Information transmission method and device, related equipment and storage medium

Abstract

The invention discloses an information transmission method and device, a terminal, network equipment and a storage medium. The method comprises: a terminal receiving first information and second information sent by a network device, the first information being used for indicating a first model and a second model, and the second information being used for indicating channel compression and feedback requirements; testing the performance of the first model and the second model, and under the condition that the performance of the first model and the second model meets the requirements of channel compression and feedback, performing channel compression by using the first model to obtain a plurality of output elements; and determining fourth information by using the plurality of output elements, a modulation and coding scheme (MCS) corresponding to the first model and third information, and sending the fourth information to the network device.

Description

Technical Field

[0001] This application relates to the field of wireless communications, and in particular, to an information transmission method, apparatus, related device, and storage medium. Background Art

[0002] In a Multiple-Input Multiple-Output (MIMO) system, obtaining Channel State Information (CSI) is a key condition for a base station to perform beamforming to improve transmission performance. For a Frequency Division Duplexing (FDD) system, due to the lack of reciprocity of the complete uplink and downlink channels, the base station needs to obtain the complete downlink CSI by means of terminal feedback. Among them, the terminal can perform channel compression and feedback the channel compression result to the base station, and the base station can recover the channel compression result to obtain the state of the original channel.

[0003] However, in related technologies, resource waste may occur when the terminal performs channel compression and feedback. Summary of the Invention

[0004] To solve the problems in related technologies, embodiments of this application provide an information transmission method, apparatus, related device, and storage medium.

[0005] The technical solution of the embodiments of this application is implemented as follows:

[0006] Embodiments of this application provide an information transmission method applied to a terminal, including:

[0007] Receiving a first piece of information and a second piece of information sent by a network device, where the first piece of information is used to indicate a first model and a second model, the first model is used to perform channel compression, the second model is used to perform channel decompression, and the second piece of information is used to indicate the requirements for channel compression and feedback;

[0008] Testing the performance of the first model and the second model, and when the performance of the first model and the second model meets the requirements for channel compression and feedback, using the first model to perform channel compression to obtain a plurality of output elements; determining a fourth piece of information by using the plurality of output elements, a Modulation and Coding Scheme (MCS) corresponding to the first model, and a third piece of information, and sending the fourth piece of information to the network device; where

[0009] The total number of constellation points corresponding to the MCS is the same as the total number of the multiple output elements. The third information characterizes the association relationship between the output elements of the first model and the constellation points corresponding to the MCS. The fourth information includes the constellation points corresponding to each of the multiple output elements.

[0010] In the above solution, the method further includes:

[0011] Using fifth information to determine the MCS and the third information. The fifth information includes multiple domains and sixth information corresponding to each domain. The sixth information includes multiple model combinations and seventh information corresponding to each model combination. Each model combination includes a first model and a second model. The seventh information includes the MCS, the third information, multiple intervals of a first parameter, and a second parameter corresponding to each interval. The first parameter represents the channel sparsity, and the second parameter represents the accuracy of the corresponding first model and second model during model training.

[0012] In the above solution, the second information includes at least a third parameter. The third parameter represents the accuracy of channel compression and feedback. Testing the performance of the first model and the second model includes:

[0013] Testing the accuracy of the first model and the second model to obtain a fourth parameter. The fourth parameter represents the accuracy of the first model and the second model during testing.

[0014] Using the third parameter and the fourth parameter to determine eighth information. Wherein, when the fourth parameter is greater than the third parameter, the eighth information characterizes that the performance of the first model and the second model meets the requirements of channel compression and feedback. When the fourth parameter is less than or equal to the third parameter, the eighth information characterizes that the performance of the first model and the second model does not meet the requirements of channel compression and feedback.

[0015] In the above solution, the method further includes:

[0016] Sending the eighth information to the network device.

[0017] An embodiment of the present application further provides an information transmission method applied to a network device, including:

[0018] Sending first information and second information to a terminal. The first information is used to indicate a first model and a second model. The first model is used for channel compression, and the second model is used for channel decompression. The second information is used to indicate the requirements of channel compression and feedback.

[0019] When the performance of the first model and the second model meets the requirements of channel compression and feedback, receive the fourth information sent by the terminal; wherein,

[0020] The fourth information is determined by using a plurality of output elements, the MCS corresponding to the first model, and the third information. The plurality of output elements are obtained by performing channel compression using the first model. The total number of constellation points corresponding to the MCS is the same as the total number of the plurality of output elements. The third information represents the association relationship between the output elements of the first model and the constellation points corresponding to the MCS. The fourth information includes the constellation points corresponding to each of the plurality of output elements.

[0021] In the above solution, the method further includes:

[0022] Determine the ninth information and the first parameter, where the ninth information represents the channel quality and the first parameter represents the channel sparsity;

[0023] Use the fifth information, the ninth information, the first parameter, and the second information to determine the first information. The fifth information includes a plurality of domains and the sixth information corresponding to each domain. The sixth information includes a plurality of model combinations and the seventh information corresponding to each model combination. Each model combination includes a first model and a second model. The seventh information includes multiple intervals of the MCS, the third information, and the first parameter, and the second parameter corresponding to each interval. The second parameter represents the accuracy of the corresponding first model and second model during model training.

[0024] In the above solution, the second information at least includes a third parameter, and the third parameter represents the accuracy of channel compression and feedback; the method further includes:

[0025] Receive the eighth information sent by the terminal. The eighth information is determined by using the third parameter and the fourth parameter, and the fourth parameter represents the accuracy of the first model and the second model during testing. Wherein, when the fourth parameter is greater than the third parameter, the eighth information represents that the performance of the first model and the second model meets the requirements of channel compression and feedback; when the fourth parameter is less than or equal to the third parameter, the eighth information represents that the performance of the first model and the second model does not meet the requirements of channel compression and feedback.

[0026] In the above solution, the method further includes:

[0027] Based on the eighth information, configure time domain and / or frequency domain resources associated with channel compression and feedback for the terminal.

[0028] An embodiment of the present application further provides an information transmission device, including:

[0029] A first receiving unit, configured to receive a first piece of information and a second piece of information sent by a network device, where the first piece of information is used to indicate a first model and a second model, the first model is used for channel compression, the second model is used for channel decompression, and the second piece of information is used to indicate requirements for channel compression and feedback;

[0030] A first processing unit, configured to test the performance of the first model and the second model, and when the performance of the first model and the second model meets the requirements for channel compression and feedback, use the first model for channel compression to obtain a plurality of output elements; determine a fourth piece of information by using the plurality of output elements, the MCS corresponding to the first model, and a third piece of information;

[0031] A first sending unit, configured to send the fourth piece of information to the network device; where,

[0032] The total number of constellation points corresponding to the MCS is the same as the total number of the plurality of output elements, the third piece of information characterizes the association relationship between the output elements of the first model and the constellation points corresponding to the MCS, and the fourth piece of information includes the constellation points corresponding to each of the plurality of output elements.

[0033] An embodiment of this application further provides an information transmission device, including:

[0034] A second sending unit, configured to send a first piece of information and a second piece of information to a terminal, where the first piece of information is used to indicate a first model and a second model, the first model is used for channel compression, the second model is used for channel decompression, and the second piece of information is used to indicate requirements for channel compression and feedback;

[0035] A second receiving unit, configured to receive the fourth piece of information sent by the terminal when the performance of the first model and the second model meets the requirements for channel compression and feedback; where,

[0036] The fourth piece of information is determined by using a plurality of output elements, the MCS corresponding to the first model, and a third piece of information, the plurality of output elements are obtained by using the first model for channel compression, the total number of constellation points corresponding to the MCS is the same as the total number of the plurality of output elements, the third piece of information characterizes the association relationship between the output elements of the first model and the constellation points corresponding to the MCS, and the fourth piece of information includes the constellation points corresponding to each of the plurality of output elements.

[0037] An embodiment of this application further provides a terminal, including: a first communication interface and a first processor; where,

[0038] The first processor is configured to:

[0039] Receive the first information and the second information sent by the network device through the first communication interface, where the first information is used to indicate a first model and a second model, the first model is used for channel compression, the second model is used for channel decompression, and the second information is used to indicate the requirements for channel compression and feedback;

[0040] Test the performance of the first model and the second model. When the performance of the first model and the second model meets the requirements for channel compression and feedback, use the first model for channel compression to obtain a plurality of output elements; determine the fourth information by using the plurality of output elements, the MCS corresponding to the first model, and the third information, and send the fourth information to the network device through the first communication interface; where

[0041] The total number of constellation points corresponding to the MCS is the same as the total number of the plurality of output elements, the third information characterizes the association relationship between the output elements of the first model and the constellation points corresponding to the MCS, and the fourth information includes the constellation points corresponding to each of the plurality of output elements.

[0042] An embodiment of the present application further provides a network device, including: a second communication interface and a second processor; where

[0043] The second communication interface is used for:

[0044] Send the first information and the second information to the terminal, where the first information is used to indicate a first model and a second model, the first model is used for channel compression, the second model is used for channel decompression, and the second information is used to indicate the requirements for channel compression and feedback;

[0045] Receive the fourth information sent by the terminal when the performance of the first model and the second model meets the requirements for channel compression and feedback; where

[0046] The fourth information is determined by using a plurality of output elements, the MCS corresponding to the first model, and the third information. The plurality of output elements are obtained by using the first model for channel compression. The total number of constellation points corresponding to the MCS is the same as the total number of the plurality of output elements. The third information characterizes the association relationship between the output elements of the first model and the constellation points corresponding to the MCS. The fourth information includes the constellation points corresponding to each of the plurality of output elements.

[0047] An embodiment of the present application further provides a terminal, including: a first processor and a first memory for storing a computer program that can run on the processor,

[0048] Wherein, when the first processor is used to run the computer program, it executes the steps of any of the above methods on the terminal side.

[0049] An embodiment of the present application further provides a network device, including: a second processor and a second memory for storing a computer program that can run on the processor,

[0050] Wherein, when the second processor is used to run the computer program, it executes the steps of any of the above methods on the network device side.

[0051] An embodiment of the present application further provides a storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of any of the above methods on the terminal side, or implements the steps of any of the above methods on the network device side.

[0052] The information transmission method, apparatus, related equipment, and storage medium provided by the embodiments of this application. The terminal receives the first information and the second information sent by the network device. The first information is used to indicate the first model and the second model. The first model is used for channel compression, and the second model is used for channel decompression. The second information is used to indicate the requirements for channel compression and feedback. Test the performance of the first model and the second model. When the performance of the first model and the second model meets the requirements for channel compression and feedback, use the first model to perform channel compression to obtain a plurality of output elements. Use the plurality of output elements, the MCS corresponding to the first model, and the third information to determine the fourth information, and send the fourth information to the network device. Wherein, the total number of constellation points corresponding to the MCS is the same as the total number of the plurality of output elements. The third information characterizes the association relationship between the output elements of the first model and the constellation points corresponding to the MCS. The fourth information includes the constellation points corresponding to each of the plurality of output elements. In the solution provided by the embodiments of this application, the network device indicates to the terminal the channel compression model (i.e., the above-mentioned first model), the channel decompression model (i.e., the above-mentioned second model), and the requirements for channel compression and feedback. The terminal tests the performance of these two models. When the performance of these two models meets the requirements for channel compression and feedback, use the channel compression model to perform channel compression to obtain a plurality of output elements, and then use the plurality of output elements, the MCS corresponding to this channel compression model, and the mapping relationship between the output elements of this channel compression model and the constellation points corresponding to this MCS to determine the channel compression feedback information (i.e., the above-mentioned fourth information) including the constellation points corresponding to each of the plurality of output elements, and feedback this channel compression feedback information to the network device. In this way, since the terminal directly uses the channel compression model to obtain a plurality of output elements when testing that the channel compression model and the channel decompression model indicated by the network device meet the requirements for channel compression and feedback, and then directly uses the MCS corresponding to this channel compression model to perform modulation and coding on the plurality of output elements, it can effectively simplify the transmission channel processing flow on the premise of ensuring the requirements for channel compression and feedback. For example, it can omit at least one of the processes such as cyclic redundancy check (CRC, Cyclic Redundancy Check), channel coding, rate matching, and hybrid automatic repeat request (HARQ, Hybrid Automatic Repeat-reQuest), and scrambling on the output result of the channel compression model on the premise of ensuring the accuracy of channel transmission, so as to reduce the redundancy added by these processes when the terminal performs channel compression and feedback, such as reducing the redundancy added by channel coding, thereby significantly reducing the overhead of channel transmission and avoiding waste of resources when the terminal performs channel compression and feedback. Description of the Drawings

[0053] Figure 1 Schematic diagram of channel compression and feedback process in related technologies;

[0054] Figure 2 Schematic diagram of transmission channel processing process in related technologies;

[0055] Figure 3 Schematic diagram of the process of an information transmission method according to an embodiment of the present application;

[0056] Figure 4 Schematic diagram of transmission channel processing process according to an embodiment of the present application;

[0057] Figure 5 Schematic diagram of the difference in channel sparsity calculation between the line-of-sight (LOS) channel and the non-line-of-sight (NLOS) channel according to an embodiment of the present application;

[0058] Figure 6 Schematic diagram of quadrature phase shift keying (QPSK) constellation points according to an embodiment of the present application;

[0059] Figure 7 Schematic diagram of the process of another information transmission method according to an embodiment of the present application;

[0060] Figure 8 Schematic diagram of channel compression and feedback process of the application example of the present application;

[0061] Figure 9 Schematic diagram of the specific process of channel compression and feedback of the application example of the present application;

[0062] Figure 10 Schematic diagram of the structure of an information transmission device according to an embodiment of the present application;

[0063] Figure 11 Schematic diagram of the structure of another information transmission device according to an embodiment of the present application;

[0064] Figure 12 Schematic diagram of the structure of a terminal according to an embodiment of the present application;

[0065] Figure 13 Schematic diagram of the structure of a network device according to an embodiment of the present application;

[0066] Figure 14 Schematic diagram of the structure of an information transmission system according to an embodiment of the present application. Detailed implementation manners

[0067] The present application will be further described in detail below with reference to the accompanying drawings and embodiments.

[0068] In the related art, a terminal can feedback CSI to a base station based on a codebook. The terminal usually supports codebook types such as CSI type I, type II, and enhanced type II (which can be expressed in English as type II enhanced, abbreviated as etype II) for feedback of CSI such as rank indicator (RI), precoding matrix indicator (PMI), and channel quality indicator (CQI).

[0069] In the related art, as Figure 1 shown, the terminal can also use a data-driven deep learning method to perform channel compression and feedback the channel compression result to the base station so that the base station can obtain accurate downlink CSI. That is, the terminal can use a deep neural network (i.e., the coding network in Figure 1 , that is, the channel compression model) to extract low-dimensional features from a large amount of channel data for compression and feedback the channel compression result to the base station. The base station can use the corresponding deep neural network (i.e., the decoding network in Figure 1 , that is, the channel decompression model) to recover the channel compression result to obtain the state of the original channel. Since this solution can be optimized for specific scenarios, that is, optimized for the actual distribution law of the channel, it can improve the channel feedback accuracy under the same feedback overhead. In other words, compared with the codebook-based CSI feedback solution, this solution can enable the base station to obtain more accurate downlink CSI.

[0070] In the related art, the terminal can first perform CRC, channel coding, rate matching and HARQ, scrambling, modulation, layer mapping, discrete Fourier transform (DFT) uplink coding, multi-antenna precoding, resource mapping, and physical antenna mapping on the codebook-based CSI / AI (Artificial Intelligence)-based channel compression feedback information (i.e., the output variable of the channel compression model) through the transmission channel processing flow shown in Figure 2 , and then complete the CSI feedback through the air interface interaction between the terminal and the base station.

[0071] However, in the related art, through Figure 2When processing the AI-based channel compression feedback information (i.e., the output variables of the channel compression model) according to the transmission channel processing flow shown, the channel compression feedback information and other transmission information are usually modulated and encoded using a unified MCS based on information such as the signal-to-interference plus noise ratio (SINR). Due to the change of the communication environment, the amount of information contained in the channel matrix also changes. For the same number of output variables of the unified channel compression model, the amount of information contained in each output variable of the channel matrix compression model with different amounts of information is different, and the requirements for channel compression and feedback are also different. Therefore, using the above transmission channel processing flow may cause the terminal to transmit redundant bits, that is, resource waste may occur when the terminal performs channel compression and feedback.

[0072] Based on this, in various embodiments of the present application, the network device instructs the terminal about the channel compression model, the channel decompression model, and the requirements for channel compression and feedback. The terminal tests the performance of these two models. When the performance of these two models meets the requirements for channel compression and feedback, the terminal uses the channel compression model to perform channel compression to obtain multiple output elements, and then uses the multiple output elements, the MCS corresponding to the channel compression model, and the mapping relationship between the output elements of the channel compression model and the constellation points corresponding to the MCS to determine the channel compression feedback information including the constellation points corresponding to each output element among the multiple output elements, and feeds back the channel compression feedback information to the network device. In this way, since the terminal directly uses the channel compression model to obtain multiple output elements when testing that the channel compression model and the channel decompression model instructed by the network device meet the requirements for channel compression and feedback, and then directly uses the MCS corresponding to the channel compression model to modulate and encode the multiple output elements, it can effectively simplify the transmission channel processing flow on the premise of ensuring the requirements for channel compression and feedback. For example, it can omit at least one of the processes such as CRC, channel coding, rate matching and HARQ, and scrambling on the output result of the channel compression model on the premise of ensuring the accuracy of channel transmission, so as to reduce the redundancy added by these processes when the terminal performs channel compression and feedback, such as reducing the redundancy added by channel coding, and thus can significantly reduce the overhead of channel transmission and avoid resource waste when the terminal performs channel compression and feedback.

[0073] Specifically, an embodiment of the present application provides an information transmission method, which is applied to a terminal, as Figure 3 shown, the method includes:

[0074] Step 301: Receive the first information and the second information sent by a network device. The first information is used to indicate a first model and a second model. The first model is used for channel compression, and the second model is used for channel decompression. The second information is used to indicate the requirements for channel compression and feedback.

[0075] Step 302: Test the performance of the first model and the second model. When the performance of the first model and the second model meets the requirements for channel compression and feedback, use the first model for channel compression to obtain a plurality of output elements; determine fourth information by using the plurality of output elements, the MCS corresponding to the first model, and third information, and send the fourth information to the network device. The total number of constellation points corresponding to the MCS is the same as the total number of the plurality of output elements. The third information characterizes the association relationship between the output elements of the first model and the constellation points corresponding to the MCS. The fourth information includes the constellation point corresponding to each output element among the plurality of output elements.

[0076] In practical applications, the network device may specifically include a base station, etc. The specific type of the network device can be set according to requirements, and the embodiments of the present application do not limit this.

[0077] In practical applications, the terminal may also be referred to as a user equipment (UE), and may also be referred to as a user. The first model may also be referred to as a compression model or a channel compression model, etc. The second model may also be referred to as a decompression model or a channel decompression model, etc. The fourth information may also be referred to as compressed feedback channel information, channel compression feedback information, channel feedback information, or compression feedback information, etc. The embodiments of the present application do not limit these names, as long as their respective functions are realized.

[0078] In practical applications, when the terminal sends the fourth information to the network device, specific transmission channel processing may be performed on the fourth information. Since the terminal directly determines the fourth information by using the plurality of output elements, the MCS corresponding to the first model, and the third information, that is, the terminal directly performs modulation and coding on the plurality of output elements by using the MCS corresponding to the first model, the embodiments of the present application can simplify the transmission channel processing flow. Exemplarily, the transmission channel processing flow of the embodiments of the present application may be as Figure 4 shown, and is the same as Figure 2Compared with the transmission channel processing flow shown, CRC, channel coding, rate matching and HARQ, and scrambling processing are omitted. That is, the terminal can directly perform modulation processing on the AI-based channel compression feedback information (i.e., the multiple output elements) to obtain the fourth information, perform layer mapping, DFT uplink coding, multi-antenna precoding, resource mapping, and physical antenna mapping processing on the fourth information, and then send the processed fourth information to the network device.

[0079] In practical applications, it can be understood that the multiple output elements are the channel compression results obtained by quantifying through the first model; in other words, when using the first model for channel compression, the first model outputs multiple quantized elements. In addition, since the multiple output elements obtained by using the first model for channel compression are quantized, it shows that the first model and the second model are models with an auto-encoder architecture (which can be expressed as auto-encoder in English), and the elements output by such models can only be described by limited numerical values. Exemplarily, the first model and the second model can be models trained based on Vector Quantized-Variational Auto Encoder (VQ-VAE).

[0080] In practical applications, multiple model combinations can be pre-configured on the terminal for different channel compression and feedback requirements. Each model combination includes a first model and a second model, and the first information can specifically be used to indicate a first model and a second model included in one of the multiple model combinations. Here, it can be understood that the MCS corresponding to the first model is the same as the MCS corresponding to the second model, that is, the MCS corresponding to the corresponding model combination; the third information corresponding to the first model is the same as the third information corresponding to the second model, that is, the third information corresponding to the corresponding model combination. In addition, the specific content included in the first information can be set according to requirements. Exemplarily, the first information can include an identifier (such as a number, etc.) of a first model and an identifier of a second model, or can include an identifier of a model combination.

[0081] In actual application, the requirements for channel compression and feedback may include channel feedback accuracy (i.e., the accuracy of channel compression and feedback, which is also the channel transmission accuracy) and / or demodulation difficulty, etc. The demodulation difficulty is associated with the total number of constellation points corresponding to the MCS (i.e., the total number of constellation points included in the constellation diagram corresponding to the MCS). That is, the larger the total number of constellation points corresponding to the MCS, the higher the demodulation difficulty. In addition, the specific content included in the second information (i.e., the specific manifestation form of the requirements for channel compression and feedback) can be set according to requirements. Exemplarily, the second information may include a third parameter and / or the total number of constellation points corresponding to the MCS, and the third parameter represents the accuracy of channel compression and feedback.

[0082] In actual application, considering that the amount of information included in the channel matrix is different, the requirements for channel compression and feedback of the channel matrix may also be different. Therefore, on the terminal, there may be specifically pre-configured a mapping relationship (i.e., an association relationship) between different domains, different model combinations, different MCSs, different third information, different amounts of information of the channel matrix, and different accuracies of different model combinations during model training. When the terminal determines the fourth information by using the multiple output elements, the MCS corresponding to the first model, and the third information, it can first use this mapping relationship to determine the MCS corresponding to the first model and the third information, and then determine the fourth information. Among them, the channel sparsity can be used as an index to measure the amount of information included in the channel matrix. The higher the channel sparsity, the less the amount of information included in the channel matrix.

[0083] Based on this, in one embodiment, the method may further include:

[0084] Using fifth information to determine the MCS and the third information. The fifth information includes multiple domains and sixth information corresponding to each domain. The sixth information includes multiple model combinations and seventh information corresponding to each model combination. Each model combination includes a first model and a second model. The seventh information includes multiple intervals of MCS, third information, and first parameter, and second parameter corresponding to each interval. The first parameter represents the channel sparsity, and the channel sparsity can reflect the amount of information included in the channel matrix. The second parameter represents the accuracy of the corresponding first model and second model during model training.

[0085] In actual application, the network side may perform offline training of the multiple model combinations in advance, generate (i.e., determine) the fifth information, and configure the fifth information to the terminal. It can be understood that the network side includes at least one network device. The at least one network device may include a network device that sends the first information and the second information to the terminal, or may not include a network device that sends the first information and the second information to the terminal.

[0086] Specifically, the network side can pre-collect downlink channel estimation data, preprocess channel training samples for different domains (such as the spatio-frequency domain (i.e., the space-frequency domain) and / or the spatio-frequency-time domain (i.e., the space-frequency-time domain), etc.), that is, set (i.e., determine) the channel sparsity threshold according to the information amount distribution of the channel matrix (i.e., the threshold corresponding to the first parameter, which can be understood as the multiple endpoints corresponding to multiple intervals of the first parameter), design a set of channel compression models and channel decompression models (i.e., a model combination) for the channel matrices of different domains respectively, weight the loss function of the channel compression / decompression models according to the channel sparsity of the training samples (i.e., the first parameter), the higher the channel sparsity, the smaller the weight, so that the number of quantifiable values of the output variable of the channel compression model (i.e., the total number of output elements after quantization by the first model) is consistent with the number of constellation points corresponding to the MCS (i.e., the total number of constellation points corresponding to the MCS), and the values can be 2, 4, 16, 64, …, etc. Then, the network side can perform offline training of multiple model combinations, record the model training accuracy (i.e., the second parameter) corresponding to different channel sparsity intervals (i.e., the intervals of the first parameter) of each model combination under the corresponding MCS, and map (which can be understood as associating) the quantization codebook of each trained model combination under the corresponding MCS (the quantization codebook can be understood as a set of quantized values, i.e., the set of output elements after quantization by the first model) with the constellation points corresponding to the MCS one by one, to obtain the MCS mapping relationship table (i.e., the third information); in other words, the network side can determine the third information of each model combination under the corresponding MCS. Then, the network side can organize the above various types of information, determine the seventh information corresponding to each model combination, use the seventh information corresponding to each model combination to determine the sixth information corresponding to each domain, and use the sixth information corresponding to each domain to determine the fifth information. Finally, the network side can synchronize the multiple trained model combinations and the fifth information to the terminal (such as configuring them locally on the terminal before leaving the factory), and synchronize the multiple trained model combinations and the fifth information to the network device (such as pre-configuring them locally on the base station).

[0087] Among them, the specific forms of the fifth information, the sixth information, and the seventh information can be set according to requirements. Exemplarily, the fifth information, the sixth information, and the seventh information can be implemented through different tables. The fifth information can be as shown in Table 1, one sixth information can be as shown in Table 2, and one seventh information can be as shown in Table 3; or, the fifth information can be as shown in Table 4, that is, the fifth information, the sixth information, and the seventh information can be implemented through Table 4.

[0088] Table 1

[0089] Domain Domain Number Sixth Information Spatial-Frequency Domain (i.e., Space-Frequency Domain) <![CDATA[R 1 > Sixth Information - 1 Spatial-Frequency-Time Domain (i.e., Space-Frequency-Domain-Time Domain) <![CDATA[R 2 > Sixth Information - 2 … … …

[0090] Table 2

[0091] Model Number (i.e., Model Combination Number) Seventh Information <![CDATA[E 1 / D 1 > Seventh Information - 1 <![CDATA[E 2 / D 2 > Seventh Information - 2 … … <![CDATA[E N / D N > Seventh Information - N

[0092] Table 3

[0093]

[0094]

[0095] Table 4

[0096]

[0097]

[0098] In actual application, the network side may pre-deploy an AI model training system, and the AI model training system performs offline training on the multiple model combinations and determines the fifth information. The specific deployment method of the AI model training system can be set according to requirements, and the embodiments of the present application do not limit this. Exemplarily, the AI model training system may be deployed on a central unit (CU) and / or a distributed unit (DU) of a base station, that is, a base station performs offline training on the multiple model combinations and determines the fifth information; or, the AI model training system may be deployed on a logical entity across the CU and / or DU, that is, deployed on the CU and / or DU of multiple base stations. In other words, at least two base stations cooperate to perform offline training on the multiple model combinations and determine the fifth information.

[0099] In practical applications, the specific calculation method of the channel sparsity (i.e., the first parameter) can be set according to requirements, and the embodiments of the present application do not make any limitations. Exemplarily, the first parameter can specifically represent the proportion of the number of elements in the channel matrix that are less than a specific threshold (the threshold value can be set according to requirements) to the total number of all elements, and the first parameter can be represented by a relative percentage. Specifically, assuming that at time t, x(t) represents the transmitted signal and y(t) represents the received signal, then a mathematical model can be established for the relationship between the transmitted signal, the received signal, and the channel matrix H(t) as y(t) = H(t)x(t) + n(t); n(t) represents noise. When calculating the channel sparsity on the network side, it is necessary to preprocess the channel matrix. Specifically, a two-dimensional DFT can be performed on the channel matrix H(t) in the original spatio-frequency domain (i.e., the space-frequency domain) to obtain a channel matrix that is sparse in the angle-delay domain; or orthogonal time-frequency space modulation (OTFS, Orthogonal Time Frequency Space) can be performed on the channel matrix in the original spatio-frequency time domain (i.e., the space-frequency-time domain) to obtain a channel matrix that is sparse in the angle-delay Doppler domain; the preprocessed channel matrix can be denoted as H'. After the preprocessing is completed, the network side can calculate the modulus value of each single element in each sample H' of the sampled channel matrix dataset, and record the corresponding modulus matrix as a new sample, that is, the elements in the sample matrix After that, the sum of all elements in each new sample matrix can be calculated and denoted as sum; each element in the new sample matrix is sorted from largest to smallest to form a vector h, and the elements of this vector can be denoted as h i ; after that, the sorted elements can be accumulated and divided by sum, that is, calculate Record the subscript of the first h′ k ≥80%, denoted as K; the difference obtained by subtracting K from the total number of channel matrix elements and then dividing by the total number of channel matrix elements is the channel sparsity S (i.e., the first parameter) of the channel matrix. The channel sparsity S defined in the above manner can distinguish channels to a certain extent. For example, the calculation differences of the channel sparsity between the line-of-sight (LOS, Line Of Sight) channel and the non-line-of-sight (NLOS, Non Line of Sight) channel can be as Figure 5 shown. After the NLOS channel matrix and the LOS channel matrix are preprocessed and transformed into the angle-delay domain, the matrix sparsity of the NLOS channel can be calculated as 74.45% (i.e., (3072 - 785) / 3072 * 100%) and the matrix sparsity of the LOS channel can be calculated as 90.43% (i.e., (3072 - 294) / 3072 * 100%) by the above method, and there are obvious differences between the two.

[0100] In practical applications, when the network side determines the loss function corresponding to the first model and / or the second model, assuming that the number of sparsity intervals corresponding to a model combination is K, and a channel sample belongs to the k-th interval, the corresponding loss function can be determined as follows:

[0101]

[0102] where decoder represents the decompression operation of the compressed channel matrix; ‖·,·‖ represents a function for calculating the difference between two elements, such as a function like taking the negative of the cosine similarity; sg represents stop gradient (which can be expressed in English as stopgradient), that is, the gradient is not calculated during backpropagation; z represents the output before quantization, and z q represents the output after quantization, and γ k <β k is used to restrict as much as possible to make z q approach z, rather than making z approach z q ; λ k is used to weight the channel sparsity. In principle, the higher the sparsity of the channel matrix, the smaller the weight. When the network side trains each model combination, it can train the first model and / or the second model by minimizing the above loss function (1).

[0103] In practical applications, the specific manifestation form of the third information (i.e., the above MCS mapping relationship table, such as the above etc.) can be set according to requirements. Specifically, when the network side trains a model combination, it can determine the output range of the first model as [min, max] according to the corresponding training samples, map the output of the first model to the [0, 1] interval through a scaling function (such as etc.), train the quantization autoencoder offline by restricting the quantifiable values of the output variables of the first model (i.e., training the first model and the second model), and determine the third information of this model combination under the corresponding MCS. Exemplarily, assuming the use of Figure 6 the orthogonal phase shift keying (QPSK, Quadrature Phase Shift Keying) constellation diagram shown, that is, assuming that the total number of output elements after quantization by the first model and the total number of constellation points corresponding to the MCS are both 4, and assuming that the output elements after quantization by the first model are 0, 0.4, 0.8, and 1, then the network side can map the output elements after quantization by the first model to the QPSK constellation points as shown in Table 5, that is, the MCS mapping relationship table corresponding to the first model (i.e., the third information) can be as shown in Table 5.

[0104] Table 5

[0105] QPSK Constellation Point Output Element after Quantization by the First Model 00 0 01 0.4 10 0.8 11 1

[0106] In actual application, the specific manner for the network device to determine the second information can be set according to requirements. Exemplarily, the network device may be pre-configured with the accuracy of channel compression and feedback (i.e., the third parameter); and / or, the network device may determine the total number of constellation points corresponding to the MCS according to the current capabilities of the network (such as computing capabilities and / or transmission capabilities, etc.), that is, determine the demodulation difficulty.

[0107] In actual application, the network device may also determine the ninth information and the first parameter characterizing the channel quality, and then use the fifth information, the ninth information, the first parameter, and the second information to determine the first information; wherein, the specific manifestation form of the ninth information can be set according to requirements, such as SINR, etc. Exemplarily, the base station may complete channel information preprocessing according to the selected domain (such as the spatial-frequency domain and / or the spatial-frequency time domain, etc.), select the available MCS according to the SINR of the uplink channel (i.e., the ninth information), and calculate the sparsity of the channel matrix (i.e., the first parameter) according to the sounding reference signal (SRS) of the terminal; thereafter, the base station may compare the calculated sparsity of the channel matrix with the pre-set channel sparsity threshold included in the fifth information (such as those in Table 5, etc., that is, the multiple endpoints corresponding to the multiple intervals of the first parameter), and find (i.e., determine) the corresponding channel sparsity interval according to the comparison result with the channel sparsity threshold, that is, determine the interval of the first parameter; thereafter, the base station may, according to the requirements of channel compression and feedback (such as accuracy and / or demodulation difficulty, etc., that is, the second information), select the optimal model combination that meets the conditions (i.e., can meet the requirements of channel compression and feedback) from this interval, such as selecting the model combination with the highest model training accuracy (i.e., the second parameter), or selecting a model combination whose model training accuracy is not the highest but can meet the lower demodulation difficulty requirement, that is, selecting the model combination with the highest model training accuracy on the premise of being able to meet the lower demodulation difficulty requirement.

[0108] In actual application, the network device may send the first information and the second information to the terminal through the physical downlink control channel (PDCCH, Physical Downlink Control CHannel) or the physical downlink shared channel (PDSCH, Physical Downlink Shared CHannel); that is, the terminal may receive the first information and the second information through the PDCCH or the PDSCH.

[0109] In practical applications, when the network device cannot determine the first information by using the fifth information, the ninth information, the first parameter, and the second information, that is, when all model combinations trained on the network side do not meet the requirements of channel compression and feedback, and / or all model combinations corresponding to the channel sparsity interval determined according to the sparsity of the channel matrix do not meet the requirements of channel compression and feedback, the network device may configure the terminal to perform CSI feedback in other ways, such as performing CSI feedback based on an etype II codebook.

[0110] In practical applications, when the second information at least includes the accuracy of channel compression and feedback (i.e., the third parameter), when the terminal tests the performance of the first model and the second model, it can re-determine the accuracy of the first model and the second model, and compare the re-determined accuracy with the third parameter, and determine whether the performance of the first model and the second model meets the requirements of channel compression and feedback according to the comparison result.

[0111] Based on this, in an embodiment, the second information may at least include a third parameter, and the third parameter represents the accuracy of channel compression and feedback; the testing of the performance of the first model and the second model may include:

[0112] Test the accuracy of the first model and the second model to obtain a fourth parameter, and the fourth parameter represents the accuracy of the first model and the second model during testing;

[0113] Use the third parameter and the fourth parameter to determine the eighth information; wherein, when the fourth parameter is greater than the third parameter, the eighth information represents that the performance of the first model and the second model meets the requirements of channel compression and feedback; when the fourth parameter is less than or equal to the third parameter, the eighth information represents that the performance of the first model and the second model does not meet the requirements of channel compression and feedback.

[0114] Among them, in practical applications, the specific manifestation form of the eighth information can be set according to requirements. Exemplarily, when the eighth information is marked as 1, the eighth information can reflect that the fourth parameter is greater than the third parameter, that is, the eighth information represents that the performance of the first model and the second model meets the requirements of channel compression and feedback; at this time, the terminal can use the first model to perform channel compression to obtain multiple quantized output elements, use the multiple quantized output elements, the MCS corresponding to the first model, and the third information to determine the fourth information, and adopt Figure 4The transmission channel processing flow shown processes the fourth information and sends the processed fourth information to the network device. When the eighth information is identified as 0, the eighth information can reflect that the fourth parameter is less than or equal to the third parameter, that is, the eighth information indicates that the performance of the first model and the second model does not meet the requirements of channel compression and feedback; at this time, the terminal can use other methods for CSI feedback, such as CSI feedback based on the etype II codebook, and can use Figure 2 the transmission channel processing flow shown to process the CSI based on the etype II codebook.

[0115] In actual application, the terminal needs to feedback the eighth information to the network device for the network device to configure the time domain and / or frequency domain resources associated with channel compression and feedback for the terminal based on the eighth information. Exemplarily, when the eighth information indicates that the performance of the first model and the second model meets the requirements of channel compression and feedback, the network device can configure the Figure 4 time domain and / or frequency domain resources corresponding to the transmission channel processing flow shown; when the eighth information indicates that the performance of the first model and the second model does not meet the requirements of channel compression and feedback, the network device can configure the terminal to perform CSI feedback based on the etype II codebook and can configure the Figure 2 time domain and / or frequency domain resources corresponding to the transmission channel processing flow shown.

[0116] In actual application, it can be understood that when the eighth information indicates that the performance of the first model and the second model meets the requirements of channel compression and feedback, after receiving the fourth information, the network device can perform demodulation processing using a transmission channel processing flow Figure 4 opposite to that shown, select the MCS corresponding to the second model and the third information (i.e., the MCS and the third information corresponding to the first model) for demapping according to the fifth information, and can use the second model for channel decompression. When the eighth information indicates that the performance of the first model and the second model does not meet the requirements of channel compression and feedback, after receiving the CSI based on the etype II codebook, the network device can perform demodulation processing using a transmission channel processing flow Figure 2 opposite to that shown.

[0117] In actual application, the network device can monitor the first parameter and the ninth information in real time. When the first parameter and / or the ninth information changes, the network device can use the new first parameter and / or the ninth information to update the first information and send the updated first information to the terminal. After receiving the updated first information, the terminal can re-execute step 302 above based on the updated first information. Exemplarily, the base station can adjust the model combination by monitoring SRS according to the uplink channel quality (i.e., the ninth information) and the sparsity of the channel matrix (i.e., the first parameter). When the threshold interval corresponding to the sparsity changes, the base station can re-select the optimal model combination corresponding to the threshold interval and indicate it to the terminal. When the uplink channel quality changes, the base station can adjust the available MCS selection range according to the quality of the uplink channel, and can re-select the optimal model combination according to the interval where the sparsity belongs and indicate it to the terminal. The terminal tests whether the performance of the model combination re-indicated by the base station meets the requirements of channel compression and feedback, and feeds back the test result (i.e., the eighth information) to the base station. When the performance of the re-indicated model combination meets the requirements of channel compression and feedback, the terminal uses the channel compression model re-indicated by the base station for channel compression. When the performance of the re-indicated model combination does not meet the requirements of channel compression and feedback, the terminal performs CSI feedback based on the type II codebook. The base station allocates corresponding time domain and / or frequency domain resources according to the test result (i.e., the eighth information) fed back by the terminal.

[0118] Correspondingly, an embodiment of the present application further provides an information transmission method, which is applied to a network device (such as a base station, etc.), as Figure 7 shown, the method includes:

[0119] Step 701: Send a first information and a second information to the terminal, where the first information is used to indicate a first model and a second model, the first model is used for channel compression, the second model is used for channel decompression, and the second information is used to indicate the requirements of channel compression and feedback;

[0120] Step 702: Receive the fourth information sent by the terminal when the performance of the first model and the second model meets the requirements of channel compression and feedback;

[0121] Wherein, the fourth information is determined by using a plurality of output elements, the MCS corresponding to the first model, and the third information. The plurality of output elements are obtained by performing channel compression using the first model. The total number of constellation points corresponding to the MCS is the same as the total number of the plurality of output elements. The third information characterizes the association relationship between the output elements of the first model and the constellation points corresponding to the MCS. The fourth information includes the constellation points corresponding to each of the plurality of output elements.

[0122] In one embodiment, the method may further include:

[0123] Determine a ninth piece of information and a first parameter, where the ninth piece of information characterizes the channel quality, the first parameter characterizes the channel sparsity, and the channel sparsity can reflect the amount of information contained in the channel matrix;

[0124] Use the fifth piece of information, the ninth piece of information, the first parameter, and the second piece of information to determine the first piece of information; the fifth piece of information includes multiple domains and sixth information corresponding to each domain, the sixth piece of information includes multiple model combinations and seventh information corresponding to each model combination, each model combination includes a first model and a second model, the seventh piece of information includes MCS, a third piece of information, multiple intervals of the first parameter, and a second parameter corresponding to each interval, and the second parameter characterizes the accuracy of the corresponding first model and second model during model training.

[0125] In one embodiment, the second piece of information at least includes a third parameter, and the third parameter characterizes the accuracy of channel compression and feedback; the method may further include:

[0126] Receive an eighth piece of information sent by the terminal, where the eighth piece of information is determined using the third parameter and a fourth parameter, and the fourth parameter characterizes the accuracy of the first model and the second model during testing; wherein, when the fourth parameter is greater than the third parameter, the eighth piece of information characterizes that the performance of the first model and the second model meets the requirements of channel compression and feedback; when the fourth parameter is less than or equal to the third parameter, the eighth piece of information characterizes that the performance of the first model and the second model does not meet the requirements of channel compression and feedback.

[0127] In one embodiment, the method may further include:

[0128] Based on the eighth piece of information, configure time domain and / or frequency domain resources associated with channel compression and feedback for the terminal.

[0129] The information transmission method provided by the embodiments of this application. The terminal receives the first information and the second information sent by the network device. The first information is used to indicate the first model and the second model. The first model is used for channel compression, and the second model is used for channel decompression. The second information is used to indicate the requirements for channel compression and feedback. Test the performance of the first model and the second model. When the performance of the first model and the second model meets the requirements for channel compression and feedback, use the first model to perform channel compression to obtain multiple output elements. Use the multiple output elements, the MCS corresponding to the first model, and the third information to determine the fourth information, and send the fourth information to the network device. Wherein, the total number of constellation points corresponding to the MCS is the same as the total number of the multiple output elements. The third information represents the association relationship between the output elements of the first model and the constellation points corresponding to the MCS. The fourth information includes the constellation points corresponding to each of the multiple output elements. In the solution provided by the embodiments of this application, the network device instructs the terminal about the channel compression model (i.e., the above-mentioned first model), the channel decompression model (i.e., the above-mentioned second model), and the requirements for channel compression and feedback. The terminal tests the performance of these two models. When the performance of these two models meets the requirements for channel compression and feedback, use the channel compression model to perform channel compression to obtain multiple output elements, and then use the multiple output elements, the MCS corresponding to this channel compression model, and the mapping relationship between the output elements of this channel compression model and the constellation points corresponding to this MCS to determine the channel compression feedback information (i.e., the above-mentioned fourth information) including the constellation points corresponding to each of the multiple output elements, and feedback this channel compression feedback information to the network device. In this way, since the terminal directly uses the channel compression model to obtain multiple output elements when testing that the channel compression model and the channel decompression model indicated by the network device meet the requirements for channel compression and feedback, and then directly uses the MCS corresponding to this channel compression model to perform modulation and coding on the multiple output elements, it can effectively simplify the transmission channel processing flow on the premise of ensuring the requirements for channel compression and feedback. For example, it can omit at least one of the processes such as CRC, channel coding, rate matching and HARQ, and scrambling on the output result of the channel compression model on the premise of ensuring the accuracy of channel transmission, so as to reduce the redundancy added by these processes when the terminal performs channel compression and feedback, such as reducing the redundancy added by channel coding, thus significantly reducing the overhead of channel transmission and avoiding waste of resources when the terminal performs channel compression and feedback.

[0130] In addition, in the solution provided by the embodiments of this application, in some specific cases (such as when the requirement for demodulation difficulty is relatively low), the network device can reduce the difficulty of demodulating the channel feedback information by selecting a model combination that can meet the relatively low requirement for demodulation difficulty and instructing it to the terminal.

[0131] The present application will be further described in detail below in combination with application examples.

[0132] In this application example, considering that the higher the channel sparsity (i.e., the sparsity of the channel matrix), the less information contained in a single output element of the channel compression model under the same compression ratio, due to the different amounts of information contained in the channel matrix, the processing flow and processing method of the compressed channel matrix can be modified. For example, the training loss function can be adjusted according to the information contained in the channel, and the constellation points of MCS are directly used to represent the quantized output, and then Figure 4 the transmission channel processing flow shown in which CRC, channel coding, rate matching and HARQ, and scrambling processing are omitted is used for processing. Thus, on the premise of ensuring the accuracy of channel transmission, not only can the transmission overhead be significantly reduced (such as reducing the redundancy added in steps such as channel coding), but also the transmission channel processing flow can be effectively simplified. In addition, in some specific cases (such as when the demodulation difficulty requirement is low), the difficulty of demodulating the channel feedback information can also be reduced. Among them, the index used to measure the information contained in the channel in this application example is the channel matrix sparsity, and the channel matrix sparsity can be used to represent the proportion of the number of elements less than a preset specific threshold value in the channel matrix to the total number of elements, and can be expressed in relative percentage.

[0133] Specifically, in this application example, first, the network side preprocesses the channel matrix training samples in different domains (such as the spatio-frequency domain and / or the spatio-frequency time domain, etc.). That is, for each domain, the channel matrix sparsity threshold is set according to the information amount distribution of the channel matrix, and a compression model and a decompression model based on the quantization autoencoder (i.e., the above model combination) are designed for the channel matrices in different domains respectively. The loss functions of the compression model and the decompression model are weighted according to the channel matrix sparsity of the training samples. The higher the channel matrix sparsity, the smaller the weight. The number of quantizable values of the output variable of the channel compression model is the same as the number of constellation points corresponding to MCS. Then, based on the above configuration, the network side completes the training of the compression model and the decompression model, and records the model training accuracy corresponding to different sparsity intervals of the compression model and the decompression model under the corresponding MCS. The trained quantization codebook (i.e., the set of multiple output elements quantized by the compression model) is mapped one by one with the points of the constellation diagram and the mapping relationship table (i.e., the above third information) is recorded. The trained model and its number, the corresponding output variable quantization codebook, the channel sparsity threshold interval and the corresponding MCS, and the mapping relationship table are synchronized to the terminal and the base station (i.e., the above fifth information is synchronized to the terminal and the base station).

[0134] After that, the base station selects an available MCS according to the selected domain and the SINR of the uplink channel. According to channel reciprocity, it estimates the sparsity of the downlink channel through the sparsity of the uplink channel, compares it with the sparsity threshold, finds the corresponding sparsity interval according to the comparison result with the sparsity threshold, and selects the available model with the highest accuracy in this sparse interval. When the accuracy cannot meet the requirements, the terminal is configured to use other methods for CSI feedback, such as CSI feedback based on the type II codebook. The base station sends channel compression feedback model configuration information (this information can include the above first information and the above second information) to the terminal according to the selected available model. The terminal tests the model accuracy according to the model number (i.e., the above first information) in the received channel compression feedback model configuration information and feeds back to the base station whether the tested model accuracy meets the accuracy requirements (i.e., the above second information). The base station configures the time domain and / or frequency domain resources required for the terminal to feedback CSI according to the feedback of the terminal. If the model accuracy tested by the terminal meets the accuracy requirements, as Figure 8 shown, the terminal selects the compression model indicated by the base station to complete channel compression, maps the output variables according to the MCS and the mapping relationship table corresponding to the model (i.e., the above third information), and feeds back the compressed channel information (i.e., the above fourth information) to the base station according to the corresponding resource configuration. The base station selects the corresponding MCS and mapping relationship table to complete demapping and decompresses the channel information through the corresponding decompression model.

[0135] In this application example, as Figure 9 shown, the specific process of channel compression and feedback may include the following steps:

[0136] Step 901: The network side collects channel data in different domains, sets the threshold of sparsity (i.e., the above first parameter) according to the data distribution, offline trains a channel compression model (i.e., the above first model) and a channel decompression model (i.e., the above second model) based on the AI quantization autoencoder according to the number of constellation points corresponding to different MCSs, records the accuracy (i.e., the above second parameter) of the model combination for different sparsity threshold intervals, and synchronizes the model combination and Table 4 (i.e., the above fifth information) to the base station and the terminal, and then executes Step 902;

[0137] Step 902: The terminal accesses the network and enters the radio resource control (RRC) connected state (which can be expressed in English as RRC-CONNECTED), sends SRS according to the configuration, and then executes Step 903;

[0138] Step 903: The base station selects a model combination according to the uplink channel quality (i.e., the above-mentioned ninth information), the sparsity of the SRS channel matrix (i.e., the above-mentioned first parameter), and the model accuracy requirement (i.e., the third parameter included in the above-mentioned second information), and feeds back resources according to the MCS configuration corresponding to the model combination; sends channel compression feedback model configuration information to the terminal, which includes the above-mentioned first information and the above-mentioned second information, and then executes Step 904;

[0139] Step 904: The base station sends a channel state information reference signal (CSI-RS, CSI-Reference Signal) according to the configuration, and then executes Step 905;

[0140] Step 905: The terminal tests the model accuracy according to the compression model and decompression model indicated by the base station, and sends an indication (i.e., the above-mentioned eighth information) of whether the test model accuracy (i.e., the above-mentioned fourth parameter) meets the accuracy requirement to the base station, and then executes Step 906;

[0141] Step 906: The base station configures the time domain and / or frequency domain resources for CSI feedback according to the indication fed back by the terminal, and then executes Step 907;

[0142] Step 907: If the test model accuracy meets the accuracy requirement, the terminal selects a compression model according to the model number (i.e., the above-mentioned first information) and compresses the channel information, and completes the channel information coding and modulation according to the MCS corresponding to the compression model; otherwise, it completes the channel feedback through the Figure 2 shown process; then, the terminal feeds back the channel information according to the resource configuration set by the base station, and then executes Step 908;

[0143] Step 908: If the test model accuracy meets the accuracy requirement, the base station demaps according to the mapping table (i.e., the above-mentioned third information) to obtain the decompressed relevant channel information, and then executes Step 909;

[0144] Step 909: The base station updates the channel compression feedback model configuration information (i.e., updates the above-mentioned first information) according to the SRS channel sparsity and the uplink channel quality and sends it to the terminal. The terminal feeds back an indication according to the test accuracy (i.e., the above-mentioned eighth information). The base station reallocates the time domain and / or frequency domain resources according to the fed-back indication, and the terminal reports the channel information according to the new configuration.

[0145] Among them, in step 901, the AI model training system of the radio access network (this system can be deployed on the CU and / or DU of a base station, or can be deployed on a logical entity across the CU and / or DU, that is, deployed on the CU and / or DU of multiple base stations) pre-collects downlink channel estimation data, and preprocesses channel training samples in different domains (such as the spatio-frequency domain and / or the spatio-frequency time domain, etc.), that is, sets the matrix sparsity threshold (i.e., the channel sparsity threshold) according to the sparsity distribution of the training samples, and assigns different weights to the loss function for channel samples in different sparsity intervals. The higher the sparsity of the sample, the smaller the weight λ of the loss function. According to the model accuracy of the compression / decompression model corresponding to the sample record in the channel matrix sparsity interval. Suppose after channel analysis, 1 matrix sparsity threshold is selected and denoted as δ i = 0.83. According to the different channel matrix sparsities corresponding to the collected data, the sample data is grouped and different training weights λ 1 = 1, λ 1 = 0.5 are assigned. For the channel matrix in the spatio-frequency domain, compression models and decompression models based on the quantized autoencoder (model combinations can be expressed as E 2 / D 11 , E 11 / D 12 , E 12 / D 13 , E 13 / D 14 / D 14 ) are designed for different MCSs (including Binary Phase Shift Keying (BPSK), QPSK, 16-Quadrature Amplitude Modulation (QAM), and 64-QAM) respectively. The number of quantifiable values of the model output variable is the same as the number of constellation points corresponding to the MCS (2, 4, 16, 64 respectively). The AI model training system offline trains the channel compression model and decompression model based on the AI quantized autoencoder (i.e., VQ-VAE) based on the above assumptions, and records the sparsity intervals [0, δ 1 ) and [δ 1,1]For the corresponding model training accuracy, map the trained quantization codebook to the points of the constellation diagram one by one and record the MCS mapping relationship table (i.e., the above-mentioned third piece of information). The mapping relationship tables among the domain, threshold, MCS, MCS mapping relationship table, and accuracy can be shown in Table 6. Table 6 can be understood as a specific implementation of Table 4 and contains the above-mentioned fifth piece of information. The AI model training system can synchronize the trained compression model, decompression model, model number, corresponding output variable quantization codebook, channel sparsity threshold, corresponding MCS, and MCS mapping relationship table to the terminal and the base station, that is, synchronize Table 6 and multiple model combinations to the terminal and the base station.

[0146] Table 6

[0147]

[0148] In step 903, the base station completes the preprocessing of the spatial-frequency domain channel information, selects available MCS according to the SINR of the uplink channel (i.e., the above-mentioned ninth piece of information). For example, determine that the available MCS includes BPSK, QPSK, and 16-QAM according to the SINR of the uplink channel; calculate the sparsity of the channel matrix according to the SRS (i.e., the above-mentioned first parameter). Suppose the calculated sparsity of the channel matrix is 0.91. 0.91 can be compared with the pre-set sparsity threshold δ 1 = 0.83, and according to the sparsity comparison result, find the corresponding sparsity interval [δ 1 ,1]; according to the requirement (i.e., the above-mentioned second piece of information), select the optimal model combination that meets the requirement in the corresponding sparse interval, for example, it is E 13 / D 13 . Then, the base station can send the channel compression feedback model configuration information to the terminal through PDCCH / PDSCH. The channel compression feedback model configuration information includes the channel compression feedback model number (i.e., the above-mentioned first piece of information). In some cases, the channel compression feedback model configuration information may also include the required model accuracy (i.e., the third parameter included in the above-mentioned second piece of information).

[0149] In step 905, the terminal selects models E 13 and D 13 according to the model number in the received channel compression feedback model configuration information and completes the AI quantization autoencoder-based (i.e., models E 13 and D 13) Channel compression and decompression, that is, perform performance testing to obtain the test accuracy (i.e., the above-mentioned fourth parameter); the terminal compares the test accuracy with the required accuracy sent by the base station (i.e., the above-mentioned third parameter). If the test accuracy is higher than the required accuracy, it sends identification 1 to the base station, that is, sends the above-mentioned eighth information indicating that the performance of the first model and the second model meets the requirements of the channel compression and feedback; otherwise, it sends 0 to the base station, that is, sends the above-mentioned eighth information indicating that the performance of the first model and the second model does not meet the requirements of the channel compression and feedback to the base station.

[0150] In step 906, if the base station receives identification 1, it configures the time domain and / or frequency domain resources required for the channel processing flow as shown in Figure 4 to the terminal; if it receives identification 0, it configures the terminal to perform CSI feedback based on the etype II codebook and configures the time domain and / or frequency domain resources required for the channel processing flow as shown in Figure 2 to the terminal.

[0151] In step 907, when the identification sent by the terminal in step 905 is 1, that is, when the above-mentioned eighth information indicates that the performance of the first model and the second model meets the requirements of the channel compression and feedback, the terminal maps the quantized output of the compression model E 13 according to the corresponding mapping relationship table (i.e., the above-mentioned third information), and feedbacks the compressed channel information (i.e., the fourth information) to the base station according to the MCS and resource configuration corresponding to the model selected by the base station (i.e., E 13 / D 13 ). The feedback process of this channel information is as shown in Figure 4 . Otherwise, the terminal performs CSI feedback according to the process shown in Figure 2 .

[0152] In step 908, when the identification received by the base station in step 905 is 1, that is, when the above-mentioned eighth information indicates that the performance of the first model and the second model meets the requirements of the channel compression and feedback, the base station selects the MCS mapping table corresponding to the decompression model D 13 for demapping, and decompresses the channel information according to the corresponding decompression model D 13 . Otherwise, the base station can refer to the channel information processing process when receiving CSI based on the etype II codebook to solve the relevant channel information.

[0153] In step 909, the subsequent base station adjusts the model by listening to the SRS according to the uplink channel quality and the sparsity of the channel matrix. When the threshold interval corresponding to the sparsity changes, the base station re-selects the optimal model corresponding to the threshold. The terminal completes channel compression and decompression according to the compression and decompression models selected by the base station to obtain the test accuracy. The base station allocates the corresponding feedback time domain and / or frequency domain resources according to whether the test accuracy meets the required accuracy. The terminal adjusts the channel feedback information processing method (i.e., adjusts the compression model or switches to feedback CSI based on the type II codebook) and reports the channel information to be feedback according to the configuration. When the uplink channel quality changes, the base station adjusts the available MCS selection range according to the uplink channel quality and selects the optimal model according to the interval to which the sparsity belongs. The terminal completes channel compression and decompression according to the compression and decompression models selected by the base station to obtain the test accuracy. The base station allocates the corresponding time domain and / or frequency domain resources according to whether the test accuracy meets the required accuracy. The terminal adjusts the channel feedback information processing method (i.e., adjusts the compression model or switches to feedback CSI based on the type II codebook) and reports the channel information to be feedback according to the configuration. It can be understood that step 909 can be implemented with reference to the specific implementation process of steps 903 to 908, which will not be elaborated here.

[0154] The solution provided in this application example can, on the premise of ensuring the channel transmission accuracy, not only significantly reduce the transmission overhead and thus reduce the energy consumption, but also effectively simplify the processing flow of the channel feedback information. In addition, in some cases, it can also reduce the difficulty of demodulating the channel feedback information.

[0155] To implement the method on the terminal side in the embodiments of the present application, the embodiments of the present application further provide an information transmission device disposed on the terminal, as Figure 10 shown. The device includes:

[0156] A first receiving unit 1001, configured to receive a first piece of information and a second piece of information sent by a network device, where the first piece of information is used to indicate a first model and a second model, the first model is used for channel compression, the second model is used for channel decompression, and the second piece of information is used to indicate the requirements for channel compression and feedback;

[0157] A first processing unit 1002, configured to test the performance of the first model and the second model. When the performance of the first model and the second model meets the requirements for channel compression and feedback, use the first model for channel compression to obtain a plurality of output elements; use the plurality of output elements, the MCS corresponding to the first model, and a third piece of information to determine a fourth piece of information;

[0158] A first sending unit 1003, configured to send the fourth piece of information to the network device; where

[0159] The total number of constellation points corresponding to the MCS is the same as the total number of the multiple output elements. The third information characterizes the association relationship between the output elements of the first model and the constellation points corresponding to the MCS. The fourth information includes the constellation points corresponding to each of the multiple output elements.

[0160] Wherein, in one embodiment, the first processing unit 1002 is further configured to determine the MCS and the third information by using fifth information. The fifth information includes multiple domains and sixth information corresponding to each domain. The sixth information includes multiple model combinations and seventh information corresponding to each model combination. Each model combination includes a first model and a second model. The seventh information includes multiple intervals of the MCS, the third information, and the first parameter, and a second parameter corresponding to each interval. The first parameter represents the channel sparsity, and the channel sparsity can reflect the amount of information contained in the channel matrix. The second parameter represents the accuracy of the corresponding first model and the second model during model training.

[0161] In one embodiment, the second information includes at least a third parameter, and the third parameter represents the accuracy of channel compression and feedback. Correspondingly, the first processing unit 1002 is specifically configured to:

[0162] Test the accuracy of the first model and the second model to obtain a fourth parameter, where the fourth parameter represents the accuracy of the first model and the second model during testing;

[0163] Determine the eighth information by using the third parameter and the fourth parameter. Wherein, when the fourth parameter is greater than the third parameter, the eighth information characterizes that the performance of the first model and the second model meets the requirements of channel compression and feedback; when the fourth parameter is less than or equal to the third parameter, the eighth information characterizes that the performance of the first model and the second model does not meet the requirements of channel compression and feedback.

[0164] In one embodiment, the first sending unit 1003 is further configured to send the eighth information to the network device.

[0165] In practical applications, the first receiving unit 1001 and the first sending unit 1003 can be implemented by a communication interface in the information transmission device; the first processing unit 1002 can be implemented by a processor in the information transmission device.

[0166] To implement the method on the network device side in the embodiments of the present application, the embodiments of the present application further provide an information transmission device, which is disposed on the network device, as Figure 11 shown. The device includes:

[0167] A second transmitting unit 1101, configured to transmit first information and second information to a terminal, where the first information is used to indicate a first model and a second model, the first model is used for channel compression, the second model is used for channel decompression, and the second information is used to indicate requirements for channel compression and feedback;

[0168] A second receiving unit 1102, configured to receive fourth information sent by the terminal when the performance of the first model and the second model meets the requirements for channel compression and feedback; where

[0169] The fourth information is determined by using a plurality of output elements, a modulation and coding scheme (MCS) corresponding to the first model, and third information. The plurality of output elements are obtained by performing channel compression using the first model. The total number of constellation points corresponding to the MCS is the same as the total number of the plurality of output elements. The third information characterizes the association relationship between the output elements of the first model and the constellation points corresponding to the MCS. The fourth information includes the constellation points corresponding to each of the plurality of output elements.

[0170] Where, in one embodiment, as Figure 11 shown, the apparatus may further include:

[0171] A second processing unit 1103, configured to:

[0172] Determine ninth information and a first parameter, where the ninth information characterizes channel quality, and the first parameter characterizes channel sparsity, and the channel sparsity can reflect the amount of information contained in a channel matrix;

[0173] Determine the first information by using fifth information, the ninth information, the first parameter, and the second information. The fifth information includes a plurality of domains and sixth information corresponding to each domain. The sixth information includes a plurality of model combinations and seventh information corresponding to each model combination. Each model combination includes a first model and a second model. The seventh information includes a plurality of intervals of the MCS, the third information, and the first parameter, and second parameters corresponding to each interval. The second parameter characterizes the accuracy of the corresponding first model and the second model during model training.

[0174] In one embodiment, the second information at least includes a third parameter, where the third parameter represents the accuracy of channel compression and feedback; correspondingly, the second receiving unit 1102 is further configured to receive eighth information sent by the terminal, where the eighth information is determined by using the third parameter and a fourth parameter, and the fourth parameter represents the accuracy of the first model and the second model during testing; wherein, when the fourth parameter is greater than the third parameter, the eighth information represents that the performance of the first model and the second model meets the requirements of channel compression and feedback; when the fourth parameter is less than or equal to the third parameter, the eighth information represents that the performance of the first model and the second model does not meet the requirements of channel compression and feedback.

[0175] In one embodiment, the second processing unit 1103 is further configured to configure time domain and / or frequency domain resources associated with channel compression and feedback for the terminal based on the eighth information.

[0176] In actual application, the second sending unit 1101 and the second receiving unit 1102 may be implemented by a communication interface in the information transmission device; the second processing unit 1103 may be implemented by a processor in the information transmission device.

[0177] It should be noted that: when the information transmission device provided in the above embodiment performs information transmission, only the above division of each program module is used for illustration. In actual application, the above processing may be allocated to different program modules according to needs, that is, the internal structure of the device is divided into different program modules to complete all or part of the above-described processing. In addition, the information transmission device provided in the above embodiment and the information transmission method embodiment belong to the same concept, and the specific implementation process thereof can be seen in the method embodiment, which will not be elaborated here.

[0178] Based on the hardware implementation of the above program module, and in order to implement the method on the terminal side of the embodiments of the present application, the embodiments of the present application further provide a terminal, as Figure 12 shown, the terminal 1200 includes:

[0179] A first communication interface 1201 capable of performing information interaction with a network device and / or other terminals;

[0180] A first processor 1202, connected to the first communication interface 1201 to implement information interaction with a network device and / or other terminals, and configured to execute the method provided by one or more of the above technical solutions on the terminal side when running a computer program;

[0181] A first memory 1203, where the computer program is stored on the first memory 1203.

[0182] Specifically, the first processor 1202 is configured to:

[0183] Receive a first piece of information and a second piece of information sent by a network device through the first communication interface 1201. The first piece of information is used to indicate a first model and a second model. The first model is used for channel compression, and the second model is used for channel decompression. The second piece of information is used to indicate the requirements for channel compression and feedback.

[0184] Test the performance of the first model and the second model. When the performance of the first model and the second model meets the requirements for channel compression and feedback, use the first model to perform channel compression to obtain a plurality of output elements. Determine a fourth piece of information based on the plurality of output elements, the MCS corresponding to the first model, and a third piece of information, and send the fourth piece of information to the network device through the first communication interface 1201. Wherein,

[0185] The total number of constellation points corresponding to the MCS is the same as the total number of the plurality of output elements. The third piece of information characterizes the association relationship between the output elements of the first model and the constellation points corresponding to the MCS. The fourth piece of information includes the constellation point corresponding to each output element in the plurality of output elements.

[0186] Wherein, in one embodiment, the first processor 1202 is further configured to determine the MCS and the third piece of information using a fifth piece of information. The fifth piece of information includes a plurality of domains and a sixth piece of information corresponding to each domain. The sixth piece of information includes a plurality of model combinations and a seventh piece of information corresponding to each model combination. Each model combination includes a first model and a second model. The seventh piece of information includes the MCS, the third piece of information, a plurality of intervals of a first parameter, and a second parameter corresponding to each interval. The first parameter characterizes the channel sparsity, and the channel sparsity can reflect the amount of information contained in the channel matrix. The second parameter characterizes the accuracy of the corresponding first model and the second model during model training.

[0187] In one embodiment, the second piece of information at least includes a third parameter, and the third parameter characterizes the accuracy of channel compression and feedback. Correspondingly, the first processor 1202 is further configured to:

[0188] Test the accuracy of the first model and the second model to obtain a fourth parameter, and the fourth parameter characterizes the accuracy of the first model and the second model during testing.

[0189] Determine the eighth information by using the third parameter and the fourth parameter; wherein, when the fourth parameter is greater than the third parameter, the eighth information represents that the performance of the first model and the second model meets the requirements of channel compression and feedback; when the fourth parameter is less than or equal to the third parameter, the eighth information represents that the performance of the first model and the second model does not meet the requirements of channel compression and feedback.

[0190] In one embodiment, the first communication interface 1201 is further configured to send the eighth information to the network device.

[0191] It should be noted that the specific processing procedures of the first processor 1202 and the first communication interface 1201 can be understood with reference to the above method, and will not be elaborated here.

[0192] Of course, in actual application, each component in the terminal 1200 is coupled together through the bus system 1204. It can be understood that the bus system 1204 is used to realize the connection and communication between these components. In addition to the data bus, the bus system 1204 also includes a power bus, a control bus, and a status signal bus. However, for the sake of clarity, in Figure 12 all kinds of buses are labeled as the bus system 1204.

[0193] The first memory 1203 in the embodiment of the present application is used to store various types of data to support the operation of the terminal 1200. Examples of these data include: any computer program for operating on the terminal 1200.

[0194] The method disclosed in the embodiments of the present application above can be applied to or implemented by the first processor 1202. The first processor 1202 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit in hardware or instructions in software form in the first processor 1202. The first processor 1202 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The first processor 1202 can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present application. The general-purpose processor may be a microprocessor or any conventional processor, etc. Combining the steps of the method disclosed in the embodiments of the present application, it can be directly embodied as being executed and completed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software module may be located in a storage medium, and this storage medium is located in the first memory 1203. The first processor 1202 reads the information in the first memory 1203 and combines its hardware to complete the steps of the foregoing method.

[0195] In an exemplary embodiment, the terminal 1200 may be implemented by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontroller units (MCUs), microprocessors, or other electronic components for executing the foregoing method.

[0196] Based on the hardware implementation of the above program module, and in order to implement the method on the network device side in the embodiments of the present application, the embodiments of the present application further provide a network device, as Figure 13 shown. The network device 1300 includes:

[0197] A second communication interface 1301 capable of information interaction with the terminal and / or other network devices;

[0198] A second processor 1302, connected to the second communication interface 1301 to enable information interaction with the terminal and / or other network devices, and when running a computer program, executes the method provided by one or more of the above technical solutions on the network device side;

[0199] A second memory 1303, on which the computer program is stored.

[0200] Specifically, the second communication interface 1301 is used for:

[0201] Sending a first information and a second information to the terminal, the first information being used to indicate a first model and a second model, the first model being used for channel compression, the second model being used for channel decompression, and the second information being used to indicate the requirements for channel compression and feedback;

[0202] Receiving a fourth information sent by the terminal when the performance of the first model and the second model meets the requirements for channel compression and feedback; wherein,

[0203] The fourth information is determined by using a plurality of output elements, the MCS corresponding to the first model, and a third information. The plurality of output elements are obtained by using the first model for channel compression. The total number of constellation points corresponding to the MCS is the same as the total number of the plurality of output elements. The third information characterizes the association relationship between the output elements of the first model and the constellation points corresponding to the MCS. The fourth information includes the constellation points corresponding to each of the plurality of output elements.

[0204] Wherein, in one embodiment, the second processor 1302 is used for:

[0205] Determining a ninth information and a first parameter, the ninth information characterizing the channel quality, the first parameter characterizing the channel sparsity, and the channel sparsity can reflect the amount of information contained in the channel matrix;

[0206] Using a fifth information, the ninth information, the first parameter, and the second information to determine the first information; the fifth information includes a plurality of domains and a sixth information corresponding to each domain. The sixth information includes a plurality of model combinations and a seventh information corresponding to each model combination. Each model combination includes a first model and a second model. The seventh information includes a plurality of intervals of MCS, third information, and first parameter, and a second parameter corresponding to each interval. The second parameter characterizes the accuracy of the corresponding first model and the second model during model training.

[0207] In one embodiment, the second information at least includes a third parameter, where the third parameter represents the accuracy of channel compression and feedback; correspondingly, the second communication interface 1301 is further configured to receive eighth information sent by the terminal, where the eighth information is determined by using the third parameter and a fourth parameter, and the fourth parameter represents the accuracy of the first model and the second model during testing; where, when the fourth parameter is greater than the third parameter, the eighth information represents that the performance of the first model and the second model meets the requirements of channel compression and feedback; when the fourth parameter is less than or equal to the third parameter, the eighth information represents that the performance of the first model and the second model does not meet the requirements of channel compression and feedback.

[0208] In one embodiment, the second processor 1302 is further configured to, based on the eighth information, configure time domain and / or frequency domain resources associated with channel compression and feedback for the terminal.

[0209] It should be noted that: The specific processing procedures of the second communication interface 1301 and the second processor 1302 can be understood with reference to the above method, and will not be elaborated here.

[0210] Of course, in actual application, each component in the network device 1300 is coupled together through a bus system 1304. It can be understood that the bus system 1304 is used to implement connection communication between these components. The bus system 1304 includes, in addition to a data bus, a power bus, a control bus, and a status signal bus. However, for the sake of clear illustration, in Figure 13 all kinds of buses are labeled as the bus system 1304.

[0211] The second memory 1303 in the embodiment of the present application is used to store various types of data to support the operation of the network device 1300. Examples of these data include: any computer program for operating on the network device 1300.

[0212] The method disclosed in the embodiments of the present application can be applied to or implemented by the second processor 1302. The second processor 1302 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in the second processor 1302 or instructions in software form. The second processor 1302 may be a general-purpose processor, DSP, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The second processor 1302 can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present application. The general-purpose processor may be a microprocessor or any conventional processor, etc. Combining the steps of the method disclosed in the embodiments of the present application can be directly embodied as being executed and completed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software module may be located in a storage medium, and this storage medium is located in the second memory 1303. The second processor 1302 reads the information in the second memory 1303 and combines its hardware to complete the steps of the foregoing method.

[0213] In an exemplary embodiment, the network device 1300 may be implemented by one or more ASICs, DSPs, PLDs, CPLDs, FPGAs, general-purpose processors, controllers, MCUs, Microprocessors, or other electronic components for executing the foregoing method.

[0214] It can be understood that the memories (the first memory 1203 and the second memory 1303) in the embodiments of the present application can be volatile memories or non-volatile memories, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM, Read Only Memory), a programmable read-only memory (PROM, Programmable Read-Only Memory), an erasable programmable read-only memory (EPROM, Erasable Programmable Read-Only Memory), an electrically erasable programmable read-only memory (EEPROM, Electrically Erasable Programmable Read-Only Memory), a ferromagnetic random access memory (FRAM, ferromagnetic random access memory), a flash memory (Flash Memory), a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM, Compact Disc Read-Only Memory); the magnetic surface memory can be a disk memory or a tape memory. The volatile memory can be a random access memory (RAM, Random Access Memory), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as a static random access memory (SRAM, Static Random Access Memory), a synchronous static random access memory (SSRAM, Synchronous Static Random Access Memory), a dynamic random access memory (DRAM, Dynamic Random Access Memory), a synchronous dynamic random access memory (SDRAM, Synchronous Dynamic Random Access Memory), a double data rate synchronous dynamic random access memory (DDR SDRAM, Double Data Rate Synchronous Dynamic Random Access Memory), an enhanced synchronous dynamic random access memory (ESDRAM, Enhanced Synchronous Dynamic Random Access Memory), a sync link dynamic random access memory (SLDRAM, SyncLink Dynamic Random Access Memory), and a direct rambus random access memory (DRRAM, Direct Rambus Random Access Memory).The memories described in the embodiments of the present application are intended to include, but are not limited to, these and any other suitable types of memories.

[0215] To implement the method provided by the embodiments of the present application, the embodiments of the present application further provide an information transmission system, as Figure 14 shown. The system includes: a terminal 1401 and a network device 1402.

[0216] Here, it should be noted that: the specific processing procedures of the terminal 1401 and the network device 1402 have been described in detail above and will not be elaborated here.

[0217] In an exemplary embodiment, the embodiments of the present application further provide a storage medium, namely a computer storage medium, specifically a computer-readable storage medium. For example, it includes a first memory 1203 storing a computer program, and the above computer program can be executed by a first processor 1202 of the terminal 1200 to complete the steps of the foregoing method on the terminal side. Another example is a second memory 1303 storing a computer program, and the above computer program can be executed by a second processor 1302 of the network device 1300 to complete the steps of the foregoing method on the network device side. The computer-readable storage medium can be a FRAM, ROM, PROM, EPROM, EEPROM, Flash Memory, magnetic surface memory, optical disc, or CD-ROM, etc.

[0218] It should be noted that: "first", "second", etc. are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence.

[0219] In addition, the technical solutions described in the embodiments of the present application can be arbitrarily combined without conflict.

[0220] The above is only a preferred embodiment of the present application and is not intended to limit the protection scope of the present application.

Claims

1. An information transmission method, characterized in that, applied to a terminal, comprising: receiving a first piece of information and a second piece of information sent by a network device, where the first piece of information is used to indicate a first model and a second model, the first model is used for channel compression, the second model is used for channel decompression, and the second piece of information is used to indicate the requirements for channel compression and feedback; testing the performance of the first model and the second model, and when the performance of the first model and the second model meets the requirements for channel compression and feedback, using the first model to perform channel compression to obtain a plurality of output elements; determining a fourth piece of information based on the plurality of output elements, the modulation and coding scheme MCS corresponding to the first model, and a third piece of information, and sending the fourth piece of information to the network device; wherein, the total number of constellation points corresponding to the MCS is the same as the total number of the plurality of output elements, the third piece of information characterizes the association relationship between the output elements of the first model and the constellation points corresponding to the MCS, and the fourth piece of information includes the constellation point corresponding to each output element in the plurality of output elements.

2. The method according to claim 1, characterized in that, the method further comprises: determining the MCS and the third piece of information using a fifth piece of information, the fifth piece of information includes a plurality of domains and sixth pieces of information corresponding to each domain, the sixth piece of information includes a plurality of model combinations and seventh pieces of information corresponding to each model combination, each model combination includes a first model and a second model, and the seventh piece of information includes an MCS, a third piece of information, a plurality of intervals of a first parameter, and a second parameter corresponding to each interval; the first parameter characterizes the channel sparsity, and the second parameter characterizes the accuracy of the corresponding first model and second model during model training.

3. The method according to claim 1 or 2, characterized in that, the second piece of information at least includes a third parameter, and the third parameter characterizes the accuracy of channel compression and feedback; the testing the performance of the first model and the second model includes: testing the accuracy of the first model and the second model to obtain a fourth parameter, and the fourth parameter characterizes the accuracy of the first model and the second model during testing; determining an eighth piece of information using the third parameter and the fourth parameter; wherein, when the fourth parameter is greater than the third parameter, the eighth piece of information characterizes that the performance of the first model and the second model meets the requirements for channel compression and feedback; when the fourth parameter is less than or equal to the third parameter, the eighth piece of information characterizes that the performance of the first model and the second model does not meet the requirements for channel compression and feedback.

4. The method according to claim 3, characterized in that, the method further comprises: sending the eighth piece of information to the network device.

5. An information transmission method, characterized in that, applied to a network device, comprising: Send a first piece of information and a second piece of information to a terminal, where the first piece of information is used to indicate a first model and a second model, the first model is used for channel compression, the second model is used for channel decompression, and the second piece of information is used to indicate the requirements for channel compression and feedback; When the performance of the first model and the second model meets the requirements for channel compression and feedback, receive a fourth piece of information sent by the terminal; where The fourth piece of information is determined using a plurality of output elements, the MCS corresponding to the first model, and a third piece of information. The plurality of output elements are obtained by performing channel compression using the first model. The total number of constellation points corresponding to the MCS is the same as the total number of the plurality of output elements. The third piece of information characterizes the association relationship between the output elements of the first model and the constellation points corresponding to the MCS. The fourth piece of information includes the constellation points corresponding to each of the plurality of output elements.

6. The method according to claim 5, wherein, the method further includes: determine a ninth piece of information and a first parameter, where the ninth piece of information characterizes channel quality and the first parameter characterizes channel sparsity; determine the first piece of information using a fifth piece of information, the ninth piece of information, the first parameter, and the second piece of information. The fifth piece of information includes a plurality of domains and a sixth piece of information corresponding to each domain. The sixth piece of information includes a plurality of model combinations and a seventh piece of information corresponding to each model combination. Each model combination includes a first model and a second model. The seventh piece of information includes multiple intervals of MCS, the third information, and the first parameter, and a second parameter corresponding to each interval. The second parameter characterizes the accuracy of the corresponding first model and second model during model training.

7. The method according to claim 5 or 6, wherein, the second piece of information at least includes a third parameter, where the third parameter characterizes the accuracy of channel compression and feedback; the method further includes: receive an eighth piece of information sent by the terminal, where the eighth piece of information is determined using the third parameter and a fourth parameter, and the fourth parameter characterizes the accuracy of the first model and the second model during testing; where, when the fourth parameter is greater than the third parameter, the eighth piece of information characterizes that the performance of the first model and the second model meets the requirements for channel compression and feedback; when the fourth parameter is less than or equal to the third parameter, the eighth piece of information characterizes that the performance of the first model and the second model does not meet the requirements for channel compression and feedback.

8. The method according to claim 7, wherein, the method further includes: Based on the eighth piece of information, configure time domain and / or frequency domain resources associated with channel compression and feedback for the terminal.

9. An information transmission device, wherein, it includes: A first receiving unit, configured to receive a first piece of information and a second piece of information sent by a network device, where the first piece of information is used to indicate a first model and a second model, the first model is used for channel compression, the second model is used for channel decompression, and the second piece of information is used to indicate the requirements for channel compression and feedback; A first processing unit, configured to test the performance of the first model and the second model, and in the case where the performance of the first model and the second model meets the requirements of channel compression and feedback, perform channel compression using the first model to obtain a plurality of output elements; Determine fourth information by using the plurality of output elements, the MCS corresponding to the first model, and third information; A first sending unit, configured to send the fourth information to the network device; wherein, The total number of constellation points corresponding to the MCS is the same as the total number of the plurality of output elements, the third information represents the association relationship between the output elements of the first model and the constellation points corresponding to the MCS, and the fourth information includes the constellation points corresponding to each of the plurality of output elements.

10. An information transmission device, Characterized in that, It includes: A second sending unit, configured to send first information and second information to a terminal, where the first information is used to indicate a first model and a second model, the first model is used for channel compression, the second model is used for channel decompression, and the second information is used to indicate the requirements of channel compression and feedback; A second receiving unit, configured to receive the fourth information sent by the terminal in the case where the performance of the first model and the second model meets the requirements of channel compression and feedback; wherein, The fourth information is determined by using a plurality of output elements, the MCS corresponding to the first model, and third information, the plurality of output elements are obtained by performing channel compression using the first model, the total number of constellation points corresponding to the MCS is the same as the total number of the plurality of output elements, the third information represents the association relationship between the output elements of the first model and the constellation points corresponding to the MCS, and the fourth information includes the constellation points corresponding to each of the plurality of output elements.

11. A terminal, Characterized in that, It includes: A first communication interface and a first processor; wherein, The first processor is configured to: Receive the first information and the second information sent by the network device through the first communication interface, where the first information is used to indicate a first model and a second model, the first model is used for channel compression, the second model is used for channel decompression, and the second information is used to indicate the requirements of channel compression and feedback; Test the performance of the first model and the second model, and in the case where the performance of the first model and the second model meets the requirements of channel compression and feedback, perform channel compression using the first model to obtain a plurality of output elements; determine fourth information by using the plurality of output elements, the MCS corresponding to the first model, and third information, and send the fourth information to the network device through the first communication interface; wherein, The total number of constellation points corresponding to the MCS is the same as the total number of the plurality of output elements, the third information represents the association relationship between the output elements of the first model and the constellation points corresponding to the MCS, and the fourth information includes the constellation points corresponding to each of the plurality of output elements.

12. A network device, Characterized in that, It includes: A second communication interface and a second processor; wherein, The second communication interface is configured to: Send a first message and a second message to the terminal, the first message being used to indicate a first model and a second model, the first model being used for channel compression, the second model being used for channel decompression, and the second message being used to indicate the requirements for channel compression and feedback; Receive a fourth message sent by the terminal when the performance of the first model and the second model meets the requirements for channel compression and feedback; wherein, The fourth message is determined by using a plurality of output elements, the MCS corresponding to the first model, and a third message. The plurality of output elements are obtained by performing channel compression using the first model. The total number of constellation points corresponding to the MCS is the same as the total number of the plurality of output elements. The third message characterizes the association relationship between the output elements of the first model and the constellation points corresponding to the MCS. The fourth message includes the constellation points corresponding to each of the plurality of output elements.

13. A terminal, Characterized in that, It includes: A first processor and a first memory for storing a computer program that can run on the processor, Wherein, when the first processor is used to run the computer program, it executes the steps of the method according to any one of claims 1 to 4.

14. A network device, Characterized in that, It includes: A second processor and a second memory for storing a computer program that can run on the processor, Wherein, when the second processor is used to run the computer program, it executes the steps of the method according to any one of claims 5 to 8.

15. A storage medium, on which a computer program is stored, Characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4, or implements the steps of the method according to any one of claims 5 to 8.

Images