive Substances 0.000 claims description 4
- 230000000996 additive effect Effects 0.000 claims description 4
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 230000000052 comparative effect Effects 0.000 claims description 2
- 238000010801 machine learning Methods 0.000 abstract description 7
- 238000012545 processing Methods 0.000 abstract description 7
- 238000004891 communication Methods 0.000 description 5
- 238000005457 optimization Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 235000015429 Mirabilis expansa Nutrition 0.000 description 1
- 244000294411 Mirabilis expansa Species 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 235000013536 miso Nutrition 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000000523 sample Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/02—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas
- H04B7/04—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas
- H04B7/0413—MIMO systems
- H04B7/0456—Selection of precoding matrices or codebooks, e.g. using matrices antenna weighting
-
- 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
-
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/02—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas
- H04B7/04—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas
- H04B7/06—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the transmitting station
- H04B7/0613—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the transmitting station using simultaneous transmission
- H04B7/0615—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the transmitting station using simultaneous transmission of weighted versions of same signal
- H04B7/0617—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the transmitting station using simultaneous transmission of weighted versions of same signal for beam forming
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/02—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas
- H04B7/04—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas
- H04B7/08—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the receiving station
- H04B7/0837—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the receiving station using pre-detection combining
- H04B7/0842—Weighted combining
- H04B7/086—Weighted combining using weights depending on external parameters, e.g. direction of arrival [DOA], predetermined weights or beamforming
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Electromagnetism (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The invention provides a fast beam alignment method based on multitask learning, which comprises the following steps: s1, a base station generates a plurality of pilot signals based on beams in a detection codebook and outputs the pilot signals to a user terminal; s2, the user terminal measures the power of the pilot signals and feeds back the measurement result to the base station terminal; s3, selecting the optimal transmission beam by the base station through a deep learning model based on the measurement result; and S4, the optimal transmission beam is adopted between the base station end and the user end to implement data transmission so as to realize beam alignment. The method applies machine learning to the signal processing task, and can quickly realize accurate beam alignment.
Description
Technical Field
The invention relates to the technical field of wireless communication, in particular to a rapid beam alignment method based on multi-task learning.
Background
In fifth and future generations of wireless communication, millimeter wave communication is considered as one of the important solutions to provide higher transmission rates and more device access. Millimeter wave signals face severe propagation losses and attenuation blockage compared to electromagnetic waves below 6 gigahertz (Hz). For example, the attenuation of electromagnetic waves at 60GHz is more than 10dB/Km, while at 700MHz it is only about 0.01dB/Km. Fortunately, the millimeter wave wavelength is small and the required antenna size is also small. Thus, the inventors can package a large number of antenna elements, utilize a large scale array of transmitter and receiver antennas, and achieve a highly directional shaped beam, thereby maintaining a high strength received signal. Therefore, how to find a high-quality coding mode to realize beam alignment is a key problem for realizing a high-quality millimeter wave communication system.
The currently released 5G standard employs a beam alignment framework based on beam scanning and searching, measurement and reporting. Base Stations (BSs) and User Equipments (UEs) must search for all possible beam pair combinations, thus causing significant beam scanning overhead and delay. Therefore, researchers have proposed reducing the complexity of beam scanning by hierarchical beam searching. In the hierarchical search strategy, the BS and the UE first scan a wide beam and then iteratively narrow the search space to obtain the best transmission beam, however, the performance of the hierarchical search is sensitive to high noise, which easily causes error accumulation. Inspired by the recent success of Machine Learning (ML) in signal processing tasks, researchers explored ML solutions to the beam alignment problem. For example, researchers have proposed a regression model based on vehicle location to predict the received power of all beams in a precoding codebook. Considering the direction of the mobile device, researchers have realized predicting the best beam in the precoding codebook based on the classification model of the position and direction of the mobile user, however, additional sensors are required to be equipped for obtaining the position and direction information, and additional feedback overhead is also caused.
In order to improve the accuracy of beam alignment without additional sensors, researchers have proposed that the BS may scan a small sounding codebook to sense the current channel, then select a number of beamforming vectors based on the feedback measurements, perform a second round of test on the selected vectors and select the best beamforming vector from the selected vectors. Nevertheless, the existing methods still require a certain number of probe beams to achieve high accuracy alignment. Therefore, the related work of the current beam alignment method faces the following challenges: 1) A large number of detection beams are needed to realize the sensing of the channel, which results in a large wave speed scanning overhead; 2) Repeated interaction between the BS and the UE is required, resulting in additional control and feedback overhead; 3) The accuracy of beam alignment is limited.
Therefore, there is a need to provide a beam alignment method that solves the complexity and accuracy problems of beam alignment.
Disclosure of Invention
Solves the technical problem
Aiming at the defects in the prior art, the invention provides a rapid beam alignment method based on multi-task learning, which applies machine learning to a signal processing task and can rapidly realize accurate beam alignment.
Technical scheme
In order to realize the purpose, the invention is realized by the following technical scheme:
the invention provides a fast beam alignment method based on multi-task learning, which comprises the following steps:
s1, a base station generates a plurality of pilot signals based on beams in a detection codebook and outputs the pilot signals to a user terminal;
s2, the user terminal measures the power of the pilot signals and feeds back the measurement result to the base station terminal;
s3, selecting the optimal transmission beam by the base station through a deep learning model based on the measurement result;
and S4, the optimal transmission beam is adopted between the base station end and the user end to implement data transmission so as to realize beam alignment.
Further, the pilot signal
Comprises the following steps:
wherein,
、
and &>
Respectively representing a transmit power, a channel vector and a noise power of->
Is greater than or equal to the additive complex noise>
Represents a base signal->
Representing a beamforming vector.
Further, the sounding codebook is a discrete fourier transform codebook such that each beam is steered to a discrete direction, wherein the beamforming vectors
Is defined as:
wherein,
、
and &>
Respectively representing the carrier wavelength, the antenna spacing and the size of the discrete fourier transform codebook.
Further, an index of the optimal transmission beam
Comprises the following steps:
further, the base station scans a smaller detection codebook
And sensing a channel, and then directly selecting a beamforming vector according to a measurement result of compressed sensing and statistical prior information.
Further, an index of the beamforming vector selected directly
Comprises the following steps:
wherein,
represents a beam selection function, based on the beam selection function>
Is greater than or equal to>
Wherein, in the process,
indicating the ^ th or greater in the sounding codebook>
An acceptance signal of each detection codeword;
And &>
Respectively representing the channel distribution at the base station side and the power limit of the pilot signal.
Further, the training tasks of the deep learning model specifically include a main task and an auxiliary task, the main task includes a training beam scanning and beam selection task, and the auxiliary task is used for sharing domain knowledge with the main task to assist the training of the main task.
Further, the training beam scanning is modeled by a complex neural network with parameterized intensity, and the beam scanning refers to that the base station generates a plurality of pilot signals based on beams in a sounding codebook and outputs the pilot signals to the user terminal; the training beam selection models a beam selection equation into a classifier constructed by two fully connected layers, wherein the input of the beam selection equation is the noisy signal power obtained in the beam scanning stage, and the output is the likelihood value for selecting each codeword in the discrete Fourier transform codebook.
Further, the auxiliary tasks include a channel reconstruction task and a contrast representation task.
Further, the channel reconstruction task models a regression model comprising two fully-connected layers based on the received signals in the beam scanning stage; the comparative representation task is used to share similar representations for the same scan measurements at different noise powers.
Advantageous effects
The invention provides a fast beam alignment method based on multi-task learning, which applies machine learning to a signal processing task and can fast realize accurate beam alignment; furthermore, the method provided by the invention adopts the steps that a small detection codebook is firstly scanned to sense the channel, and then the beamforming vector is directly selected according to the measurement result of compressed sensing and the statistic prior information, so that the overhead of beam alignment by an exhaustion method is greatly reduced; finally, the method provided by the invention combines the training detection codebook and the beam selector, and improves the accuracy by adopting a training model of multi-task learning.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the prior art descriptions will be briefly described below. It is obvious that the drawings in the following description are only some embodiments of the invention, and that for a person skilled in the art, other drawings can be derived from them without inventive effort.
Fig. 1 is a schematic diagram illustrating steps of a fast beam alignment method based on multi-task learning according to an embodiment of the present invention;
fig. 2 is a schematic diagram of a deep learning model in the fast beam alignment method based on multi-task learning according to an embodiment of the present invention;
fig. 3 is a schematic diagram of a specific application of the fast beam alignment method based on multi-task learning according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. It is to be understood that the embodiments described are only a few embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, an embodiment of the present invention provides a fast beam alignment method based on multitask learning, including the following steps:
s1, a base station generates a plurality of pilot signals based on beams in a detection codebook and outputs the pilot signals to a user terminal;
s2, the user terminal measures the power of the pilot signals and feeds back the measurement result to the base station terminal;
s3, selecting the optimal transmission beam by the base station through a deep learning model based on the measurement result;
and S4, data transmission is carried out between the base station end and the user end by adopting the optimal transmission beam so as to realize beam alignment.
In this embodiment, for step S1, the base station first couples to the basic signal
Processing is performed based on the beam in the sounding codebook such that the base signal->
Generating a pilot signal->
The specific processing formula is as follows:
wherein,
、
and &>
Respectively representing a transmit power, a channel vector and a noise power of->
Is greater than or equal to the additive complex noise>
Represents a base signal, <' > based>
Representing a beamforming vector. I.e. the base signal->
And the beamforming vector->
Multiply to realize the basic signal->
The beamforming of (1).
In this embodiment, the sounding codebook employed by the present invention is generally a discrete fourier transform codebook, such that each beam is steered to a discrete direction, wherein the beamforming vector
Is defined as:
wherein N represents the number of base station antennas,
、
and &>
Respectively representing carrier wavelengths,Antenna spacing and size of the discrete Fourier transform codebook.
In this embodiment, the main task of the beam training is to search the best beam (i.e. find the beam with the highest measured signal power) for the subsequent data transmission phase, and use the channel
The user side adopts the fifth->
Individual beamforming vector
The signal-to-noise ratio at time is:
thus, in the discrete Fourier transform codebook, the index of the best transmission beam
Comprises the following steps:
the inventor researches and finds that it is extremely difficult to obtain perfect channel state in limited time, and the solution in the current 5G standard is to use all alternative beamforming vectors exhaustively
Transmitting a pilot signal and then measuring all beamforming vectors->
Lower pilot signal power->
Such an exhaustive search would generate a large training overhead.
In the present embodiment, in order to cope with the largerTraining overhead, the base station end can sense a channel by scanning a smaller discrete Fourier transform codebook, then directly select a beamforming vector according to a measurement result of compressed sensing and statistical prior information, and then directly select an index of the beamforming vector at the moment
Comprises the following steps:
wherein, the
Represents the size, or->
Represents a beam selection function, <' > based on>
Independent variable of (2)
In which>
Indicating a smaller detection codebook pick>
In a fifth or fifth sun>
An acceptance signal of each detection codeword;
And &>
Respectively representing the channel distribution at the base station side and the power limit of the pilot signal.
In this embodiment, the training tasks of the deep learning model specifically include a main task and an auxiliary task, the main task includes a training beam scanning task and a beam selection task, and the auxiliary task is used for sharing domain knowledge with the main task to assist the training of the main task.
In this embodiment, referring to fig. 2, the training beam scanning is modeled by a beam scanning process with a plurality of neural networks parameterized in intensity, where the beam scanning refers to that the base station generates a plurality of pilot signals based on beams in a sounding codebook and outputs the pilot signals to the ue; the training beam selection models a beam selection equation into a classifier constructed by two fully-connected layers, wherein the input of the beam selection equation is the noisy signal power obtained in a beam scanning stage, and the output is the likelihood value for selecting each codeword in the sounding codebook.
First, in the training phase, the inventors base the historical channel data set
Constructing training data, the beam selection problem is modeled as a multi-classification problem, where the number of classification classes is->
I.e., the number of candidate beams. Thus, label information for a multi-category problem may be expressed as ≧ or>
Wherein
the beam scanning process is then modeled by a Complex Neural Network (Complex NN). It is worth mentioning that unlike the usual angular parameterized Complex NN, the present invention proposes a strength parameterized Complex NN whose learnable parameters can be expressed as
. In particular, complex NN is based on the channel condition->
As an input, outputs the corresponding pilot signal power->
. The specific calculation process is as follows:
the Complex NN layer first computes the array of noiseless received signals and then mixes the received signals with additive Complex noise to compute the array of noisy received signals.
Finally, complexNN obtains noisy signal power measurements by calculating the 2-norm of the received signal
And a noiseless signal power measurement>
:
In this embodiment, the beam selection task specifically includes:
as described in the data preparation phase, the inventors apply the beam selection equation
Modeling as a classifier constructed from two fully-connected layers, parameters of the fully-connected layersExpressed as->
. Beam selection equation pick>
The input of (a) is the noisy signal power->
The output is the likelihood value of each code word in the selected transmission codebook:
since the problem considered was a multi-classification problem, the inventors have looked at beam scanning parameters
And a beam selection parameter->
Joint optimization is performed to minimize the cross entropy loss function:
in this embodiment, the auxiliary tasks generally include a channel reconstruction task and a contrast representation task.
In this embodiment, the channel reconstruction task is targeted based on the received signals during the beam scanning phase
，The inventors modeled this reconstruction process as a regression model with two fully connected layers, the output of which can be expressed as
Since the problem under consideration is a regression problem, asThe inventors scanned the beam parameters
And channel reconstruction parameters>
Performing a joint optimization to minimize a mean square error loss function:
in this embodiment, the comparison indicates that the task requires that the learned features be able to distinguish between anchor, positive and negative examples. The issue points for this task are: the same scan measurement at different noise powers should share a similar representation. And the characteristics of different channels should be different even though they belong to the same class. In general, features with discriminative power represent increased resistance of the classifier to noise. In particular, on a given channel
Noisy signal power measurement under the same channel>
Noiseless signal power measurement>
And a noisy signal power measurement of another channel>
The inventors scan a beam parameter->
And a beam selection parameter>
The optimization is performed so as to minimize the triplet losses:
finally, the model is trained in an end-to-end fashion. I.e. joint optimization by minimizing multi-objective losses:
wherein
And &>
Is a hyper-parameter that balances the preferences of different tasks.
Finally, passing the learned parameters
The inventors can construct a scan beam->
And the reception power array measured at the user terminal is greater than or equal to the actual physical scanning beam scan>
Said user side will->
Feeding back to the base station end through an uplink channel, and the base station end will->
An input beam selection model selecting an optimal transmission beam +>
. The tasks of channel reconstruction and contrast representation are only calculated in a training stage, actual deployment is not needed, and no additional cost is caused. />
In this embodiment, referring to fig. 3, the inventor generally considers a mm-wave mimo communication system using this method, where each of the base stations has N antennas and serves the ue with multiple single antennas. Although the user side usually has also an antenna array, for simplicity, the inventors only consider the MISO scenario where beam alignment is performed at the base station side. It should be understood by those skilled in the art that if the ue has multiple antennas, beam alignment of the ue can be performed similarly.
The invention has the advantages that the invention provides a fast beam alignment method based on multi-task learning, which applies machine learning to a signal processing task and can fast realize accurate beam alignment; furthermore, the method provided by the invention adopts the steps that a small detection codebook is firstly scanned to sense the channel, and then the beamforming vector is directly selected according to the measurement result of compressed sensing and the statistic prior information, so that the overhead of beam alignment by an exhaustion method is greatly reduced; finally, the method provided by the invention combines the training detection codebook and the beam selector, and improves the accuracy by adopting a training model of multi-task learning.
The above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and the modifications or the substitutions do not cause the essence of the corresponding technical solutions to depart from the scope of the technical solutions of the embodiments of the present invention.
Claims (6)
1. A fast beam alignment method based on multitask learning is characterized by comprising the following steps:
s1, a base station generates a plurality of pilot signals based on beams in a detection codebook and outputs the pilot signals to a user terminal;
s2, the user terminal measures the power of the pilot signals and feeds back the measurement result to the base station terminal;
s3, selecting the optimal transmission beam by the base station through a deep learning model based on the measurement result;
s4, the optimal transmission beam is adopted between the base station end and the user end to carry out data transmission so as to realize beam alignment, and the pilot signal
Comprises the following steps:
wherein,
、
and &>
Respectively representing a transmit power, a channel vector and a noise power of->
The additive complex noise of (a) is, device for selecting or keeping>
Represents a base signal, <' > based>
Representing beamforming vectors, the sounding codebook being a discrete Fourier transform codebook such that each beam is steered in a discrete direction, wherein a beamforming vector ≦>
Is defined as:
wherein N represents the number of base station antennas,
、 
and &>
Respectively representing the wavelength of a carrier, the distance between antennas and the size of the discrete Fourier transform codebook, wherein a training task of the deep learning model specifically comprises a main task and an auxiliary task, the main task comprises a training beam scanning task and a beam selection task, the auxiliary task is used for sharing domain knowledge with the main task to assist the training of the main task, the training beam scanning process is modeled by a complex neural network with parameterized intensity, and the beam scanning refers to that the base station generates a plurality of pilot signals based on beams in a detection codebook and outputs the pilot signals to the user terminal; the training beam selection models a beam selection equation into a classifier constructed by two fully connected layers, wherein the input of the beam selection equation is the noisy signal power obtained in the beam scanning stage, and the output is the likelihood value for selecting each codeword in the discrete Fourier transform codebook.
2. The method of claim 1, wherein the index of the best transmission beam is an index of the best transmission beam
Comprises the following steps:
3. the fast beam alignment method based on multitask learning as claimed in claim 1, wherein said base station end scans a sounding codebook
Sensing a channel, and then directly selecting a beamforming vector according to a measurement result of compressed sensing in combination with statistical prior information, wherein->
Indicating the size of the sounding codebook.
4. The fast multi-task learning-based beam alignment method according to claim 3, wherein the index of the beamforming vector selected directly is
Comprises the following steps:
wherein,
represents a beam selection function, <' > based on>
Is greater than or equal to>
Wherein is present>
Indicating the ^ th or greater in the sounding codebook>
An acceptance signal of each detection codeword;
And &>
Respectively representing the channel distribution at the base station side and the power limit of the pilot signal.
5. The fast beam alignment method based on multi-task learning of claim 1, wherein the auxiliary tasks include a channel reconstruction task and a contrast representation task.
6. The fast beam alignment method based on multi-task learning as claimed in claim 5, wherein the channel reconstruction task models a regression model including two fully connected layers based on the received signals during the beam scanning phase; the comparative representation task is used to share similar representations for the same scan measurements at different noise powers.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211591391.XA CN115622596B (en) | 2022-12-12 | 2022-12-12 | Rapid beam alignment method based on multi-task learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211591391.XA CN115622596B (en) | 2022-12-12 | 2022-12-12 | Rapid beam alignment method based on multi-task learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115622596A CN115622596A (en) | 2023-01-17 |
| CN115622596B true CN115622596B (en) | 2023-04-18 |
Family
ID=84880551
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211591391.XA Active CN115622596B (en) | 2022-12-12 | 2022-12-12 | Rapid beam alignment method based on multi-task learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115622596B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118786696A (en) * | 2023-02-01 | 2024-10-15 | 新华三技术有限公司 | Training data acquisition method, device, base station, terminal and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110365612A (en) * | 2019-06-17 | 2019-10-22 | 浙江大学 | A kind of deep learning Beam Domain channel estimation methods based on approximate Message Passing Algorithm |
| CN114844545A (en) * | 2022-05-05 | 2022-08-02 | 东南大学 | Communication beam selection method based on sub6GHz channel and partial millimeter wave pilot frequency |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3751753A1 (en) * | 2019-06-13 | 2020-12-16 | Mitsubishi Electric R&D Centre Europe B.V. | Method and system of massive mimo communciation |
| WO2021112592A1 (en) * | 2019-12-03 | 2021-06-10 | 엘지전자 주식회사 | Artificial intelligence-based beam management method in wireless communication system and apparatus therefor |
| KR102204783B1 (en) * | 2020-07-09 | 2021-01-18 | 전남대학교산학협력단 | Deep learning-based beamforming communication system and method |
| US11588539B2 (en) * | 2020-09-23 | 2023-02-21 | Northeastern University | Coordination-free mmWave beam management with deep waveform learning |
| US20220368399A1 (en) * | 2021-05-14 | 2022-11-17 | Qualcomm Incorporated | Techniques for selecting a beam pair using single layer measurements |
-
2022
- 2022-12-12 CN CN202211591391.XA patent/CN115622596B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110365612A (en) * | 2019-06-17 | 2019-10-22 | 浙江大学 | A kind of deep learning Beam Domain channel estimation methods based on approximate Message Passing Algorithm |
| CN114844545A (en) * | 2022-05-05 | 2022-08-02 | 东南大学 | Communication beam selection method based on sub6GHz channel and partial millimeter wave pilot frequency |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115622596A (en) | 2023-01-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111954228B (en) | Beam management method, beam management device, electronic equipment and computer readable storage medium | |
| US9155097B2 (en) | Methods and arrangements for beam refinement in a wireless network | |
| Rezaie et al. | Location-and orientation-aided millimeter wave beam selection using deep learning | |
| Heng et al. | Learning site-specific probing beams for fast mmWave beam alignment | |
| Morais et al. | Position-aided beam prediction in the real world: How useful GPS locations actually are? | |
| CN110417444A (en) | A kind of millimeter wave channel rays training method based on deep learning | |
| CN113438002B (en) | LSTM-based analog beam switching method, device, equipment and medium | |
| WO2020253156A1 (en) | Data-driven beam tracking method and apparatus for mobile millimeter wave communication system | |
| CN115622596B (en) | Rapid beam alignment method based on multi-task learning | |
| Xu et al. | Cramér-Rao bound and antenna selection optimization for dual radar-communication design | |
| CN111446999A (en) | Position-assisted beam alignment method and system based on multi-arm forced theft | |
| Zhang et al. | Deep learning of near field beam focusing in terahertz wideband massive MIMO systems | |
| Papasotiriou et al. | Outdoor THz fading modeling by means of gaussian and gamma mixture distributions | |
| Heng et al. | Learning probing beams for fast mmWave beam alignment | |
| CN109936399A (en) | A kind of insincere junction network antenna selecting method based on deep neural network | |
| Garkisch et al. | Codebook-based user tracking in irs-assisted mmwave communication networks | |
| Wang et al. | Intelligent beam training with deep convolutional neural network in mmwave communications | |
| Aquino et al. | A Review of Direction of Arrival Estimation Techniques in Massive MIMO 5G Wireless Communication Systems | |
| Song et al. | Fast beamforming strategy: Learning the AoD of the dominant path | |
| Chen et al. | Beam Tracking Based on Multi-Point Radar Sensing and DNN Prediction | |
| US20230108467A1 (en) | Beamforming multipath wireless channels using neural networks | |
| Wang et al. | Fast MmWave Beam Alignment Method with Adaptive Discounted Thompson Sampling | |
| CN117220738A (en) | Model training method, multi-base station cooperative beam tracking prediction method and device | |
| Zhao et al. | Intelligent Multi-peak Beam Training in mmWave Communications with Deep Neural Networks | |
| Medra et al. | New Machine Learning Approach for Low Overhead Multi-Beam Prediction |
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 |