CN111478783B

CN111478783B - Method and equipment for configuring wireless transmission parameters

Info

Publication number: CN111478783B
Application number: CN201910062567.4A
Authority: CN
Inventors: 张思明; 韩双锋; 王桂珍
Original assignee: China Mobile Communications Group Co Ltd; China Mobile Communications Ltd Research Institute
Current assignee: China Mobile Communications Group Co Ltd; China Mobile Communications Ltd Research Institute
Priority date: 2019-01-23
Filing date: 2019-01-23
Publication date: 2023-01-13
Anticipated expiration: 2039-01-23
Also published as: CN111478783A

Abstract

The invention discloses a method and equipment for configuring wireless transmission parameters, which relate to the technical field of wireless communication and are used for solving the problems that the wireless transmission configuration is inaccurate due to the fact that the channel estimation relying on a reference signal in the existing communication system lags, is influenced by time delay and has low accuracy of channel estimation, and the method comprises the following steps: collecting at least one channel information set of at least one cell in real time; estimating channel output of at least one cell by at least one channel information set through a trained space-time correlation model, wherein the space-time correlation model is obtained by training the space-time correlation of a channel by utilizing a machine learning algorithm; and configuring wireless transmission parameters according to the channel output of at least one cell. The invention can accurately estimate the space-time correlation of the cell channel information, thereby reducing the frequency of channel estimation of a base station or a user, reducing the time delay and improving the accuracy of channel estimation and wireless transmission parameter configuration.

Description

Method and equipment for configuring wireless transmission parameters

Technical Field

The present invention relates to the field of wireless communications technologies, and in particular, to a method and a device for configuring wireless transmission parameters.

Background

Channel estimation is an evaluation of a real-time wireless channel in a wireless communication system for guiding a transmission configuration of a physical layer. The current real-time channel acquisition method in the communication system can be mainly divided into three categories, namely estimation based on reference signals, blind estimation and semi-blind estimation. Estimation of reference signals is most widely used in practical systems.

In the actual transmission process, the accuracy of channel estimation is ensured by depending on a large number of intensive reference signals and fast signal processing calculation power, and the requirements on the accuracy and timeliness of channel estimation are high particularly in a high-speed high-frequency high-reliability application scene.

In the prior art, channel estimation must be performed through a reference signal, channel estimation in a current communication system is actually delayed, channel inaccuracy and time delay cause physical layer performance bottleneck, and in addition, transmission parameters configured according to a channel estimation result are inaccurate, so that the requirements of future networks on high speed, low time delay, high spectrum efficiency and high energy efficiency cannot be met.

In summary, in the current communication system, the channel estimation depending on the reference signal is delayed, and is affected by the time delay, and the accuracy of the channel estimation is low, so that the wireless transmission configuration is inaccurate.

Disclosure of Invention

The invention provides a method and equipment for configuring wireless transmission parameters, which are used for solving the problems that in the prior art, the wireless transmission configuration is inaccurate due to the fact that channel estimation relying on a reference signal in a communication system lags, is influenced by time delay and has low accuracy of channel estimation.

In a first aspect, a method for configuring a wireless transmission parameter provided in an embodiment of the present invention includes:

collecting at least one channel information set of at least one cell in real time;

estimating channel output of the at least one cell by the at least one channel information set through a trained space-time correlation model, wherein the space-time correlation model is obtained by training space-time correlation of a channel by utilizing a machine learning algorithm;

and configuring wireless transmission parameters according to the channel output of the at least one cell.

The method is different from the method that the channel estimation is carried out by a reference signal in the prior art, the channel information collected in real time is used as the input of a space-time correlation model through a machine learning algorithm to accurately estimate the channel of at least one cell, the frequency of channel estimation of a base station or a user can be reduced, the time delay is reduced, the channel output with space-time correlation is obtained, the accuracy of the channel estimation is improved, the wireless transmission parameters are configured according to the accurate channel output, and the accuracy of the wireless transmission configuration is improved.

In one possible implementation, the space-time correlation model includes a spatial correlation model and a temporal correlation model;

estimating channel output of at least one cell from the at least one channel information set via a trained space-time correlation model, comprising:

inputting at least one channel information set of the at least one cell at any time into a spatial correlation model to obtain at least one channel spatial correlation information set corresponding to the at least one cell at the time;

and aiming at any cell, inputting a plurality of channel space correlation information sets corresponding to n continuous moments of the cell before the t-th moment into the time correlation model to obtain the channel output of the cell at the t-th moment, wherein t and n are positive integers.

According to the method, a flexible channel model is adopted, the existing channel estimation method based on the reference signal is abandoned, the simple machine learning algorithm models are combined, the correlation between space and time among multiple cells or multiple users is fully considered, the channel estimation can be accurately carried out, the spectral efficiency and the energy efficiency of the system can be improved, the channel estimation of the next position of the channel at the next moment is obtained through the estimation of the real-time channel, and the support is provided for the network planning, network optimization and real-time wireless transmission information configuration.

In a possible implementation manner, if there is one cell, the spatial correlation model is a local spatial neural network model;

for any time, inputting at least one channel information set of the at least one cell at the time into a spatial correlation model to obtain at least one channel spatial correlation information set corresponding to the at least one cell at the time, including:

and aiming at any time, inputting the at least one channel information set of the cell at the time into the local spatial neural network model to obtain the at least one channel spatial correlation information set corresponding to the cell at the time.

According to the method, when one cell is adopted, at least one channel information set of the cell is estimated through a local spatial neural network model, channel spatial correlation among multiple users in the cell is modeled, at least one channel spatial correlation information set with spatial correlation is output, and channel information of the next position can be accurately estimated.

In a possible implementation manner, if there are a plurality of cells, the spatial correlation model includes a local spatial neural network model and a global spatial neural network model;

aiming at any one time, inputting a plurality of channel information sets of all cells at the time into the global spatial neural network model to obtain at least one first channel spatial correlation information set corresponding to each cell; inputting the at least one channel information set of the at least one cell at the time into the local spatial neural network model to obtain at least one second channel spatial correlation information set corresponding to the at least one cell at the time;

vector superposition is carried out on a first channel space correlation information set and a second channel space correlation information set corresponding to the same cell to obtain a third channel space correlation information set;

and taking a set formed by the first channel spatial correlation information set, the second channel spatial correlation information set and the third channel spatial correlation information set as the channel spatial correlation information set corresponding to the same cell at the time.

According to the method, if a plurality of cells exist, a local spatial neural network model is adopted to estimate at least one channel information set of each cell to output a first channel spatial correlation information set, for any one cell, channel spatial correlation among multiple users in the cell is modeled, channel information among the multiple cells is estimated through a global spatial neural network model to output a second channel spatial correlation information set, spatial correlation among the multiple cells is established, and channel information of the next position can be accurately estimated through vector sum of the first spatial correlation information and the second spatial correlation information.

In a possible implementation manner, after estimating the channel output of the at least one cell by using the trained space-time correlation model, the method further includes:

monitoring key performance indicators KPIs of the at least one cell to obtain KPI parameters of the at least one cell;

and adjusting the KPI parameters in the loss functions in the time correlation model and/or the spatial correlation model according to the KPI parameters of the at least one cell.

According to the method, after the channel is estimated through the machine learning algorithm, the accuracy of channel estimation can be judged by monitoring the key performance index of at least one cell in real time, and then the spatial correlation model and/or the time correlation model are optimized according to the monitored KPI, so that a more accurate estimation effect is achieved.

In one possible implementation, the channel information set includes part or all of the following:

channel parameters, channel real-time impulse response.

The method can estimate the channel through the real-time impulse response of the channel or the parameters representing the channel quality, wherein the information output obtained through the machine learning algorithm is consistent with the type of the channel information in the channel information set input into the machine learning algorithm.

In a second aspect, an apparatus for configuring wireless transmission parameters provided in an embodiment of the present invention includes: a processor and a transceiver:

the processor: for collecting in real time at least one set of channel information for at least one cell using the transceiver; estimating channel output of the at least one cell by the at least one channel information set through a trained space-time correlation model, wherein the space-time correlation model is obtained by training space-time correlation of a channel by utilizing a machine learning algorithm; and configuring wireless transmission parameters according to the channel output of the at least one cell.

the processor is specifically configured to:

and for any cell, inputting a plurality of channel space correlation information sets corresponding to n continuous moments of the cell before the t-th moment into the time correlation model to obtain the channel output of the cell at the t-th moment, wherein t and n are positive integers.

and aiming at any time, inputting at least one channel information set of the cell at the time into the local spatial neural network model to obtain at least one channel spatial correlation information set corresponding to the cell at the time.

for any time, inputting the at least one channel information set of the at least one cell at the time into a spatial correlation model to obtain a channel spatial correlation information set corresponding to the at least one cell at the time, including:

and taking a set formed by the first channel spatial correlation information set, the second channel spatial correlation information set and the third channel spatial correlation information set as the channel spatial correlation information set corresponding to the same region at the time.

In one possible implementation, the processor is further configured to:

channel parameters, channel real-time impulse response.

In a third aspect, an apparatus for configuring a wireless transmission parameter provided in an embodiment of the present invention includes: at least one processing unit and at least one memory unit, wherein the memory unit stores program code that, when executed by the processing unit, causes the apparatus to perform the following.

the processing unit is specifically configured to:

In one possible implementation, the processing unit is further configured to:

channel parameters, channel real-time impulse response.

In a fourth aspect, the present application further provides a computer storage medium having a computer program stored thereon, where the program is executed by a processing unit to implement the steps of the method according to the above-mentioned aspects.

In addition, for technical effects brought by any one implementation manner of the second aspect to the fourth aspect, reference may be made to technical effects brought by different implementation manners of the first aspect, the second aspect, and the third aspect, and details are not repeated here.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without inventive exercise.

Fig. 1 is a schematic diagram illustrating a method for configuring wireless transmission parameters according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of obtaining channel output by a machine learning algorithm according to an embodiment of the present invention;

fig. 3A is a schematic diagram of space division of a cell according to an embodiment of the present invention;

fig. 3B is a schematic diagram of channel estimation performed by a first single-cell spatial correlation model according to an embodiment of the present invention;

fig. 3C is a schematic diagram of channel estimation performed by a second single-cell spatial correlation model according to an embodiment of the present invention;

fig. 3D is a schematic diagram of channel estimation performed by a third single-cell spatial correlation model according to an embodiment of the present invention;

fig. 4A is a schematic diagram of space division of multiple cells according to an embodiment of the present invention;

fig. 4B is a schematic diagram of a multi-cell spatial correlation model for channel estimation according to an embodiment of the present invention;

fig. 5A is a schematic diagram of a time correlation model of channel estimation according to an embodiment of the present invention;

fig. 5B is a schematic diagram of another time correlation model for channel estimation according to an embodiment of the present invention;

fig. 5C is a diagram illustrating channel estimation according to an embodiment of the present invention;

FIG. 6A is a schematic diagram of a method for constructing a spatial correlation model according to an embodiment of the present invention;

FIG. 6B is a schematic diagram of a method for constructing a time correlation model according to an embodiment of the present invention;

fig. 6C is a flowchart of model construction, channel estimation and channel monitoring according to an embodiment of the present invention;

fig. 7 is a flowchart of a complete method for channel estimation according to an embodiment of the present invention;

fig. 8 is a schematic diagram of a first apparatus for configuring wireless transmission parameters according to an embodiment of the present invention;

fig. 9 is a schematic diagram of a second apparatus for channel estimation according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention clearer, the present invention will be described in further detail with reference to the accompanying drawings, and it is apparent that the described embodiments are only a part of the embodiments of the present invention, not all of the 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.

Some of the words that appear in the text are explained below:

1. the term "and/or" in the embodiments of the present invention describes an association relationship of associated objects, and indicates that three relationships may exist, for example, a and/or B may indicate: a exists alone, A and B exist simultaneously, and B exists alone. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship.

2. The terminal in the embodiment of the invention refers to a terminal capable of supporting channel estimation, namely a mobile phone, a tablet, a computer and the like.

3. The "network side device" referred to in the embodiments of the present invention is a device such as a base station that can perform channel estimation, for example, a macro base station, a home base station, and the like.

4. The Term "reciprocity" in the embodiments of the present invention refers to channel reciprocity of TD-LTE (Time Division Long Term Evolution), where uplink and downlink of a TD-LTE system are transmitted in different Time slots of the same frequency resource, so that within a relatively short Time (coherence Time of channel propagation), it can be considered that channel fading experienced by transmission signals of uplink and downlink are the same.

The application scenario described in the embodiment of the present invention is for more clearly illustrating the technical solution of the embodiment of the present invention, and does not form a limitation on the technical solution provided in the embodiment of the present invention, and it can be known by a person skilled in the art that with the occurrence of a new application scenario, the technical solution provided in the embodiment of the present invention is also applicable to similar technical problems. In the description of the present invention, the term "plurality" means two or more unless otherwise specified.

The performance of a wireless communication system is greatly affected by wireless channels, such as shadow fading and frequency selective fading, so that the propagation path between a transmitter and a receiver is very complicated. Wireless channels are not fixed and predictable as wired channels, but rather have a large degree of randomness, which presents a significant challenge to the design of a receiver.

In coherent detection of an OFDM (Orthogonal Frequency Division Multiplexing) system, a channel needs to be estimated, and the accuracy of channel estimation directly affects the performance of the whole system. In order to accurately recover a transmission signal at a receiving end, people adopt various measures to resist the influence of multipath effect on a transmission signal, and the realization of a channel estimation technology needs to know information of a wireless channel, such as the order of the channel, the doppler shift, the multipath delay or the parameters of channel impulse response and the like. Therefore, channel parameter estimation is a key technology for implementing a wireless communication system. Whether detailed channel information can be obtained or not is an important index for measuring the performance of a wireless communication system, so that a transmitting signal can be correctly demodulated at a receiving end. Therefore, the research on the channel parameter estimation algorithm is a significant work.

The current real-time channel acquisition method in the communication system can be mainly divided into three categories, namely estimation based on reference signals, blind estimation and semi-blind estimation. Estimation of reference signals is most widely used in practical systems.

For the estimation of the channel of a TDD (Time Division duplex) system, the channel depends on the reciprocity of the uplink and downlink channels and the uplink reference signal sent by the user, the base station calculates the channel impulse response, and based on the reciprocity of the uplink and downlink channels, the characteristics of the downlink channel and the uplink channel on the same frequency band can be considered to be consistent within a certain coherent Time, and the channel estimation is used for physical layer transmission such as downlink beamforming (beamforming) and MIMO (Multiple-Input Multiple-output system); since the channel varies in time, an accurate estimate needs to be obtained within coherence time and coherence frequency, and the assumption of channel reciprocity holds.

For estimation of a Channel of an FDD (Frequency Division duplex) system, a base station transmits a pilot signal to a user, the user performs Channel estimation after receiving the pilot signal transmitted by the base station, and then feeds back Channel Quality Information such as CSI (Channel State Information), CQI (Channel Quality Indicator), PMI (Precoding Matrix Indicator), and the like to the base station, and the base station performs physical layer downlink transmission according to the user feedback.

In the prior art, channel estimation must be performed through a reference signal, channel estimation in the current communication system is actually delayed, channel uncertainty and time delay will cause a physical layer performance bottleneck, and the requirements of future networks on high speed, low time delay, high spectrum efficiency and high energy efficiency cannot be met.

Meanwhile, a future communication network will be a complex heterogeneous network (e.g., macro-micro stereo networking, cell overlay coverage, multi-RAT (Radio Access Technology) coverage, and dynamic and flexible cell configuration), and if coverage and interference among multiple cells are not accurately estimated, system capacity and coverage in a region will be damaged.

Therefore, the invention provides a method and equipment for configuring wireless transmission parameters, which can simultaneously and accurately estimate multi-user channel information in a multi-cell environment through the combination of machine learning algorithms, and can reduce the frequency of channel estimation of a base station or a user; the method has the advantages that the accurate channel information and interference information are provided, wireless transmission parameters can be accurately configured, and the method can be used for network planning and network optimization.

In view of the above-mentioned situations, the following describes the embodiments of the present invention in further detail with reference to the drawings.

As shown in fig. 1, a method for configuring wireless transmission parameters in an embodiment of the present invention specifically includes the following steps:

step 100: collecting at least one channel information set of at least one cell in real time;

step 101: estimating channel output of the at least one cell by the at least one channel information set through a trained space-time correlation model, wherein the space-time correlation model is obtained by training space-time correlation of a channel by utilizing a machine learning algorithm;

step 102: and configuring wireless transmission parameters according to the channel output of the at least one cell.

Through the scheme, different from the prior art that channel estimation is carried out through a reference signal, channel information collected in real time is used as input of a space-time correlation model through a machine learning algorithm to accurately estimate a channel of at least one cell, the frequency of channel estimation of a base station or a user can be reduced, time delay is reduced, channel output with space-time correlation is obtained, the accuracy of channel estimation is improved, wireless transmission parameters are configured according to the accurate channel output, the accuracy of wireless transmission configuration is improved, and the method can be used for network planning and network optimization.

Wherein the channel information set includes, but is not limited to, some or all of the following:

channel parameters, channel real-time impulse response.

In the embodiment of the present invention, at least one channel information set of at least one cell is collected in real time, where the channel information set is determined according to the channel information of the cell collected by the network side device in real time.

For example, assuming that there are 3 cells (cells) 1,2, and 3 with adjacent positions, the 3 cells are divided into spatial grids, as shown in fig. 4A, CQI information of each spatial grid is determined by a base station in a real-time collection manner, and a set formed by CQI information of all spatial grids in one cell is used as a channel information set corresponding to the cell.

In the embodiment of the present invention, the channel parameter may be a large-scale channel parameter: RSRP (Reference Signal Receiving Power), CQI, path Loss (path Loss), and the like, and may be small-scale channel parameters: time-frequency angle domain expansion, related bandwidth, related time, related distance, interference characteristic parameters and the like.

In the embodiment of the invention, the space-time correlation model is trained through a machine learning algorithm, at least one channel information set is subjected to the trained space-time correlation model to estimate the channel output of at least one cell, the trained model is used for realizing the accurate estimation of the channel output, and further, the wireless transmission parameters can be accurately configured.

Specifically, a channel information set corresponding to a cell is input into a space-time correlation model obtained by performing space-time correlation training through a machine learning algorithm for channel estimation, and the type of the obtained channel output is consistent with that of the input channel information set, for example, if the input is user RSRP, the obtained channel estimation is also RSRP, and if the input is an angle expansion parameter, the obtained channel estimation is also angle expansion parameter estimation of the channel; if the input is CIR (Impulse response of channel), the obtained CIR estimation of the channel is obtained; if the input is the RSRP and the angle spread parameter, the estimation of the RSRP and the angle spread parameter of the channel is obtained.

If the channel is linear, then the channel estimate is an estimate of the system impulse response. Through channel estimation, the receiver can obtain the impulse response of the channel, thereby providing the required CSI for subsequent coherent demodulation.

In the embodiment of the present invention, when the machine learning algorithm is used to train the space-time correlation of the channel, an AE (automatic encoder) method may be adopted, where AE is an unsupervised learning method in an artificial intelligent neural network, and is mainly used to learn a group of data to obtain an encoding method, which is a common dimension reduction algorithm. The AE includes an input layer, an hidden layer, and an output layer in a simple structure, and the number of nodes (nodes) of the output layer and the input layer must be kept uniform. Depending on the learning method, the main categories include DAE (Denoising auto encoder), SAE (Sparse auto encoder), CAE (constrained auto encoder), and the like. In the training process, back-propagation methods such as conjugate gradient method are often used to obtain information of multiple hidden layers.

In the embodiment of the present invention, at least one channel information set of at least one cell is subjected to channel estimation by using a space-time correlation model to obtain channel output of the at least one cell, specifically, the at least one channel information set of the cell is estimated by using a spatial correlation model and a temporal correlation model, for any time, the at least one channel information set of the at least one cell is input into a trained spatial correlation model to obtain the at least one channel spatial correlation information set of the at least one cell at the time, and then, for any cell, at least n channel spatial correlation information sets of n consecutive times before the time t are input into the temporal correlation model to obtain channel output of the cell at the time t, where n and t are positive integers.

Wherein the spatial correlation model includes, but is not limited to, some or all of the following:

convolutional neural networks, self-coding neural networks.

Wherein the time correlation model includes, but is not limited to, some or all of the following:

LSTM (Long Short-Term Memory), ARIMA (Autoregressive Integrated Moving Average Model), SVR (support vector regression Model).

As shown in fig. 2, the spatial correlation model and the temporal correlation model are integrated to estimate the channel through the combination of the machine learning algorithm, and the spatial and temporal correlation between cells or between users in the same cell is considered, so that the accurate estimation of the channel information of the cell is realized, the frequency of channel estimation of the base station or the user can be reduced, and the accurate channel information and the interference information are provided, and the method can be used for network planning and network optimization. By estimating the real-time channel, the method provides support for the configuration of the wireless transmission information with network specification and network optimization and real time.

For example, at time t1, one channel information set H1 of cell 1 and one channel information set H2 of cell 2 are input into the spatial correlation model, and spatial correlation information sets H1 'and H2' of the next positions of cell 1 and cell 2 at time t1 are output through the spatial correlation model.

Optionally, the channel information in the set H1 and the channel information in the set H2 may be the same type of information, for example, the information in H1 and H2 may both be CQIs; the channel information in the set H1 and the channel information in H2 may also be different types of information, for example, the information in H1 is RSRP, and the information in H2 is an angle spread parameter.

The output of the spatial correlation model is used as the input of the temporal correlation model, and when the time correlation model estimates the channel at the time t for any cell, n sets of spatial correlation information at n consecutive times before the time t can be input into the temporal correlation model for monitoring, and the final result is the channel information set estimation result at the next position at the time t, i.e. the channel output.

For example, the spatial correlation information set of the cell 1 at the time t1 is H1', the spatial correlation information set of the cell 1 at the time t2 is H1", and the spatial correlation information set of the cell 1 at the time t3 is H1"', where t1, t2, t3, and t4 are consecutive 4 times, when estimating the channel output of the cell 1 at the time t4, H1', H1", and H1"' are input into the temporal correlation model for estimation, and the input data set H1"" is the channel output of the next position of the cell 1 at the time t 4.

It should be noted that, the manner of performing channel estimation by combining the spatial correlation model and the temporal correlation model, which is listed in the embodiment of the present invention, is only an example, and any manner of estimating a channel by a machine learning method is applicable to the embodiment of the present invention.

And aiming at the spatial correlation model, when the spatial correlation model is used for estimating at least one channel information set of at least one cell, inputting the at least one channel information set of the at least one cell at any moment into the spatial correlation model to obtain at least one channel spatial correlation information set corresponding to the at least one cell at the moment.

Optionally, the spatial correlation model recited in the embodiment of the present invention may estimate spatial correlation of channels of multiple cells, or may estimate spatial correlation between multiple users of a cell, where the spatial correlation models corresponding to channel estimation for multiple cells and for a cell are different, and are respectively described as follows:

if the cell is one, the spatial correlation model is a local spatial neural network model;

specifically, at any time, at least one channel information set of a certain cell at the time t is input into the local spatial neural network model to obtain at least one channel spatial correlation information set corresponding to the cell at the time t.

Taking Local Automatic Encoder (LAE) as an example, the Local spatial neural network model is trained as follows:

in the coverage area of a cell, the cell 1 may be divided into M × N spatial grids, which mainly refers to one representation of spatial positions, as shown in fig. 3A, the cell 1 is divided into M × N spatial grids, the channel information of each spatial grid (x, y) at time t1 is used as the minimum input Data unit of the model, the channel information of all the spatial grids (x, y) at time t in the cell 1 constitutes a Data patch (Data patch) H1t1,1 (x, y) (i.e., a channel information set H1t1,1 (x, y) of the cell 1, where x =1,2, …, M; y =1,2, …, N), H1t1,1 (x, y) is input into LAE to obtain encoded Data Ht1'1 (x, y) of the cell 1 (i.e., the spatial grid 1 at time t1 is subjected to channel estimation), and as shown in fig. 3B.

In fig. 3C, ht1,2 (x, y) is a channel information set of the cell 2 at the time t1, and Ht1'2 (x, y) is a channel spatial correlation information set after channel estimation is performed on the cell 2 at the time t 1; in fig. 3D, ht2,1 (x, y) is a channel information set of the cell 1 at the time t2, and Ht2'1 (x, y) is a channel spatial correlation information set after the cell 1 performs channel estimation at the time t 2.

For cell 1, if the set of spatial correlation information of the channel at time t1 of cell 1 obtained by the spatial correlation model is Ht1'1 (x, y), and the set of spatial correlation information of the channel at time t2 is Ht2'1 (x, y), then the temporal correlation model should be input to Ht1'1 (x, y) and Ht2'1 (x, y) when estimating the channel at time t 3.

When channel estimation is carried out, at least one channel information set of a cell is input into LAE to establish channel space correlation among users in the cell, so that channel output among a plurality of users in the same cell has space correlation, and the channel space correlation information set of the cell is used as input of a time correlation model to estimate the time correlation.

If the number of the cells is multiple, the spatial correlation model comprises a local spatial neural network model and a global spatial neural network model;

specifically, for any one time, inputting a plurality of channel information sets of all cells at the time into a global spatial neural network model to obtain at least one first channel spatial correlation information set corresponding to each cell; inputting at least one channel information set of each cell at the moment into a local spatial neural network model to obtain at least one second channel spatial correlation information set corresponding to the cell at the moment;

and taking a set formed by the first channel spatial correlation information set, the second channel spatial correlation information set and the third channel spatial correlation information set as a channel spatial correlation information set corresponding to the same cell at any time.

In the embodiment of the present invention, taking a Local spatial neural network model as LAE (Local auto encoder) and a Global spatial neural network model as GAE (Global auto encoder) as an example, a training process of a spatial correlation model is as follows:

assume that there are three

cells

1,2,3 located adjacently as shown in fig. 4A, and each cell is divided into M times N spatial grids. The channel information of each spatial grid (x, y) at time t is used as the minimum input data unit of the model, and all the spatial grid (x, y) channel information in cell i constitutes data slice Hi (x, y) (i.e. a set of channel information of cell i at time t, where i =1,2,3,x =1,2, …, M; y =1,2, …, N), where all data in a region is to be used as input of a machine learning model, for example, H1 (x, y), H2 (x, y), and H3 (x, y) are respectively input into three LAE to obtain coded output data pieces H 'i (x, y) of each cell (i.e., a second channel spatial correlation information set of the cell i at time t), so as to establish spatial correlation between spatial grids inside the cell, i.e., spatial correlation of users inside the cell, and H1 (x, y), H2 (x, y), and H3 (x, y) are simultaneously input into the GAE to obtain H "i (x, y) (i.e., a first channel spatial correlation information set of the cell i at time t), thereby achieving establishment of spatial correlation between cells, and H' i (x, y) is superimposed with H" i (x, y) to obtain H "i (x, y) vectors ^～ i (x, y) (the third channel spatial correlation information set at time t of cell i) represents the spatial correlation between multi-cell and multi-user, as shown in fig. 4B.

For the cell i, the channel spatial correlation information of the cell i at the time t obtained by the spatial correlation model is as follows: from H 'i (x, y), H' i (x, y) and H ^～ i (x, y), when estimating the channel at time t +1, the spatial correlation information of the channel at time t should be aggregated as follows: from H 'i (x, y), H' i (x, y), H ^～ i (x, y) and a channel space correlation information set corresponding to the first n-1 continuous time moments at the time t are input into a time correlation model.

When channel estimation is carried out, a plurality of channel information sets of a plurality of cells are input into a plurality of LAEs to establish channel space correlation among users in each cell, a plurality of channel information sets of a plurality of cells are input into one GAE to establish channel space correlation among a plurality of cells, so that channel output among the plurality of users in the cells has space correlation, and the channel space correlation information set of each cell is used as input of a time correlation model to estimate the time correlation.

In the embodiment of the present invention, AE is an algorithm model based on a neural network, which can make its output simulate input through encoding and decoding processes, as shown in fig. 4B, if the input is multiple sets of channel information of multiple cells, then the output of the model is multiple sets of channel spatial correlation information of multiple cells, and the dimension of the output information may be consistent with the input, or may be a compressed set of channel information spatial correlation.

It should be noted that the spatial correlation models listed in the embodiments of the present invention are only examples, and any spatial correlation model capable of performing channel spatial correlation estimation is applicable to the embodiments of the present invention.

Aiming at a time correlation model, aiming at any cell, after a channel space correlation information set of the cell is obtained through the space correlation model, the channel space correlation information sets corresponding to n continuous moments before the t-th moment of the cell are input into the time correlation model to obtain the channel output of the cell at the t-th moment, wherein t and n are positive integers.

As shown in fig. 5A, when determining the channel output of the cell 1 at time t, the set of spatial correlation information of the channels of the cell at n consecutive times before time t is input to the LSTM, and the channel output of the cell 1 at time t is output as the output result.

For example, the sets of spatial correlation information of channels of cell 1 at 4 consecutive time instants before the 5 th time instant are: h1 '(x, y), H2' (x, y), H3 '(x, y), and H4' (x, y), and the channel spatial correlation information at these 4 times is collectively input to the channel output of cell 1 at time 5, which is the input result obtained by the temporal correlation model.

Optionally, in consideration of possible difference between dimensions of input and output information of the spatial correlation model, when channel estimation is performed through the temporal correlation model, for any cell, after a channel spatial correlation information set of the cell is obtained through the spatial correlation model, channel spatial correlation information sets corresponding to n consecutive times of the cell before the t-th time and a determined channel information set are input into the temporal correlation model to obtain a channel output of the cell at the t-th time, where t and n are positive integers.

As shown in fig. 5B, when determining the channel output of the cell 1 at time t, the set of channel spatial correlation information at n consecutive times before time t of the cell and the determined set of channel information are input to the LSTM, and the channel output of the cell 1 at time t is the output result.

Compared with fig. 5A, the input of the time correlation model further includes a channel information set of a cell determined by real-time measurement of the network side device, in this case, the influence of the dimension difference between the time correlation model and the spatial correlation model can be reduced, so that the estimation result is more accurate.

As shown in fig. 5C, n is 4, and the sets of channel information measured by the base station in real time at 4 consecutive times before the 5 th time in the cell 1 are respectively: h1 (x, y), H2 (x, y), H3 (x, y), and H4 (x, y), where the first channel spatial correlation information sets at the 4 time instants are: h1 "(x, y), H2" (x, y), H3 "(x, y), H4" (x, y), and the second set of channel spatial correlation information is: h1 '(x, y), H2' (x, y), H3 '(x, y), and H4' (x, y), and the third channel spatial correlation information sets are: h1 ^～ (x，y)、H2 ^～ (x，y)、H3 ^～ (x，y)、H4 ^～ (x, y) inputting the first, second and third channel spatial correlation information sets and the channel information set at the previous 4 time points into the input result obtained by the time correlation model, i.e. the channel output of the cell 1 at the 5 th time point.

It should be noted that the spatial correlation models listed in the embodiments of the present invention are only examples, and any spatial correlation model capable of performing channel estimation is applicable to the embodiments of the present invention.

Optionally, in order to reduce the reconstruction loss function of each AE in the training process of the spatial correlation model, the loss function may be set as the cross entropy of the value of the input AE and the encoded value of the AE output, and the algorithm for optimizing the loss function may adopt a common back propagation algorithm such as: SGD (stored Gradient Decent random Gradient descent), RMSprop (Root Mean Square Prop, root Mean Square direction propagation), and the like.

Optionally, the number of layers of GAEs and LAE and the associated hyperparameters of the machine learning model, such as the loss function, may be designed according to the channel output characteristics to be estimated, because too many layers of LAE may significantly increase the training time of the model.

Assuming that the channel output to be estimated is only a large-scale signal strength, such as RSRP, since RSRP does not vary much over time and is only a numerical value, the number of layers of the neural network of GAE and LAE is small, and the loss function can be implemented by simple MSE (Mean Square Error); assuming that small-scale channel information, such as signal impulse response over a bandwidth, needs to be estimated, since the real-time impulse response of the channel changes rapidly with time and the data is complex, and the information dimension of the input and output of the model is large, such as Fast Fourier Transform (FFT), a deeper neural network is needed for simulation, and the loss function can be defined as the true value of the input model and the euclidean distance of the estimated value of the output of the model.

In the embodiment of the present invention, the training of the model may be performed off-line, the LAE spatial correlation model of each cell may be trained on the base station side, the spatial and temporal correlation model of a multi-cell combination may be placed on the side of a DU (Distributed Unit) or a CU (Central Unit) or a network management system for training, after the model training, the trained model is sent to the base station side, and the channel estimation is performed through the base station.

As shown in fig. 6A, a training process of a multi-cell channel spatial correlation model provided in the embodiment of the present invention is as follows: the method comprises the steps of firstly collecting channel information of a whole cell of a cell through user feedback or base station measurement, then carrying out data processing on the obtained information through dimensionality reduction, denoising, interference removal and the like, and finally carrying out channel data reconstruction, finally establishing an information database of multi-cell channels and interference through the collection, processing and reconstruction of the information of a plurality of cells, and training through a machine learning module to obtain a multi-cell channel spatial correlation model.

As shown in fig. 6B, a training process of a time correlation model of a multi-cell channel provided in the embodiment of the present invention is as follows: and constructing information at a plurality of moments from t-n to t-1, which is input by the multi-cell spatial correlation model, into a multi-cell channel and interference history database, and then training the multi-cell channel and interference history database by a machine learning module to obtain a time correlation model of the multi-cell channel.

It should be noted that the model training process mentioned in the embodiments of the present invention is only an example, and any model training process is applicable to the embodiments of the present invention.

In the embodiment of the present invention, after performing channel estimation on a cell by using a machine learning algorithm to obtain a channel output of at least one cell, a KPI (Key performance indicator) monitoring mechanism may be further used to monitor KPI parameters of at least one cell in real time, and feed the monitored KPI parameters back to the time correlation model and/or the spatial correlation model, and adjust the KPI parameters in a loss function in the time correlation model and/or the spatial correlation model according to the KPI parameters, so as to optimize the KPI and the optimization model.

Among them, KPI parameters include, but are not limited to, some or all of the following:

user number, cell throughput, coverage, SINR (Signal to Interference plus Noise Ratio).

In the embodiment of the present invention, when determining the loss function of the model, KPI parameters may be used as parameters of the model loss function, for example:

loss = (d total interference of neighbor cell)/a regional cell user number + b cell throughput/c user SINR.

Among them, loss is a Loss function, and a, b, c, and d are coefficients of the Loss function (i.e., weights of KPI parameters), and the Loss function can be reduced by using an algorithm such as back propagation.

In the embodiment of the invention, after KPI monitoring is executed, KPI parameters in the loss function can be timely adjusted according to the monitoring result, and then the model is optimized.

Specifically, after KPI monitoring is executed, a trough minimum value of the Loss function can be searched according to a monitoring result back propagation algorithm, and when the neural network parameters are iterated, the method of continuously updating the model parameters along the opposite direction through calculating the gradient (first derivative) of the Loss function is used for achieving the convergence of the Loss function, so that the model is optimized.

For example, based on the recorded channel information of the cell and the corresponding KPI value, loss = (d × total interference of neighbor cell)/a × number of regional cell users + b × cell throughput/c × user SINR is generated, and the minimum value of the Loss function is found through a back propagation algorithm while keeping the weights a, b, c, and d unchanged. When the neural network parameters are iterated, the KPI parameters are adjusted by calculating the gradient (first derivative) of the Loss function and continuously updating the macro weight w and/or the bias b of the machine learning model along the opposite direction of the gradient so as to achieve the convergence of the Loss function.

In the embodiment of the present invention, a spatial correlation model and a temporal correlation model may be trained in advance in an offline state, and specifically, a channel database of multiple cells is formed by performing data collection and preprocessing on channel information, and data storage and forwarding are performed, where the data includes not only spatial information but also temporal information. By using the data, the establishment of a space correlation model and a time correlation model of the channel can be carried out, after the trained model is issued to the base station side, the estimation of the channel is executed through the model, the output information obtained by inputting the information measured in real time by the base station into the space correlation model is transferred to the time correlation model, and finally the channel output of the channel at the next moment and the next position is obtained. As shown in fig. 6C, after the channel estimation is performed on the channel through the model, the KPI of the multiple cells is monitored in real time, so that the base station can adjust resource allocation and the like according to the channel, and optimize the model by adjusting the KPI parameters in the loss function.

Optionally, when the monitored KPI is degraded severely, it may be possible to pass through a fallback mechanism, the KPI degradation severely may be caused by the fact that the trained machine learning model is not accurate to the result of channel estimation, (possibly because the channel information of the cell itself is inaccurate or outdated, that is, the data input into the model is not accurate, resulting in an unsatisfactory algorithm result, or the machine learning algorithm needs to continue to optimize), and when a fallback condition is reached (for example, at least one KPI parameter is degraded to be less than the minimum fallback threshold value), it may be possible to fallback to the conventional channel estimation algorithm to estimate the thought output through the KPI monitoring. And the fallback is that the base station no longer uses the model obtained by the machine learning algorithm to perform channel estimation, but reverts to the traditional channel estimation algorithm and performs channel estimation through the reference signal.

For example, the number of regional cell users of a plurality of cells is reduced by half compared with the number of preset regional cell users, and the cell throughput of at least one cell is much smaller than the preset cell throughput, when a backoff condition is reached (for example, at least one KPI parameter is decreased to be smaller than a minimum backoff threshold), a result obtained after channel estimation is performed by using a trained model is not too accurate, so that a backoff base station can be set, and when the backoff condition is reached, channel estimation can be performed by using a conventional channel estimation algorithm by using KPI monitoring.

It should be noted that the manner of monitoring the optimization model through KPIs listed in the embodiment of the present invention is only an example, and any manner that can optimize the model is applicable to the embodiment of the present invention.

As shown in fig. 7, a complete method for channel estimation according to an embodiment of the present invention includes:

step 700, the network side device determines a channel information set according to the collected channel information and performs preprocessing to form a channel database of multiple cells;

701, establishing a channel space correlation model and a time correlation model by using a channel information set in a channel database by network side equipment;

step 702, a network side device acquires a plurality of channel information sets of a plurality of cells and a plurality of users in real time, and inputs the plurality of channel information sets into a spatial correlation model to obtain a plurality of channel spatial correlation information sets with spatial correlation;

step 703, the network side device inputs a plurality of sets of channel spatial correlation information with spatial correlation into the time correlation model to obtain channel outputs of a plurality of cells, and configures wireless transmission parameters according to the channel outputs of the plurality of cells;

step 704, the network side equipment monitors KPI parameters of multiple cells in real time;

step 705, the network side device optimizes the spatial correlation model and/or the temporal correlation model according to the KPI parameters.

Based on the same inventive concept, the embodiment of the present invention further provides a device for configuring wireless transmission parameters, and since the device is the device in the method in the embodiment of the present invention, and the principle of the device for solving the problem is similar to that of the method, the implementation of the device may refer to the implementation of the method, and repeated details are omitted.

As shown in fig. 8, an embodiment of the present invention further provides an apparatus for configuring wireless transmission parameters, where the apparatus includes: processor 800 and transceiver 801:

the processor 800 is configured to: collecting at least one channel information set of at least one cell in real time using the transceiver 801; estimating channel output of the at least one cell by the at least one channel information set through a trained space-time correlation model, wherein the space-time correlation model is obtained by training space-time correlation of a channel by utilizing a machine learning algorithm; and configuring wireless transmission parameters according to the channel output of the at least one cell.

Optionally, the space-time correlation model includes a spatial correlation model and a temporal correlation model;

the processor 800 is specifically configured to:

Optionally, if there is one cell, the spatial correlation model is a local spatial neural network model;

Optionally, if there are multiple cells, the spatial correlation model includes a local spatial neural network model and a global spatial neural network model;

Optionally, the processor 800 is further configured to:

adjusting KPI parameters in a loss function in the temporal correlation model and/or the spatial correlation model according to KPI parameters of the at least one cell.

Optionally, the channel information set includes part or all of the following:

channel parameters, channel real-time impulse response.

As shown in fig. 9, an embodiment of the present invention further provides an apparatus for configuring wireless transmission parameters, where the apparatus includes: at least one processing unit 900, and at least one storage unit 901, wherein the storage unit 901 stores program code that, when executed by the processing unit 900, causes the apparatus to perform the following:

the processing unit 900 is specifically configured to:

Optionally, if there are a plurality of cells, the spatial correlation model includes a local spatial neural network model and a global spatial neural network model;

Optionally, the processing unit 900 is further configured to:

Optionally, the channel information set includes part or all of the following:

channel parameters, channel real-time impulse response.

An embodiment of the present invention further provides a computer-readable non-volatile storage medium, which includes a program code, and when the program code runs on a computing terminal, the program code is configured to enable the computing terminal to perform the steps of the method for channel estimation according to the embodiment of the present invention.

The present application is described above with reference to block diagrams and/or flowchart illustrations of methods, apparatus (systems) and/or computer program products according to embodiments of the application. It will be understood that one block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, and/or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer and/or other programmable data processing apparatus, create means for implementing the functions/acts specified in the block diagrams and/or flowchart block or blocks.

Accordingly, the present application may also be embodied in hardware and/or in software (including firmware, resident software, micro-code, etc.). Furthermore, the present application may take the form of a computer program product on a computer-usable or computer-readable storage medium having computer-usable or computer-readable program code embodied in the medium for use by or in connection with an instruction execution system. In the context of this application, a computer-usable or computer-readable medium may be any medium that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.

While the preferred embodiments of the present application have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all alterations and modifications as fall within the scope of the application.

It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.

Claims

1. A method for configuring wireless transmission parameters, the method comprising:

aiming at any cell, inputting a plurality of channel space correlation information sets corresponding to n continuous moments of the cell before the t moment into a time correlation model to obtain the channel output of the cell at the t moment, wherein t and n are positive integers, and the space correlation model and the time correlation model are obtained by training the space-time correlation of a channel by using a machine learning algorithm;

2. The method of claim 1, wherein if there is one of the cells, the spatial correlation model is a local spatial neural network model;

3. The method of claim 1, wherein if there are a plurality of cells, the spatial correlation model comprises a local spatial neural network model and a global spatial neural network model;

4. The method according to any one of claims 1 to 3, wherein, after inputting a plurality of sets of channel spatial correlation information corresponding to n consecutive times before the t-th time into the time correlation model to obtain the channel output of the cell at the t-th time, for any one cell, the method further comprises:

and adjusting the KPI parameters in the loss function of the time correlation model and/or the spatial correlation model according to the KPI parameters of the at least one cell.

5. The method of claim 1, wherein the set of channel information comprises some or all of:

channel parameters, channel real-time impulse response.

6. An apparatus for configuring wireless transmission parameters, the apparatus comprising: a processor and a transceiver:

the processor: for collecting, in real time, at least one set of channel information for at least one cell using the transceiver; aiming at any time, inputting at least one channel information set of the at least one cell at the time into a spatial correlation model to obtain at least one channel spatial correlation information set corresponding to the at least one cell at the time; for any cell, inputting a plurality of channel spatial correlation information sets corresponding to n times of the cell before the t-th time into a time correlation model to obtain the channel output of the cell at the t-th time, wherein t and n are positive integers, and the spatial correlation model and the time correlation model are obtained by training the space-time correlation of a channel by using a machine learning algorithm; and configuring wireless transmission parameters according to the channel output of the at least one cell.

7. The apparatus of claim 6, wherein if there is one of the cells, the spatial correlation model is a local spatial neural network model;

8. The apparatus of claim 6, wherein if there are a plurality of cells, the spatial correlation model comprises a local spatial neural network model and a global spatial neural network model;

9. The device of any of claims 6-8, wherein the processor is further configured to:

10. The apparatus of claim 6, wherein the set of channel information comprises some or all of:

channel parameters, channel real-time impulse response.

11. An apparatus for configuring wireless transmission parameters, the apparatus comprising: at least one processing unit and at least one memory unit, wherein the memory unit stores program code which, when executed by the processing unit, causes the apparatus to perform the steps of the method of any of claims 1 to 5.

12. A computer-storable medium on which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 5.