CN111435926A

CN111435926A - MIMO system channel prediction method, device, medium and equipment

Info

Publication number: CN111435926A
Application number: CN201910027634.9A
Authority: CN
Inventors: 张思明; 范志明
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: 2019-01-11
Filing date: 2019-01-11
Publication date: 2020-07-21
Anticipated expiration: 2039-01-11
Also published as: CN111435926B

Abstract

The present invention relates to the field of wireless technologies, and in particular, to a method, an apparatus, a medium, and a device for channel prediction in a Multiple Input Multiple Output (MIMO) system. In the scheme of the invention, according to the CIRs of partial MIMO links in the MIMO system, the CIRs of other MIMO links can be obtained by utilizing a neural network model which is correspondingly established in advance in the MIMO system. Based on the correlation among the MIMO links in the MIMO system, the CIR of the unknown MIMO link can be calculated from the CIR of the known MIMO link through the neural network model, and compared with the MIMO channel estimation method, the method does not need to solve all pilot frequencies, even does not need to send the pilot frequencies on all antennas, thereby reducing the operation amount, reducing the occupation of wireless resources and effectively improving the spectrum efficiency and the energy efficiency. And even under the condition that the channel changes rapidly, the prediction time delay can be ensured to be small, and the prediction accuracy can be effectively ensured.

Description

MIMO system channel prediction method, device, medium and equipment

Technical Field

The present invention relates to the field of wireless technologies, and in particular, to a method, an apparatus, a medium, and a device for channel prediction in a Multiple Input Multiple Output (MIMO) system.

Background

MIMO is the fundamental technology of modern wireless mobile communication, and MIMO channel estimation has been the focus of research when determining the Channel Impulse Response (CIR) of the MIMO link.

In the current MIMO Orthogonal Frequency Division Multiplexing (OFDM) communication system, MIMO channel estimation mainly uses preset time-frequency domain resources to periodically send pilot frequencies on all antennas, and obtains channel impulse response of each MIMO link through a channel estimation algorithm. Here, the base station may send pilot, and the terminal performs measurement and estimation; or the terminal sends pilot frequency, and the base station makes measurement and estimation. In real wireless communication, it is assumed that the channels do not change much at a frequency point within a certain time, that is, the CIRs of single-frequency-point single-time points derived from the pilots can be considered to be accurate within a correlation time (coherence time) and a correlation bandwidth (coherence bandwidth), and the base station and the terminal perform subsequent physical layer data processing, such as channel precoding, digital coding and decoding, using the CIRs obtained by channel estimation.

However, when determining the channel impulse response based on MIMO channel estimation, pilot frequencies need to be periodically transmitted on all antennas, which occupies a large amount of wireless resources and causes low spectral efficiency and energy efficiency due to large computation amount.

In addition, in the fifth generation mobile communication system (5G) key technologies, such as massive antenna technology (MassiveMIMO) and millimeter wave (mmWave) technology, it can be seen that the number of antennas is increasing and the channel changes very quickly.

Under the condition, if the pilot frequency is periodically transmitted on each antenna for MIMO channel estimation, the wireless resource occupation is more obvious, the spectrum efficiency and the energy efficiency are reduced, the change speed of the channel in a high-speed scene or a millimeter wave frequency band is high, strong calculation force is needed to support large-scale MIMO channel estimation, the complexity of the traditional channel estimation algorithm is high, the estimation time delay is difficult to guarantee, and negative influence is generated on the accuracy of the algorithm.

Disclosure of Invention

The embodiment of the invention provides a method, a device, a medium and equipment for predicting a channel of an MIMO system, which are used for solving the problems of low spectrum efficiency and low energy efficiency when determining the channel impulse response of an MIMO link through channel estimation.

The invention provides a channel prediction method of a multiple-input multiple-output (MIMO) system, which comprises the following steps:

determining Channel Impulse Responses (CIRs) of at least two MIMO links in the MIMO system;

and obtaining the CIRs of other MIMO links in the MIMO system by utilizing a pre-established neural network model corresponding to the MIMO system according to the CIRs of the at least two MIMO links.

The invention also provides a device for predicting the channel of the MIMO system, which comprises:

a determining module, configured to determine channel impulse responses CIR of at least two MIMO links in the MIMO system;

and the prediction module is used for obtaining the CIRs of other MIMO links in the MIMO system by utilizing a pre-established neural network model corresponding to the MIMO system according to the CIRs of the at least two MIMO links.

The present invention also provides a non-volatile computer storage medium having stored thereon an executable program for execution by a processor to perform the steps of implementing the method as described above.

The invention also provides a multi-input multi-output MIMO system channel prediction device, which comprises a memory, a processor, a transceiver and a bus interface; the processor is used for reading the program in the memory and executing:

determining, by the transceiver, Channel Impulse Responses (CIRs) for at least two MIMO links in the MIMO system; and obtaining the CIRs of other MIMO links in the MIMO system by utilizing a pre-established neural network model corresponding to the MIMO system according to the CIRs of the at least two MIMO links.

According to the scheme provided by the embodiment of the invention, the CIRs of other MIMO links can be obtained by utilizing the pre-established neural network model corresponding to the MIMO system according to the CIRs of part of MIMO links in the MIMO system. Based on the correlation among the MIMO links in the MIMO system, the CIR of the unknown MIMO link can be calculated from the CIR of the known MIMO link through the neural network model, and compared with the MIMO channel estimation method, the method does not need to solve all pilot frequencies, even does not need to send the pilot frequencies on all antennas, thereby reducing the operation amount, reducing the occupation of wireless resources and effectively improving the spectrum efficiency and the energy efficiency. And even under the condition that the channel changes rapidly, the prediction time delay can be ensured to be small, and the prediction accuracy can be effectively ensured.

Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

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 description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.

Fig. 1 is a schematic flowchart of a channel prediction method of a MIMO system according to an embodiment of the present invention;

fig. 2 is a schematic channel diagram of a 2x2 MIMO system according to an embodiment of the present invention;

fig. 3 is a flowchart illustrating a channel prediction method of a MIMO system according to a second embodiment of the present invention;

fig. 4 is a schematic diagram of a simulation result of channel prediction of an MIMO system according to a second embodiment of the present invention;

FIG. 5 is a schematic diagram illustrating a time-expanded representation of a bidirectional RNN network according to a second embodiment of the present invention;

fig. 6 is a schematic diagram of a model using a time-distributed algorithm according to a second embodiment of the present invention;

fig. 7 is a schematic structural diagram of a channel prediction apparatus of a MIMO system according to a third embodiment of the present invention;

fig. 8 is a schematic structural diagram of a MIMO system channel prediction apparatus according to a fourth 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.

It should be noted that, the "plurality" or "a plurality" mentioned herein means two or more. "and/or" describes the association relationship of the associated objects, meaning that there may be three relationships, e.g., a and/or B, which may mean: 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.

The terms "first," "second," and the like in the description and in the claims, and in the drawings described above, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein.

Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

Example one

An embodiment of the present invention provides a method for predicting a channel of an MIMO system, where the flow of the steps of the method may be as shown in fig. 1, and the method includes:

step 101, determining CIR of partial MIMO link.

When performing channel prediction for a MIMO system, in this step, CIRs of at least two MIMO links in the MIMO system may be determined.

The execution subject of the present embodiment may be understood as a device using a neural network model.

It should be noted that the entity device using the neural network model may be any entity device (which may be understood as any network node). Considering that the CIRs of at least two MIMO links in the MIMO system can be directly measured and estimated by the base station (i.e. pilot is transmitted by the terminal and measured and estimated by the base station), or measured and estimated by the terminal (i.e. pilot is transmitted by the base station and measured and estimated by the terminal), if the used entity device of the neural network model is not the base station or the terminal, it can receive the CIRs of at least two MIMO links in the MIMO system transmitted by the base station or the terminal, and thereby determine the data. If the entity device using the neural network model is a base station (or a terminal), it may receive CIRs of at least two MIMO links in the MIMO system transmitted by the terminal (or the base station) to determine the data, or may directly measure and estimate the data to determine the data.

And step 102, channel prediction is carried out.

In this step, according to the CIRs of the at least two MIMO links, CIRs of other MIMO links in the MIMO system may be obtained by using a neural network model that is pre-established and corresponds to the MIMO system.

Considering that the training data of the neural network model is time-series data, the neural network model may employ a deep cyclic neural network model, for example, a deep long-short term memory network (L STM) model, to ensure the accuracy of channel prediction.

Furthermore, the deep cycle neural network model can also adopt a Bidirectional algorithm and/or a time-distributed algorithm to further improve the accuracy of channel prediction.

The neural network model can be obtained by training through the following method:

determining the CIR of each MIMO link in the MIMO system which is periodically collected as a training sample;

and training to obtain the neural network model by using the determined training sample.

It should be noted that the periodicity of the training samples is configurable, for example, the configuration period of the training samples may be determined comprehensively according to the channel variation speed and/or the storable data size (which may be understood as the storable data size of the training entity device of the neural network model), and the like. In addition, the amount of training sample data may be relatively large, so the data amount problem needs to be considered when the training sample data is stored, reported, and forwarded.

After the training samples are determined, each type of historical data in the training samples (i.e., the CIR for each MIMO link in the MIMO system) may be data-aligned. And performing model training by using historical data obtained by data alignment, and obtaining a trained neural network model through supervised machine learning.

Taking a2 × 2 MIMO system as an example, the channel diagram can be as shown in fig. 2. When the neural network model training is carried out, the CIRs on any three MIMO links can be used for predicting the CIRs on the rest MIMO links. For example, it is only necessary to know the MIMO link between H11 (corresponding to the transmitting-end Antenna 1(Tx Antenna1) and the receiving-end Antenna 1(Rx Antenna 1)), the CIR on H12 (corresponding to the MIMO link between the transmitting-end Antenna 1(Tx Antenna1) and the receiving-end Antenna 2(Rx Antenna 2)) and H21 (corresponding to the MIMO link between the transmitting-end Antenna 2(Tx Antenna2) and the receiving-end Antenna 1(Rx Antenna 1)), so that the CIR on H22 (corresponding to the MIMO link between the transmitting-end Antenna 2(Tx Antenna2) and the receiving-end Antenna 2(RxAntenna2) may be directly calculated, and therefore, it is not necessary to analyze the relevant pilot transmitted from the Tx Antenna2 at all at Rx Antenna2, thereby reducing the operation and improving the energy efficiency.

When the neural network model is trained, the CIRs on the remaining two MIMO links can be predicted through the CIRs on any two MIMO links, for example, the CIRs on H21 and H22 can be predicted only by knowing the CIRs on H11 and H12, so that pilot frequency information sent on Tx antenna2 can be omitted, time-frequency domain resources are saved, corresponding channel estimation operation is also omitted, and spectral efficiency and energy efficiency are improved.

Taking the prediction of CIRs on H21 and H22 from CIRs on H11 and H12 as an example, when performing neural network model training, the input data may be CIRs on H11 and H12, and the output data may be CIRs on H21 and H22.

Arranging each type of historical data in the training samples according to a time sequence, and assuming that each type of input data inputs M (M is a positive integer) at a time, and each type of output data outputs M at a time, performing data alignment, then a time sequence training sample that the neural network model can receive at a time can be represented as follows:

input (input): [ H11 ]_t,H11_t+1,…,H11_t+M-1And [ H12 ]_t,H12_t+1,…,H12_t+M-1】

Output (output): [ H21 ]_t,H21_t+1,…,H21_t+M-1And [ H22 ]_t,H22_t+1,…,H22_t+M-1】

The time series training samples that the neural network model can receive next time can be represented as follows:

input (input): [ H11 ]_t+1,H11_t+2,…,H11_t+MAnd [ H12 ]_t+1,H12_t+2,…,H12_t+M】

Output (output): [ H21 ]_t+1,H21_t+2,…,H21_t+MAnd [ H22 ]_t+1,H22_t+2,…,H22_t+M】

And so on.

After training, if the loss function reaches a set threshold, the loss function may be, but is not limited to, the minimum Mean Square Error (MSE) between the predicted value and the true value, then training of the neural network model may be considered to be completed, and channel prediction may be performed by using the neural network model.

When the neural network model is used for channel prediction, the CIRs on H21 and H22 can be obtained according to the determined CIRs on H11 and H12. Therefore, the CIR of each MIMO link in the complete MIMO system is obtained and can be used for subsequent physical layer processes such as channel coding and the like.

Of course, the cyclic arrangement of the CIRs of the partial MIMO links as model inputs during channel prediction is the same as the cyclic arrangement of the training samples during model training. And the period configuration of CIRs of other MIMO links in the MIMO system as model output is the same as that of training samples during model training.

It should be noted that the neural network model may be trained by a training entity device of the neural network model.

The training entity device of the neural network model may also be any entity device (which may be understood as any network node). Considering that the CIR of each MIMO link in the MIMO system can be directly measured and estimated by the base station (i.e. pilot is sent by the terminal and measured and estimated by the base station), or can be measured and estimated by the terminal (i.e. pilot is sent by the base station and measured and estimated by the terminal), if the training entity device of the neural network model is not the base station and the terminal, it can receive the CIR of each MIMO link in the MIMO system sent by the terminal or the base station, thereby determining the training sample and training the neural network model. And if the training entity device of the neural network model is a base station (or a terminal), the training entity device can receive the CIR of each MIMO link in the MIMO system sent by the terminal (or the base station), thereby determining a training sample and training the neural network model. Or directly measuring and estimating to obtain CIR of each MIMO link in the MIMO system, thereby determining a training sample and training the neural network model.

It should be further noted that, in the solution provided in the embodiment of the present invention, in order to save the computation power and resources of the entity device using the neural network model, the entity device using the neural network model and the entity device using the neural network model may be two different entity devices. Of course, the training entity device of the neural network model and the using entity device of the neural network model may be the same entity device, which is not limited in the embodiment of the present invention.

Preferably, during the execution of step 101 and step 102, this embodiment may further include step 103, and in fig. 1, for convenience of illustration, step 103 is written after step 102:

and 103, determining Key Performance Indicators (KPIs).

In this step, a terminal KPI obtained through real-time monitoring may be determined, and when the KPI does not satisfy a set condition, the CIR of each MIMO link in the MIMO system is determined by using an MIMO channel estimation method.

The KPI may be, but is not limited to, at least one of throughput, number of data streams (rank), modulation coding scheme (mcs).

It can be understood that the terminal KPI can also be monitored in real time during the channel prediction process. If the performance of the MIMO system is found to be deteriorated according to the terminal KPI, the existing MIMO channel estimation method can be returned to determine the CIR of each MIMO link in the MIMO system, so as to ensure the reliability of the system.

It can be understood that if the performance of the MIMO system suddenly deteriorates, the conventional MIMO channel estimation method can be reverted to, and if the performance of the MIMO system does not change drastically, the channel prediction using the pre-established neural network model can be maintained.

It should be further noted that, before step 101, the method may further include step 100:

step 100, triggering the optimization of the pre-established neural network model.

Because the neural network model is pre-established, when the pre-established neural network model is used for channel prediction, the wireless environment may change relative to the time of establishing the neural network model, and therefore, preferably, the pre-established neural network model can be optimized before the pre-established neural network model is used, so that the optimized neural network model can be more suitable for the current wireless environment. Therefore, in this step, optimization of the pre-established neural network model corresponding to the MIMO system may be triggered. It is understood that the use of a physical device of the neural network model may trigger the optimization of the neural network model.

The process of optimizing the pre-established neural network model is understood to be the same as the training process of the neural network model, but with the real-time channel CIR as input, ensuring that the channel is not outdated.

Specifically, the CIR of each MIMO link in the MIMO system that is periodically collected may be determined as an optimization sample; and optimizing the pre-established neural network model by using the determined optimization sample.

If step 100 is performed, in step 102, CIRs of other MIMO links in the MIMO system may be obtained by using the neural network model corresponding to the MIMO system obtained after optimization according to the CIRs of the at least two MIMO links.

It should be further noted that before step 101 (if step 100 is included, it is understood that before step 100), the method may further include step 100':

step 100', it is determined whether the corresponding neural network model has been previously established.

It can be understood that, if it is required to perform channel prediction by using a pre-established neural network model, in this embodiment, it may also be determined whether the corresponding neural network model is already pre-established.

If it is determined that the neural network model corresponding to the MIMO system has been previously established, step 101 may be performed (of course, if step 100 is included, step 100 may be understood to be performed first).

If it is determined that the neural network model corresponding to the MIMO system is not pre-established, then the neural network model corresponding to the MIMO system may be triggered to be established, and step 101 may be executed after the neural network model corresponding to the MIMO system is obtained (of course, if step 100 is included, step 100 may be executed first).

According to the scheme provided by the first embodiment of the invention, a machine-learned RNN model can be adopted according to historical channel measurement of the MIMO system, and a channel prediction model can be learned in a supervised manner according to a specific time sequence and a preparation method of training data. The pilot frequency can be reduced on partial antennas, or the pilot frequency can be reduced, so that the method is simplified relative to the existing real-time channel estimation algorithm.

In addition, the channel prediction scheme provided by the first embodiment of the invention can reduce the inaccuracy factor brought by using the traditional statistical model, can reduce the operation complexity of using the deterministic model, and can make the prediction more accurate because of considering the user characteristics and the real transmission environment.

The channel prediction scheme provided by the first embodiment of the invention can be applied to real transmission, and can reduce pilot frequency overhead compared with the existing scheme for determining the CIR of the MIMO link based on channel estimation, namely, pilot frequency can be sent on part of antennas and part of time-frequency resources.

The scheme provided by the first embodiment of the invention is explained by a specific example.

Example two

The second embodiment of the present invention provides a method for predicting a channel of an MIMO system, which takes the CIR of each MIMO link of the MIMO system obtained through measurement and estimation by a base station to implement model training and optimization, and the CIR of a part of MIMO links of the MIMO system obtained through measurement and estimation by the base station to implement channel prediction as an example, and the flow of the steps of the method may be as shown in fig. 3, and includes:

step 201, determining whether a corresponding model exists.

In this step, the entity using the neural network model may determine whether a corresponding neural network model exists.

It can be understood that, after the terminal accesses the base station, the base station may notify the use entity of the neural network model, and the use entity of the neural network model may determine whether the neural network model corresponding to the use entity exists. Specifically, the entity using the neural network model may search, in the memory, whether a neural network model corresponding to the MIMO system formed by the terminal and the base station exists.

In a possible implementation manner, it may be understood that, if a terminal accesses a base station for the first time, the terminal may send pilot to the base station on all antennas, the base station may measure and estimate the CIR of each MIMO link in the MIMO system, and may send the measured and estimated data to a training entity of a neural network model, which is trained to obtain a neural network model corresponding to the MIMO system formed by the terminal and the base station, and may store the trained neural network model in a use entity of the neural network model. Thus, when the terminal accesses the base station again, the using entity of the neural network model already stores the corresponding neural network model.

That is, if a terminal first accesses a base station, the use entity of the neural network model does not store the corresponding neural network model, and if a terminal does not first access a base station, the use entity of the neural network model stores the corresponding neural network model.

If it is determined that the corresponding neural network model has been stored, a jump is made to step 203, otherwise step 202 is performed.

Step 202, triggering neural network model training.

If the using entity of the neural network model judges that the neural network model corresponding to the using entity does not exist, the base station can be informed to collect the training sample. The base station can receive the pilot frequency sent by the terminal, measure and estimate the CIR of each MIMO link of the MIMO system, and send the measured and estimated part of data to the training entity equipment of the neural network model for model training.

After the training of the model is completed, the neural network model obtained by offline training may be stored in the entity device using the neural network model, and the step 203 may be continued.

And step 203, triggering an optimized neural network model.

In view of the change of the wireless environment, in this embodiment, before the pre-established neural network model is used, the pre-established neural network model is optimized online, so that the optimized neural network model can be subsequently used for channel prediction, and the accuracy of channel prediction is further ensured.

In this step, the entity using the neural network model may inform the base station to collect the optimized samples. The base station can receive the pilot frequency sent by the terminal, measure and estimate the CIR of each MIMO link of the MIMO system, and send the measured and estimated part of data to the training entity equipment of the neural network model for model online optimization.

After the online optimization of the model is completed, the neural network model obtained after the optimization can be stored in the use entity device of the neural network model.

Of course, it can be understood that the training entity device of the neural network model may also store the neural network model trained in advance, and after the model is optimized on line, the stored neural network model may be updated to the neural network model obtained after optimization. And the optimized neural network model stored by the neural network model self can be forwarded to the entity equipment using the neural network model for storage, so that the entity equipment can be used for channel prediction.

And step 204, determining the CIR of the partial MIMO link.

In this step, channel impulse responses CIR of at least two MIMO links in the corresponding MIMO system may be determined.

Specifically, in this embodiment, the entity using the neural network model may receive channel impulse responses CIR of at least two MIMO links in the MIMO system sent by the base station.

And step 205, channel prediction is carried out.

In this step, the entity using the neural network model may obtain CIRs of other MIMO links in the MIMO system by using the online optimized neural network model according to the determined CIRs of part of the MIMO links.

It should be noted that, in the process of channel prediction, the terminal KPI may also be monitored in real time. In this embodiment, the entity using the neural network model may determine the terminal KPI obtained through real-time monitoring, and if it is determined that the KPI does not satisfy the set condition, the channel prediction using the neural network model may be stopped, and the existing MIMO channel estimation method may be reverted to determine the CIR of each MIMO link in the MIMO system, that is, the CIR of each MIMO link in the corresponding MIMO system may be determined through the base station.

According to the scheme provided by the second embodiment of the invention, whether a corresponding neural network model exists can be determined by judging whether a terminal is accessed to a base station for the first time and can be understood as whether the terminal is a new user or an old user of the base station, and then different processes are respectively adopted for the new user and the old user to realize channel prediction.

When training a neural network model corresponding to the MIMO system, the CIR of each MIMO link in the history of the MIMO system may be used as training data, and the training data has a time sequence and is related to the propagation environment of the terminal, and the corresponding neural network model is obtained through training.

When the neural network model is trained, an off-line training mode can be adopted, and whether the CIR of the MIMO link is determined by returning to the existing MIMO channel estimation mode can be determined by monitoring the terminal KPI mode in real time in the use process of the neural network model obtained by training.

In addition, according to the scheme provided by the second embodiment of the invention, the neural network model obtained by offline training in advance can be used for channel prediction after being optimized on line, so that the optimized neural network model is more suitable for the current wireless environment, and the accuracy of channel prediction is further ensured.

Fig. 4 is a schematic diagram of a simulation result of channel prediction of a MIMO system, and as shown in fig. 4, taking the CIR on H22 obtained according to the CIRs on H11, H12, and H21 in the channel schematic diagram shown in fig. 2 as an example, under the same normalized time delay (normalseddelay), the difference between the predicted value and the true value (both expressed by normalized amplitude) of the CIR on H22 is smaller, and the prediction accuracy is higher. The difference between the real part (prediction-real) of the predicted value and the real part (original-real) of the real value, and the difference between the imaginary part (prediction-image) of the predicted value and the imaginary part (original-image) of the real value can be respectively shown in fig. 4, and it can be seen that the differences are small, and the prediction accuracy is high.

In the simulation process, a Recurrent Neural Network (RNN) model can be used, specifically, an L STM unit (cell) can be adopted to construct a deep neural network, a L STM-based deep neural network model is used, a bidirectional algorithm and a time-distributed algorithm can be used, and trained input data, intermediate variables and predicted output data can be circulated in the model in a forward direction and a reverse direction, so that the relevance among the data can be found more favorably, and the convergence of the model and the performance of the model can be improved favorably.

As shown in FIG. 5, the hidden layer in the Input layer (Input L eye) and the Output layer (Output L eye) has a forward layer (formed L eye) and a reverse layer (backed L eye) by adopting a bidirectionality algorithm, namely, the forward layer inputs and outputs are sequential, for example, if the Input 123 and the data received by the forward layer are also 123, the data received by the reverse layer becomes 321, so that the trained model has the characteristic of identifying data in different directions.

That is, the core of the bidirectional algorithm is to train input data and output data not only in the order of feeding the models, but also in the order of reversing, and the iteration is bidirectional, and the convergence of the models is also bidirectional and effective, so that the time correlation between the data can be better learned.

Further, with respect to the TimeDistributed algorithm, in fact, the TimeDistributed layer gives the model the capability of one-to-many (one to many), many-to-one (many to one), and many-to-many (many to many), rather than just one-to-one (one to one) modes, increasing the dimensionality of the model. The model diagram using the time-distributed algorithm can be shown in fig. 6, where the middle layer (shaded module) of the model is no longer only weight-adjustable, but each module has a corresponding output value, which can be used to simulate given historical output data. Therefore, the model can better accord with the trained data, and the prediction precision is higher.

The same inventive concept as in the first and second embodiments provides the following apparatuses.

EXAMPLE III

A third embodiment of the present invention provides a MIMO system channel prediction apparatus, where the structure of the apparatus may be as shown in fig. 7, and the apparatus includes:

the determining module 11 is configured to determine channel impulse responses CIR of at least two MIMO links in the MIMO system;

the prediction module 12 is configured to obtain CIRs of other MIMO links in the MIMO system according to the CIRs of the at least two MIMO links by using a neural network model that is pre-established and corresponds to the MIMO system.

The determining module 11 may also be configured to determine a terminal key performance indicator KPI obtained through real-time monitoring;

the prediction module 12 may be further configured to determine a CIR of each MIMO link in the MIMO system by using a MIMO channel estimation method when the KPI does not satisfy a set condition.

The determining module 11 may be further configured to trigger optimization of a pre-established neural network model corresponding to the MIMO system before determining channel impulse responses CIR of at least two MIMO links in the MIMO system;

the prediction module 12 is specifically configured to obtain CIRs of other MIMO links in the MIMO system by using the optimized neural network model corresponding to the MIMO system according to the CIRs of the at least two MIMO links.

The determining module 11 may be further configured to determine that a neural network model corresponding to the MIMO system is established in advance before determining channel impulse responses CIR of at least two MIMO links in the MIMO system.

Based on the same inventive concept, embodiments of the present invention provide the following apparatus and medium.

Example four

A fourth embodiment of the present invention provides a MIMO system channel prediction apparatus, where the structure of the apparatus may be as shown in fig. 8, and the apparatus includes a memory 21, a processor 22, a transceiver 23, and a bus interface; the processor 22 is configured to read the program in the memory 21, and execute:

determining, by the transceiver 23, channel impulse responses, CIRs, of at least two MIMO links in the MIMO system; and obtaining the CIRs of other MIMO links in the MIMO system by utilizing a pre-established neural network model corresponding to the MIMO system according to the CIRs of the at least two MIMO links.

Optionally, the processor 22 may specifically include a Central Processing Unit (CPU), an Application Specific Integrated Circuit (ASIC), one or more integrated circuits for controlling program execution, a hardware circuit developed by using a Field Programmable Gate Array (FPGA), or a baseband processor.

Optionally, the processor 22 may include at least one processing core.

Alternatively, the memory 21 may include a Read Only Memory (ROM), a Random Access Memory (RAM), and a disk memory. The memory 21 is used for storing data required by the at least one processor 22 during operation. The number of the memory 21 may be one or more.

A fifth embodiment of the present invention provides a nonvolatile computer storage medium, where the computer storage medium stores an executable program, and when the executable program is executed by a processor, the method provided in the first embodiment of the present invention is implemented.

In particular implementations, computer storage media may include: various storage media capable of storing program codes, such as a Universal Serial Bus flash drive (USB), a mobile hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.

In the embodiments of the present invention, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described embodiments of the apparatus are merely illustrative, and for example, the described unit or division of units is only one division of logical functions, and there may be other divisions when actually implemented, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical or other form.

The functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be an independent physical module.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, all or part of the technical solutions of the embodiments of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device, such as a personal computer, a server, or a network device, or a processor (processor) to execute all or part of the steps of the methods according to the embodiments of the present invention. And the aforementioned storage medium includes: various media that can store program codes, such as a universal serial bus flash drive (usb flash drive), a removable hard disk, a ROM, a RAM, a magnetic disk, or an optical disk.

As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, 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, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

While preferred embodiments of the present invention 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 such alterations and modifications as fall within the scope of the invention.

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 channel prediction in a MIMO system, the method comprising:

2. The method of claim 1, wherein the method further comprises: and determining a terminal key performance indicator KPI obtained by real-time monitoring, and determining the CIR of each MIMO link in the MIMO system by using an MIMO channel estimation method when the KPI does not meet set conditions.

3. The method of claim 1, wherein prior to determining Channel Impulse Responses (CIRs) for at least two MIMO links in the MIMO system, the method further comprises:

triggering and optimizing a pre-established neural network model corresponding to the MIMO system;

obtaining the CIRs of other MIMO links in the MIMO system by utilizing a pre-established neural network model corresponding to the MIMO system according to the CIRs of the at least two MIMO links, including:

and obtaining the CIRs of other MIMO links in the MIMO system by using the optimized neural network model corresponding to the MIMO system according to the CIRs of the at least two MIMO links.

4. The method of claim 1, wherein prior to determining Channel Impulse Responses (CIRs) for at least two MIMO links in the MIMO system, the method further comprises:

and determining that a neural network model corresponding to the MIMO system is established in advance.

5. The method of claim 1, in which the neural network model employs a deep-cycle neural network model.

6. The method of claim 5, in which the deep recurrent neural network model employs a Bidirectional algorithm and/or a time-distributed algorithm.

7. The method of any one of claims 1 to 6, wherein the neural network model is trained by:

8. An apparatus for channel prediction in a MIMO system, the apparatus comprising:

9. A non-transitory computer storage medium storing an executable program for execution by a processor to perform the steps of the method of any one of claims 1 to 7.

10. A MIMO system channel prediction device is characterized by comprising a memory, a processor, a transceiver and a bus interface; the processor is used for reading the program in the memory and executing: