IRS communication system channel estimation method, electronic equipment and medium

Abstract

The invention discloses an IRS communication system channel estimation method, electronic equipment and a medium. The method may comprise: constructing an IRS assisted wireless communication system; a communication process is divided into a channel estimation stage, a beam forming stage and a data transmission stage, and a channel estimation protocol is established; a channel estimation stage is divided into an offline training stage and an online application stage, and channel estimation calculation is carried out. According to the method and the device, the adaptability and the accuracy of IRS-assisted wireless communication environment channel estimation are enhanced.

Description

Technical Field

[0001] The present invention relates to the field of wireless communications, and more specifically, to an IRS communication system channel estimation method, electronic equipment and medium. Background Art

[0002] In recent years, deep learning architectures have been widely used in IRS-assisted wireless communication systems. In these systems, they are used to study technologies such as beamforming, signal positioning, and channel estimation. A relatively simple one is channel estimation for assisting a single user in wireless communication. Among them, Mishra proposed a reflection method in which the IRS turns on one unit in sequence and ensures that all other reflection units are turned off to obtain the reflection channel through the reflection unit. The BS estimates the cascade channel in sequence, which is the product of the channel information matrix from the BS to the reflection unit and the channel information matrix from the reflection unit to the user. Jensen TI, all reflection units are turned on during training, and the minimum variance unbiased estimator is used for estimation. Among them, the optimal phase shift matrix of the IRS is proved to be the discrete Fourier transform matrix. Zhang Rui's team assumed that the entire reflection surface can be divided into multiple groups, each group consists of a number of adjacent units, and the units in the group are set to the same reflection coefficient. However, these studies all face a common problem, that is, the training overhead increases with the increase in the number of reflection units, which leads to excessive training overhead and reduced spectral efficiency. In Taha A, a method for channel estimation is proposed by deploying some active units on the reflection surface. Due to the sparsity of the channel, active units are used to obtain the complete CSI. This approach reduces the training overhead but increases the hardware cost.

[0003] In the millimeter wave Massive MIMO communication system, researchers such as Elbir AM proposed a new channel estimation method based on deep learning. This method uses two CNN networks with the same structure to estimate the channel information of the direct link and the reflected link respectively. Compared with the traditional least squares (LS) method, this method has higher performance. In the IRS-assisted wireless communication scenario, researchers such as Saud Khan used deep neural networks as the front-end information processing part of the receiver to detect the signal of the communication link. Their research shows that this method can quickly adapt to the rapidly changing channel environment, and its error performance is similar to that of the maximum likelihood (ML) detector.

[0004] In addition, there are also channel estimation schemes designed for the IRS reflection mode. He ZQ designed a randomly turned on / off reflection mode based on the low rank of the MIMO channel. On this basis, he studied the cascade channel estimation method based on sparse matrix decomposition, which can reduce the pilot overhead. However, this method has major disadvantages. Not only does it require amplitude control of multiple devices separately, thereby increasing the cost of the equipment, but also because only a small number of units are turned on at a time, it is impossible to give full play to the advantage of quantity. This affects the accuracy of channel estimation. In response to the above problems, Zheng B proposed a channel estimation scheme based on the IRS full reflection mode, that is, the amplitude of each reflection surface unit is at the maximum state in both the channel estimation and information transmission stages.

[0005] Although the research in the above literature has proposed its own solution to the channel estimation problem in the IRS-assisted wireless communication scenario, there are still some shortcomings. Although the passive or active channel estimation method can obtain higher estimation accuracy under certain conditions, for the frequency division duplex (FDD) system, under this condition, the channel does not have reciprocity. FDD makes it difficult to collect uplink and downlink channel information, requiring more complex channel estimation protocols and more training overhead, which greatly reduces communication efficiency. In the time division duplex (TDD) communication system, although there is channel reciprocity, due to the multi-dimensional characteristics of IRS, the complexity of the channel information and the channel estimation algorithm itself requires a large overhead. Therefore, it is worthwhile to study the algorithm to reduce the communication overhead and algorithm complexity while ensuring accurate and reliable channel information.

[0006] Therefore, it is necessary to develop a channel estimation method, electronic equipment and medium for an IRS communication system.

[0007] The information disclosed in the background technology section of the present invention is only intended to deepen the understanding of the general background technology of the present invention, and should not be regarded as acknowledging or suggesting in any form that the information constitutes the prior art already known to those skilled in the art. Summary of the invention

[0008] The present invention proposes an IRS communication system channel estimation method, electronic equipment and medium, which can enhance the adaptability and improve the accuracy of IRS-assisted wireless communication environment channel estimation.

[0009] In a first aspect, an embodiment of the present disclosure provides an IRS communication system channel estimation method, including:

[0010] Construct IRS-assisted wireless communication system;

[0011] The communication process is divided into the channel estimation phase, the beamforming phase and the data transmission phase, and a channel estimation protocol is established;

[0012] The channel estimation stage is divided into an offline training stage and an online application stage to perform channel estimation calculation.

[0013] Preferably, an IRS is installed between a base station and a user equipment of the IRS-assisted wireless communication system, wherein the base station is configured with M antennas and the IRS is configured with N antennas. IRS reflection units, and the communication service area contains K single-antenna users.

[0014] Preferably, in the channel estimation stage, it is responsible for collecting channel information, providing a data set for training a channel estimation network and estimating channel state information CSI.

[0015] Preferably, in the beamforming stage shown, based on the estimated CSI, the receive beamforming vector of the BS and the reflection coefficient of the IRS are designed for data transmission, and then sent to the IRS controller through the control link.

[0016] Preferably, during the data transmission phase, the IRS controller sets a reflection coefficient to assist independent data transmission from the user to the BS.

[0017] Preferably, in the offline training stage, random channel data and its corresponding reflection matrix and receiving signal are generated through simulation according to input and output, and input into the neural network, so that the mapping result of the neural network gradually converges as the training proceeds.

[0018] Preferably, in the online application stage, the data is preprocessed to make it the input required by the neural network, and then input into the trained model to obtain the channel estimate through calculation.

[0019] Preferably, channel estimation of the IRS communication system is performed through CBDNet.

[0020] In a second aspect, an embodiment of the present disclosure further provides an electronic device, the electronic device comprising:

[0021] A memory storing executable instructions;

[0022] A processor runs the executable instructions in the memory to implement the IRS communication system channel estimation method.

[0023] In a third aspect, an embodiment of the present disclosure further provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the IRS communication system channel estimation method is implemented.

[0024] The methods and apparatus of the present invention have other features and advantages that will be apparent from, or will be described in detail in, the accompanying drawings and subsequent detailed descriptions incorporated herein, which together serve to explain the specific principles of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The above and other objects, features and advantages of the present invention will become more apparent through a more detailed description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, wherein like reference numerals generally represent like components throughout the exemplary embodiments of the present invention.

[0026] Figure 1 A flow chart showing the steps of a channel estimation method for an IRS communication system according to an embodiment of the present invention.

[0027] Figure 2 A schematic diagram of an IRS-assisted wireless communication system according to an embodiment of the present invention is shown.

[0028] Figure 3 A schematic diagram of a channel estimation protocol according to an embodiment of the present invention is shown.

[0029] Figure 4 A schematic diagram of a channel estimation stage according to an embodiment of the present invention is shown.

[0030] Figure 5 A schematic diagram of a CBDNet network structure according to an embodiment of the present invention is shown.

[0031] Figure 6 A schematic diagram showing a deep neural network for channel estimation according to an embodiment of the present invention is shown.

[0032] Figure 7 A schematic diagram of a channel estimation network training process according to an embodiment of the present invention is shown.

[0033] Figure 8 A schematic diagram showing comparison of reflection link channel estimation algorithms according to an embodiment of the present invention is shown.

[0034] Fig. 9 A schematic diagram showing a comparison of direct link channel estimation algorithms according to an embodiment of the present invention is shown.

[0035] Fig.10 A schematic diagram showing a comparison of direct link channel estimation algorithms according to an embodiment of the present invention is shown.

[0036] Fig.11A schematic diagram showing a comparison of cascade channel estimation performance under conditions of different numbers of reflection units according to an embodiment of the present invention is shown.

[0037] Fig.12a , Figure 12b Schematic diagrams respectively show the influence of the channel estimation accuracy of the LMMSE method and the CBDNet method on beamforming according to an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The preferred embodiments of the present invention will be described in more detail below. Although the preferred embodiments of the present invention are described below, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0039] Figure 1 A flow chart showing the steps of a channel estimation method for an IRS communication system according to an embodiment of the present invention.

[0040] like Figure 1 As shown, the IRS communication system channel estimation method includes:

[0041] Step 101, constructing an IRS-assisted wireless communication system;

[0042] Step 102, dividing the communication process into a channel estimation phase, a beamforming phase and a data transmission phase, and establishing a channel estimation protocol;

[0043] Step 103, dividing the channel estimation phase into an offline training phase and an online application phase, and performing channel estimation calculation.

[0044] In one example, an IRS is installed between a base station and a user equipment of an IRS-assisted wireless communication system, wherein the base station is configured with M antennas and the IRS is configured with N antennas. IRS reflection units, and the communication service area contains K single-antenna users.

[0045] In one example, in the channel estimation stage, it is responsible for collecting channel information, providing a data set for training a channel estimation network, and estimating channel state information CSI.

[0046] In one example, in the beamforming stage shown, based on the estimated CSI, the receive beamforming vector of the BS and the reflection coefficient of the IRS are designed for data transmission and then sent to the IRS controller through the control link.

[0047] In one example, during the data transmission phase, the IRS controller sets the reflection coefficient to facilitate independent data transmission from the user to the BS.

[0048] In one example, in the offline training stage, random channel data and its corresponding reflection matrix and receiving signal are generated through simulation based on input and output, and input into the neural network. As the training proceeds, the mapping result of the neural network gradually converges.

[0049] In one example, in the online application stage, the data is preprocessed to make it the input required by the neural network, and then input into the trained model to obtain the channel estimate through calculation.

[0050] In one example, channel estimation of the IRS communication system is performed via CBDNet.

[0051] Figure 2 A schematic diagram of an IRS-assisted wireless communication system according to an embodiment of the present invention is shown.

[0052] Specifically, Figure 2 As shown in FIG. 1 , an IRS-assisted wireless communication system is constructed, including a channel model and a system model. The system adopts a time division duplex protocol, and the demand for downlink channel information can be obtained by estimating uplink channel information. An IRS is installed between the base station and the user equipment, where the base station is configured with M antennas and the IRS is configured with N IRS reflection units, and the communication service area contains K single-antenna users. Figure 2 It can be seen that the uplink channel of the IRS-assisted wireless communication system has two main communication links. One is the unobstructed direct link from the kth user to the BS, and the channel is represented by d k , the second is the reflection link from the kth user to the BS via the IRS, which consists of three main parts: the link from the user equipment to the IRS, k , the reflection matrix of the IRS reflection array and the link G from IRS to BS, the part containing channel information is f k and G.

[0053] Figure 3 A schematic diagram of a channel estimation protocol according to an embodiment of the present invention is shown.

[0054] like Figure 3 As shown in Figure 1, the channel estimation protocol includes signal feature analysis, protocol framework design and channel estimation process. The user received signal includes three parts: direct link signal, reflected link signal and noise. The reflected link channel information is a cascade channel that obeys double Rayleigh. The communication process is divided into three main stages: channel estimation stage, beamforming stage and data transmission stage.

[0055] The channel estimation stage is responsible for collecting channel information, providing a data set for training the channel estimation network and estimating the channel state information CSI.

[0056] In the beamforming phase, based on the estimated CSI, the BS’s receive beamforming vector and the IRS’s reflection coefficient are designed for data transmission and then sent to the IRS controller via the control link. It should be noted that the estimated CSI can also be used to address issues such as physical layer security.

[0057] During the data transmission phase, the IRS controller sets the reflection coefficient accordingly to assist the independent data transmission from the user to the BS.

[0058] The channel estimation technology is based on the channel estimation method of training symbols. At the transmitting end, in addition to transmitting data symbols, a preamble or pilot signal is also transmitted, and the pilot symbols of each user are mutually orthogonal. The present invention adopts the algorithm idea of ​​supervised learning and deploys a channel estimator based on a convolutional neural network in the IRS-assisted MISO wireless communication system to replace the traditional channel estimation algorithm.

[0059] Figure 4 A schematic diagram of a channel estimation stage according to an embodiment of the present invention is shown.

[0060] like Figure 4 As shown in the figure, the entire channel estimation stage is divided into two processes: offline training and online application. In the offline training stage, based on the input and output, a large amount of random channel data and its corresponding reflection matrix and received signal are generated through simulation, and then input into the neural network. As the training proceeds, the mapping results of the neural network gradually converge. In the online application stage, the data is first preprocessed to make it the input required by the neural network, and then it is input into the trained model. The channel estimate is obtained through calculation, and finally it is applied to subsequent beamforming and prevention of physical layer eavesdropping.

[0061] Figure 5 A schematic diagram of a CBDNet network structure according to an embodiment of the present invention is shown.

[0062] CBDNet is used for channel estimation in IRS communication systems. In CBDNet, a fully convolutional network is added to learn the noise intensity to achieve the purpose of adaptive noise. CBDNet has a downsampling function. A reversible downsampling operator is introduced in the network to reshape the input matrix of size W×H×C into 4 downsampling sub-matrices of size W / 2×H / 2×C. Here C is the number of channels, that is, C=1 for grayscale images and C=3 for color images. The processed data is a two-channel channel information matrix, so C=2, such as Figure 5 shown.

[0063] Using the algorithm idea of ​​supervised learning, a channel estimator based on convolutional neural network is deployed in the wireless communication system of IRS assisted MISO to replace the traditional channel estimation algorithm. This section introduces the process of channel estimation based on deep neural network. Usually, the interference-containing channel information H^ is used as training data, and the pre-stored part of the real channel information H is used as the training label, and the mean square error of the two is used as the loss function. The deep neural network model obtained by training is an estimator with good channel estimation performance that adapts to the current environment. New interference-containing channel information is input to obtain estimated channel information data.

[0064] The above model is a relatively simple case. By designing the structure of the neural network, a neural network that is more suitable for training a channel information estimator can be made, such as a channel estimator using a convolutional neural network.

[0065] Figure 6 A schematic diagram showing a deep neural network for channel estimation according to an embodiment of the present invention is shown.

[0066] For the channel estimation problem in IRS-assisted wireless communication system, the deep residual network based on convolutional neural network can achieve better results. The feature extraction function of convolutional neural network is used to extract the noise information in the channel information matrix, and the extracted noise is subtracted by residual structure. After repeated multiple times, the loss function is calculated by combining the obtained result with the label information. The network is trained by reducing the loss function, and the channel estimation function of the network is finally realized, such as Figure 6 shown.

[0067] Deep residual networks have good results in denoising, but their flexibility is limited. The reason why convolutional neural networks have been successful in image denoising is because of their powerful modeling capabilities, network training capabilities, and design advances. However, denoising methods based solely on CNNs have limited flexibility, and the learned models are usually customized for specific noise levels. It aims to learn a mapping function, and the trained model is difficult to directly apply to images with other noise levels. It cannot be well generalized to noisy images in real environments, and it can only play a denoising role when the noise level is within a preset range (for example, [0,55]). The models learned by deep residual networks based on CNNs are usually customized for specific noise levels. When the channel information is affected by a large level of noise, there will be a serious decrease in estimation accuracy.

[0068] In this network, the fully convolutional network used to generate the noise level map contains five convolutional layers. The noise level map output by it and the original channel information matrix containing interference are used as the input of the subsequent denoising network. The denoising subnetwork contains a residual structure. The network also contains a downsampling structure for improving the training speed. The original high-dimensional data can be split according to certain rules and trained separately, thereby speeding up the training speed. The output end is output after restoring the original data dimension through upsampling.

[0069] Figure 7 A schematic diagram of a channel estimation network training process according to an embodiment of the present invention is shown.

[0070] The entire training and testing process is divided into two main parts, such as Figure 7 As shown, in the training phase, the link budget and reflection matrix are used to obtain H LS . Combine the label data and the H LS Input into the neural network and calculate the mean square error of the two. Finally, the error value is used as the loss function to train the network, and the weights and offset parameters in the network are continuously updated until the end. After the network training is completed, the test data set is input to obtain the mean square error of the channel information estimation of the network as a result reflecting the accuracy of the channel estimation.

[0071] The success of a neural network denoiser depends largely on whether the distribution of the synthetic noise and the real noise are well matched. Therefore, the real noise model is the primary problem for blind channel estimation of real channel information. CBDNet was originally used for image denoising. There is a clear difference between image noise and channel information noise. In images, the Poisson-Gaussian distribution is considered to be a more realistic model of the original noise than AWGN, and its noise characteristics are related to the processing in the camera. The noise of the channel information is still Gaussian noise, but due to the cascade multiplication operation, the noise characteristics are closer to reality than general channels.

[0072] When removing image noise, unimportant details can be lost for the purpose of enhancing its display effect. When removing the noise of channel information, the purpose is to approximate the real channel information. CBDNet is more suitable for denoising the real environment noise channel information. In this chapter, the cascade channel is the multiplication of each component channel, and the noise characteristics are more hidden and complex in the cascade channel. CBDNet is very suitable for removing this kind of complex noise in the real environment. Secondly, by combining the channel information label and the real noise channel information in training, the denoising performance of the network can be improved. In addition, by introducing the noise estimation subnetwork in CBDNet, it is possible to use asymmetric loss to improve its generalization ability for real noise and facilitate interactive denoising.

[0073] The present invention also provides an electronic device, which includes: a memory storing executable instructions; and a processor, which runs the executable instructions in the memory to implement the above-mentioned IRS communication system channel estimation method.

[0074] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned IRS communication system channel estimation method is implemented.

[0075] To facilitate understanding of the solutions and effects of the embodiments of the present invention, three specific application examples are given below. Those skilled in the art should understand that the examples are only for facilitating understanding of the present invention, and any specific details thereof are not intended to limit the present invention in any way.

[0076] Example 1

[0077] Figure 8 A schematic diagram showing comparison of reflection link channel estimation algorithms according to an embodiment of the present invention is shown.

[0078] Figure 8 The performance of reflection link channel estimation of CBDNet, CDRNet and traditional algorithms under different signal-to-noise ratios is shown. As can be seen from the result graph, the performance of the network model is significantly better than that of the LS method and the LMMSE method. And it is better than the channel estimation performance of CDRNet, especially in the case of large environmental noise. CBDNet can play a better channel estimation performance due to the special structure of the noise level map. The reason is that there is a difference in dimension between the reflection channel and the direct channel. The former is M×N, while the latter is M. The reflection surface increases the dimension of the original channel information and enhances the spatial characteristics. After adding a reflection surface to the scene, the cascade characteristics of the reflection channel will make the large-scale fading of the signal more serious, resulting in deep fading. In the small-scale fading, the original Rayleigh channel becomes a cascade channel that obeys the double Rayleigh product. The change in channel information will be directly reflected in the characteristics of the signal. The deep fading of the cascade channel causes the signal-to-noise ratio to decrease, and the amplitude of each element signal in the cascade channel becomes the product of the random distribution variable obeying the double Rayleigh and the original signal. Such changes bring more nonlinear and spatial characteristics to the channel information, and the increase in the signal-to-noise ratio also makes it difficult to estimate the channel information.

[0079] Fig. 9 A schematic diagram showing a comparison of direct link channel estimation algorithms according to an embodiment of the present invention is shown.

[0080] Fig. 9The estimation results of the direct channel reflect that due to the low-dimensional characteristics of the direct channel, its channel information is only a linear superposition of the perfect channel information and the environmental noise, and the noise characteristics are not complex. The CBDNet method and the CDRNet method based on the convolutional neural network cannot fully exert their ability to extract and remove characteristic noise, so their performance is basically equivalent to that of the LMMSE, but the complexity of their algorithms is lower.

[0081] Fig.10 A schematic diagram showing a comparison of direct link channel estimation algorithms according to an embodiment of the present invention is shown.

[0082] Fig.10 The difference in convergence between the two convolutional neural network-based algorithms when the signal-to-noise ratio is 5dB is shown. Since CBDNet contains a convolutional network that extracts the noise level before the data enters the denoising network, it can effectively estimate the noise level of the data. The estimated noise level map is used as input together with the original data, so a better estimation effect can be achieved. And because there is a downsampling structure in CBDNet, the original data matrix can be decomposed and calculated, which improves the efficiency of calculation and training, so its convergence speed is also better than CDRNet.

[0083] Fig.11 A schematic diagram showing a comparison of cascade channel estimation performance under conditions of different numbers of reflection units according to an embodiment of the present invention is shown.

[0084] Fig.11 The results show the channel estimation performance comparison of the cascaded channel obtained by training with the same configuration of CBDNet under different numbers of units on the reflector. It can be seen that in the case of low signal-to-noise ratio, that is, when the noise is strong, the estimation performance of the algorithm network for smart reflectors of different sizes is quite different. As the number of units increases, the accuracy of channel estimation begins to decrease, and this difference decreases as the signal-to-noise ratio increases. The reason for this phenomenon is that the increase in the number of units causes the dimension of the channel information matrix to increase, which puts a certain pressure on the noise level extraction of the noise level network and the mapping ability of the denoising network.

[0085] Fig.12a , Figure 12b Schematic diagrams respectively show the influence of the channel estimation accuracy of the LMMSE method and the CBDNet method on beamforming according to an embodiment of the present invention.

[0086] Fig.12a , Figure 12bThe influence of channel information obtained by LMMSE and CBDNet channel estimation methods on beam optimization is demonstrated. The accuracy of channel information obtained by CBDNet algorithm is higher than that of LMMSE method. By observing the coverage and peak value of the received signal of the first user in the scene, it can be seen that the channel information obtained by CBDNet method can optimize a better beam. The high accuracy of channel information can make the beam of reflected signal more concentrated. Better directivity will provide users with better communication quality.

[0087] Example 2

[0088] The present disclosure provides an electronic device, which includes: a memory storing executable instructions; a processor, which runs the executable instructions in the memory to implement the above-mentioned IRS communication system channel estimation method.

[0089] An electronic device according to an embodiment of the present disclosure includes a memory and a processor.

[0090] The memory is used to store non-temporary computer-readable instructions. Specifically, the memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM) and / or cache memory (cache), etc. The non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, etc.

[0091] The processor may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions. In one embodiment of the present disclosure, the processor is used to run the computer-readable instructions stored in the memory.

[0092] Those skilled in the art should be able to understand that in order to solve the technical problem of how to obtain a good user experience, the present embodiment may also include well-known structures such as a communication bus and an interface, and these well-known structures should also be included in the protection scope of the present disclosure.

[0093] For detailed description of this embodiment, reference may be made to the corresponding descriptions in the aforementioned embodiments, which will not be repeated here.

[0094] Example 3

[0095] An embodiment of the present disclosure provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the IRS communication system channel estimation method is implemented.

[0096] According to the computer-readable storage medium of the embodiment of the present disclosure, non-transitory computer-readable instructions are stored thereon. When the non-transitory computer-readable instructions are executed by a processor, all or part of the steps of the above-mentioned methods of each embodiment of the present disclosure are executed.

[0097] The above-mentioned computer-readable storage media include, but are not limited to: optical storage media (e.g., CD-ROM and DVD), magneto-optical storage media (e.g., MO), magnetic storage media (e.g., magnetic tape or mobile hard disk), media with built-in rewritable non-volatile memory (e.g., memory card) and media with built-in ROM (e.g., ROM box).

[0098] Those skilled in the art should understand that the purpose of the above description of the embodiments of the present invention is only to exemplarily illustrate the beneficial effects of the embodiments of the present invention, and is not intended to limit the embodiments of the present invention to any given examples.

[0099] The embodiments of the present invention have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments.

Claims

1. A channel estimation method for an IRS communication system, characterized in that: include: Construct IRS-assisted wireless communication system; The communication process is divided into the channel estimation phase, the beamforming phase and the data transmission phase, and a channel estimation protocol is established; The channel estimation stage is divided into an offline training stage and an online application stage to perform channel estimation calculation.

2. The IRS communication system channel estimation method according to claim 1, wherein: The IRS is installed between the base station and the user equipment of the IRS-assisted wireless communication system, wherein the base station is equipped with M antennas and the IRS is equipped with N IRS reflection units, and the communication service area contains K single-antenna users.

3. The IRS communication system channel estimation method according to claim 1, wherein: In the channel estimation stage, it is responsible for collecting channel information, providing a data set for training the channel estimation network and estimating the channel state information CSI.

4. The IRS communication system channel estimation method according to claim 1, wherein: In the beamforming stage shown, based on the estimated CSI, the BS's receive beamforming vector and the IRS's reflection coefficient are designed for data transmission and then sent to the IRS controller through the control link.

5. The IRS communication system channel estimation method according to claim 1, wherein: During the data transmission phase, the IRS controller sets the reflection coefficient to facilitate independent data transmission from the user to the BS.

6. The IRS communication system channel estimation method according to claim 1, wherein: In the offline training stage, random channel data and its corresponding reflection matrix and receiving signal are generated through simulation according to input and output, and input into the neural network. As the training proceeds, the mapping result of the neural network gradually converges.

7. The IRS communication system channel estimation method according to claim 1, wherein: In the online application stage, the data is preprocessed to make it the input required by the neural network, and then input into the trained model to obtain the estimated amount of the channel through calculation.

8. The IRS communication system channel estimation method according to claim 1, wherein: The channel estimation of the IRS communication system is performed through CBDNet.

9. An electronic device, characterized in that: The electronic device comprises: A memory storing executable instructions; A processor, wherein the processor runs the executable instructions in the memory to implement the IRS communication system channel estimation method according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the IRS communication system channel estimation method according to any one of claims 1 to 8 is implemented.

Images